The MSCI Emerging Markets Index returned 33.6% in 2025, ahead of the S&P 500 at 17.9% and the MSCI World at 21.1%. EM ex-China did slightly better at 34.6%.

The index has carried that momentum into this year and is up ~25.6% YTD. Pretty much every strategist desk from JP Morgan to Lazard to State Street has been publishing constructive notes on EM since January.

Typically, when we see consensus like this, the chart has already played out and the easy money is often behind us, not in front of us.

The truth is that when we look deeper into the EM indexes, most people who say “I am bullish on emerging markets” do not know what they are actually buying. The MSCI EM Index is now ~40% concentrated in its top ten names, and TSMC alone is more than 14% of it. When we add the rest of the Taiwanese and Korean semiconductor complex, the EM index is in large part, a leveraged bet on global AI capital expenditure.

Subscribe now

The Other Side

However, the headline EM number hides a much more complicated story underneath.

On one side sits North Asian semiconductors and memory names like TSMC, SK Hynix and Samsung Electronics that have run incredibly hard on the AI build out. The forward returns for these businesses are now hostage to the same CAPEX revisions that govern Nvidia, Micron, Broadcom and others.

On the other side sits the structural EM story, which I believe has legs and a massive secular tailwind in the decades to come. These are typically businesses that operate in truly developing economies, the likes of Vietnam, Thailand, Indonesia, Brazil and Argentina. These businesses are levered to local consumption, financial inclusion, and demographics.

While the semis half of the story re-rated higher, several of the best domestic-demand compounders went the other way. They de-rated hard, through late 2025 and the first half of 2026. For instance:

• Sea Limited has fallen from a high of $199 in October 2025 to ~$85 today, marking a ~60% drawdown despite posting three of its best quarters since.

• Mercado Libre is down around 22% year to date and trades below its own 52-week low set last year, despite posting 49% revenue growth, its fastest since Q2 2022.

• Grab sits near its 52-week low around $3.50, down roughly 28% over six months, while revenue grew 24% and adjusted EBITDA grew 46%.

• Nubank is down over 26% YTD, after a quarter with record revenues of $5B (up 53% YoY) and $871M in net income.

• dLocal is down 8% YTD despite accelerating TPV growth and its re-investment cycle coming to end.

Why I Believe The Other Half Will Re-Rate

There are four main reasons behind this.

1. The Iran war ending, and oil normalising with it.

The war that began on 28 February choked the Strait of Hormuz, through which roughly a fifth of the world's oil flows, and drove Brent toward $112 with Dated Brent spiking above $140.

Just a few days ago, the US and Iran reached a deal to end the war and reopen the strait, with formal signing set for 19 June in Switzerland. Crude is already back near $92, down around 20% from its 2026 peak.

Southeast Asia is a bloc of net oil importers, so the single largest macro pain of the last four months, imported inflation forcing local central banks to stay hawkish, is now reversing.

As the strait reopens and oil grinds back toward its pre-war equilibrium in the high-$50s to $60, that will be dis-inflationary for Indonesia, the Philippines, Thailand and Vietnam. This also gives central banks in the region the flexibility and option to cut, which feeds consumer spending and lowers funding costs for the digital lenders.

2. The earnings gap.

Consensus expects EM earnings growth in the high teens to around 20% for 2026, against roughly 13% to 15% for the US and developed markets.

However, the businesses I’m highlighting today are growing revenue at two to three times that index pace. Sea grew revenue 47%, Mercado Libre 49%, Grab 24%, Nubank 53% and dLocal 55%. The market can ignore the fundamentals for a year or two, but not forever.

What is the catalyst? I believe we are seeing a return-on-equity convergence, and that is crucial because it makes a re-rating permanent rather than a short-term trade. Return on equity is simply how much profit a company generates for every dollar of shareholder capital it holds, the cleanest single measure of how well a business compounds your money.

EM companies have historically earned a lower return on equity than their developed-market peers, and that gap is now narrowing as EM returns rise toward developed-market levels. One of the clearest signs of this playing out is that buybacks are becoming standard across these exact names.

Sea Limited:

• $1 billion buyback programme announced on 17th Nov 2025

• $168.4M exercised at average price of $93.55 per share as of Q1 2026

Nubank:

• $1 billion buyback programme announced on 4th June 2026

• Programme runs for 12 months from 4th June 2026 to 3rd June 2027

dLocal:

• $300 million buyback programme announced on 18th March 2026

• Permanent dividend programme, paying out 30% of FCF each year

Grab:

• $500 million buyback programme announced in February this year

• $400 million execution over 4 months ($250M ASR with JPM + $150M CFP with MS)

Governance and capital discipline are improving. I believe this will lead to a permanent re-pricing of high quality emerging market businesses.

Subscribe now

3. The rotation out of the AI trade.

The AI trade has been the story of much of 2025 and all of 2026. Most of the deals have been led by the hyperscalers who have poured in billions in CAPEX and are expected to continue. However, there will be a physical limit to this eventually, as CAPEX guidance is beginning to run ahead of operating cash flow and free cash flow is heading negative. CAPEX revisions cannot keep ratcheting up indefinitely.

The moment those revisions stall, the optionality premium attached to anything AI-adjacent deflates, and a very large pool of capital will have to start looking for a new home.

The 2025 and 2026 YTD EM rally was substantially the semiconductor and AI complex. Globally, quality compounders have seen huge outflows and underperformance. I believe this will reverse soon, and benefit from a massive re-rating.

4. The dollar is the single largest driver of EM returns.

Emerging market debt is substantially dollar-denominated, commodities are priced in dollars, and for a foreign holder the entire return is translated back through the exchange rate.

When the dollar rises, EM is paying down hard-currency debt with depreciating local currency, importing inflation, and handing the foreign investor a currency loss on top of everything else.

When you look at the chart above, there is a near perfect inverse correlation between DXY (in candles) and EM (in blue). Cheap valuations did nothing for EM for over a decade (2011 to 2020) because the dollar was against the asset class the entire time.

The key catalyst is the dollar turning, and it has turned. There have indeed been several false dawns in the past decade. However, this time, there are 3 independent forces pushing the same way:

1. Cyclically, the Fed is easing, which narrows the rate differential that supported the dollar.

2. Structurally, the world is de-dollarising: the dollar's share of global reserves has fallen from about 71% in 1999 to around 56 to 57% by 2025. Emerging-market central banks cut their dollar holdings from 68% to 52% between 2013 and 2024.

3. Politically, a weaker dollar is the explicit aim of US policy under this administration, which views dollar strength as an obstacle to its trade agenda and is leaning on the Fed through a dovish chair willing to prioritise jobs over the inflation mandate.

Conclusion

I believe there are really two emerging-market trades today.

The first is the trade everyone already owns: Taiwan, Korea, semiconductors, memory, AI capex, and everything downstream of that cycle. That trade has worked incredibly well, but it is now crowded, consensus, and increasingly dependent on the next round of AI CAPEX revisions.

The second is the trade people are currently ignoring. High-quality domestic-demand compounders in genuinely developing markets.

There are massive secular tailwinds and structural growth in EM that reside in domestic demand. The best domestic-demand businesses have been left behind because they don’t fit the current narrative.

I can’t promise that these names will re-rate next quarter, but there are very real catalysts and reasons for them to perform incredibly well in the next few years. Assets typically bottom when nobody is buying and it feels dead. I believe that time is now.

Thank you for reading!

-Gab

Disclaimer: The content presented in this thesis is for informational and academic purposes only and does not constitute financial advice. The analysis and opinions expressed are based on research and should not be interpreted as a recommendation to buy, sell, or hold any security. Readers should conduct their own due diligence and consult with a qualified financial advisor before making any investment decisions.

I think the next phase will increasingly move into the physical world.

As AI moves from pure software into real-world applications, a different set of companies should start to matter.

These are not necessarily the companies building the models themselves, but the ones providing the infrastructure, tools, and enabling layers that allow AI to be deployed outside controlled digital environments.

If the next decade of AI is increasingly about connecting intelligence to the physical world, I think this company gives me an interesting way to express that view.

I’ve raised 16% cash in the past month to deploy opportunistically in certain opportunities, and I believe this may be one of the more important thematic bets in my portfolio.

In this article, I’ll cover the business and share why I am starting a position now.

Subscribe now

Read more

As many of you know, I’ve been covering Sea Limited for a long time and have built up a fairly deep understanding of the business over the years.

Along the way, I’ve maintained various documents tracking different parts of the company: Shopee, Garena, Monee, financials, market share data, segment KPIs, and broader industry developments.

Over the past month, I decided to consolidate all of that into one comprehensive spreadsheet, or more specifically, a Google Sheets dashboard.

This dashboard contains 15 separate tabs, covering everything from the income statement and balance sheet to more granular operating data across Shopee, Garena, and Monee. It also includes what I personally find to be some of the most useful data points for understanding Sea: e-commerce market share, food delivery market share, and other segment-level KPIs that help track the real progress of the business.

The dataset goes all the way back to 2015, two years before Sea went public, and includes both quarterly and annual data where available. I’ve also tried to include information that is not typically found on financial aggregator websites, but is instead buried in 20-F filings, earnings calls, investor presentations, and other company disclosures.

The goal of this dashboard is to create a single place where investors can track Sea’s business evolution properly, moving beyond just the headline numbers and looking at the actual drivers of the business.

Sea is not an easy company to analyse from surface-level numbers alone. The group has multiple moving pieces across e-commerce, gaming, digital financial services, logistics, advertising, credit, and food delivery.

For example, Shopee’s reported revenue only tells part of the story. To really understand the business, we need to track GMV, take rate, order growth, monetisation, logistics strategy, advertising revenue, competitive intensity, and market share across Southeast Asia, Brazil, and other markets.

Monee is also becoming increasingly important. As the credit business scales, investors need to pay attention not only to revenue and EBITDA growth, but also loan book size, non-performing loans, user penetration, and how Sea is balancing growth with credit risk.

This will also be a working document. I plan to update it every quarter as Sea reports earnings and as more data becomes available. Over time, I expect the dashboard to become more useful as the historical dataset deepens and more segment-level trends become visible.

I’m also open to feedback. If there are additional metrics, charts, or datasets that would make the dashboard more useful, I’m happy to consider adding them.

For now, I hope this serves as a helpful resource for anyone trying to understand Sea Limited at a deeper level.

Read more

Today, we’re diving into the largest IPO in history. On 20 May 2026, Space Exploration Technologies Corp. (or SpaceX for short) filed its S-1 with the SEC, formally kicking off the road to what is expected to be a 12 June 2026 listing on NASDAQ under the ticker $SPCX.

Reuters and WSJ are pegging the targeted valuation at $1.75T to $2T, with an intended raise of approximately $75B.

This is a massive raise, and to put it in context:

• It would be ~3x the size of Saudi Aramco’s 2019 IPO, the previous record

• SpaceX would instantly be one of the top 8 most valuable companies in the world

• The synthetic SPCX-USD contract on Hyperliquid is implying a $2.2T valuation

SpaceX reported $18.7B in revenue in 2025, with the proposed $1.75T valuation implying 94x P/S.

The counter-argument would be that SpaceX is three massive businesses, with huge addressable markets. In this piece, we will work through the growth prospects of each segment, determine what the moats are, whether they are truly differentiated and defendable, and most importantly, determine a fair value for the business.

Table of Contents

1. Introduction

2. Company History

3. Business Model

4. Value Proposition

5. Product Offerings

6. Moats & Differentiation

7. Market Context & Industry Positioning

8. Competitive Landscape

9. Financials

10. Ownership & Management

11. Valuation

12. Bull and Bear Case

13. Concluding Thoughts (What I am personally doing)

1. Introduction

Most of us know the founding story of SpaceX, about how Elon tried to purchase ICBMs from the Russians. But for the benefit of those who don’t, here’s a very short backstory.

If you tried to ship something into orbit in 2002, you had two real options, pay Boeing or Lockheed in the US, or pay the Russians. The price was about $18,500 per kilogram to LEO (Low Earth Orbit). The industry had not had a serious cost innovation in 30 years, and the rockets were expendable.

Basically after each launch, the rocket would be disposed of. Imagine flying for a holiday and the plane you were travelling in gets disposed of immediately after. That was the launch industry barely 20 years ago.

SpaceX broke that and set off a new space race that we haven’t seen since the Cold War in the 60s. SpaceX’s Falcon 9 in 2010 reduced the cost to orbit by 85%, dropping payload prices to $2,700 per kg. With Falcon Heavy in 2018, it fell by another half to $1,400 per kg. (The numbers above show the latest vintage cost per kg by year, not the first-version cost, and with reuse maturing + economies of scale, costs have come down massively.)

Starship is SpaceX’s latest launch vehicle that is designed to be fully reusable and is described as a super heavy-life launch vehicle. In documents from Voyager, we found that SpaceX charged just $90M for usage of Starship, which can reportedly support a 150-ton payload. At those numbers, it would equate to a cost per kg of $600.

In about 2 decades, SpaceX has reduced the cost to orbit by 97%. Almost every other business SpaceX has built since exists because of this launch cost collapse:

• Starlink would have been economically impossible without Falcon 9 reusability. Even today, 73% of SpaceX’s Falcon 9 launches in 2025 were dedicated to deploying its own satellites (165 launches, only 43 for external customers).

• Starlink Mobile (direct-to-cell) sits on top of Starlink.

• Orbital AI compute, which is the headline ambition in the S-1, is only economically viable if Starship works at scale and reduces launch cost by another 99%.

• The lunar economy and Mars are entirely dependent on Starship.

Hence, when we evaluate SpaceX as a business, we are basically evaluating a stack with Starship at the bottom, Starlink in the middle, AI compute at the top with X, Grok, Macrohard, Cursor and Terafab layered on.

2. Company History

SpaceX was founded in March 2002 by Elon Musk, who put in $100M of his PayPal exit proceeds. The original founding team also included Tom Mueller (propulsion), Gwynne Shotwell (who joined in 2002 as VP of Business Development and is now President & COO), and a small group of engineers Musk recruited largely from TRW, Boeing, and JPL.

The early years were brutal, as to be expected for a business attempting to do the impossible. The first launch vehicle, Falcon 1 failed three times between 2006 and 2008, and SpaceX was reportedly weeks from insolvency before the fourth attempt succeeded in September 2008.

The same year, NASA awarded SpaceX a $1.6B Commercial Resupply Services contract, which kept the company alive.

What followed was a 17-year run of step-function milestones.

Timeline Highlights:

• 2002: SpaceX founded by Elon Musk in El Segundo, California

• 2006: First Falcon 1 launch (failure)

• 2008: Falcon 1 reaches orbit (first private liquid-fuel rocket to do so); NASA awards $1.6B CRS contract

• 2010: Falcon 9 first flight: Dragon becomes first private spacecraft recovered from orbit

• 2012: Dragon docks with the ISS, first private spacecraft to do so

• 2015: First successful Falcon 9 booster propulsive landing, “the impossible thing”

• 2017: First re-flight of a flight-proven Falcon 9 booster

• 2018: Falcon Heavy maiden flight, putting the Tesla Roadster into solar orbit

• 2019: First Starlink launch (60 satellites)

• 2020: Crew Dragon Demo-2: first private company to fly humans to orbit

• 2023: First Starship integrated flight test; SpaceX launches >80% of global mass-to-orbit

• 2024: First successful “chopstick” booster catch at Starbase

• March 2025: xAI acquires X (formerly Twitter), valuing the combined entity at ~$113B

• July 2025: SpaceX invests $2B into xAI as part of a $5B equity round

• December 2025: SpaceX completes secondary at ~$800B valuation (per Bloomberg)

• Feb 2, 2026: SpaceX acquires xAI in all-stock deal: xAI now SpaceX’s AI segment

• March 2026: Terafab announced with Tesla (Intel joins April 2026); SpaceX Bridge Loan of $20B signed with Goldman

• April 2026: Cursor compute & option agreement signed (implied $60B equity value)

• May 12, 2026: FCC approves $17B EchoStar spectrum acquisition for V2 Mobile

• May 2026: Anthropic Cloud Services Agreement signed: $1.25B/month through May 2029 (~$45B total)

• May 20, 2026: S-1 filed

The two most important recent events on this timeline are the xAI merger and the Anthropic deal, both of which I will keep coming back to.

3. Business Model

SpaceX now reports three operating segments. The simplest way to think about the business is as three businesses stacked vertically, with the Space segment at the foundation enabling the rest.

Space (22% of 2025 revenue, $4.086B)

This is the legacy launch business, of which the launch vehicles are Falcon 9, Falcon Heavy, and (soon) Starship. Dragon is the reusable capsule that sits on top of a rocket and holds up to 8 people and thousands of kilos of cargo. The Space segment sells launch services to commercial customers, the U.S. government (~20% of total SpaceX revenue), and other governments.

The key thing to understand here is that the Space segment’s reported revenue dramatically understates its strategic value. Of 165 Falcon 9 launches in 2025, only 43 were for external customers.

The other 122 launched Starlink satellites. Those internal launches don’t generate Space segment revenue, as they get capitalised into the Connectivity segment (discussed later). Space, in other words, is partly a captive launch vehicle for Starlink and for orbital AI compute in the future. Reported Space revenue is essentially what is left over after SpaceX feeds itself.

Connectivity (61% of 2025 revenue, $11.387B)

This refers to Starlink, the main revenue and profit driver for the business today. Starlink is a satellite internet constellation that provides global mobile broadband and the nascent Starlink Mobile direct-to-cell business.

As of March 31, 2026, Starlink has the following traction metrics:

• ~9,600 broadband and mobile satellites in LEO (≈75% of all active maneuverable satellites in orbit globally)

• 10.3M Starlink Subscribers across 164 countries (up 105% YoY from 5.0M)

• 7.4M monthly unique devices on Starlink Mobile across 30 countries

• Median peak download speed: 225 Mbps; median latency: ~25 ms

  • For reference, the typical global fixed broadband speed is ~95 Mbps with 30-50 ms latency while US fixed broadband speed is ~290 Mbps with 15-30 ms latency.

  • However, the target market for Starlink is the rural customer who might be on 25 Mbps in the US, or an Indonesian customer outside Java with no fixed line at all. 225 Mbps / 25 ms is transformational for many.

• ARPU: $66/month (down from $86/month a year earlier as the subscriber mix shifts to lower-priced international tiers and lower-ARPU developing markets)

~60%+ of Connectivity revenue is from consumer subscribers, the rest from enterprise and government. It is key to note that ARPU is declining quite fast, although margins continue to expand because of network density (no marginal cost to serve).

AI (17% of 2025 revenue, $3.201B)

This segment was bolted on in February 2026 via the xAI merger and includes:

• Compute infrastructure: COLOSSUS + COLOSSUS II, ~1 GW of nameplate compute in Memphis, TN

• Grok: xAI’s frontier model (Grok-4 currently deployed, Grok-5 in training)

• X: Social platform with ~550M MAUs and 350M daily posts

• Recent additions: Cursor (via option), Macrohard (co-owned with Tesla), the future Money Product

The key here is probably the Anthropic deal that was announced just a couple weeks ago. Anthropic has been extremely low on compute for several months and have therefore enforced sudden throttling and dynamic 5-hour session limits during peak hours. As such, Anthropic signed a $1.25B per month (once fully ramped) deal through May 2029 with xAI for access to COLOSSUS I. This equates to $15B yearly, which is 80% of SpaceX’s 2025 revenues. It is a massive deal.

That said, there is certainly room for concern. The deal states that either party can cancel the deal with 90-days notice. With the massive ramp up in compute, it is a possibility that Anthropic finds other available compute and decides to cancel the contract prematurely, especially as it would prefer to avoid funding a direct competitor.

This translates to 300+ MW of compute capacity with 220,000+ NVIDIA GPUs that will ramp to full utilisation within this month. At these prices, Anthropic is paying about $50M per MW per year, which stacks against Neoclouds that are charging ~$10-12M per MW per year, a whopping 4x premium.

Basically, Anthropic is paying a major scarcity premium for immediate, large-scale, contiguous AI compute. This may seem irrational, but due to the demand for compute being excessive at the moment, they seemingly have no other option.

This does, however, say something about Grok. Clearly, the demand for compute at Grok has been lackluster hence them being willing to rent out the whole Colossus I cluster to Anthropic.

4. Value Proposition

SpaceX’s pitch is fundamentally about owning the physical stack of the future: launch, satellites, energy, chips, compute, models, and applications, all under one roof.

Firstly, space is the next frontier not just for exploration but also for infrastructure. With Falcon 9 reusability already in production and Starship targeting another 99% cost reduction, things that were uneconomical 5 years ago are becoming economical. The most obvious example is global broadband from LEO. The next examples are direct-to-cell and orbital AI data centers.

Secondly, AI is becoming compute-bound and energy-bound. Reasoning models can be 10-100x more compute-intensive per query than chat. Inference is growing 2x+ faster than training. Meanwhile, US electricity generation has grown at less than 3% annually since 2023 while AI data center demand has roughly tripled. The Sun produces 99.8% of the solar system’s energy and if compute can be put in Sun-synchronous orbit, you will be able to bypass the terrestrial grid bottleneck entirely. Of course, making this happen is a multi-year or even multi-decade challenge.

Thirdly, SpaceX is uniquely positioned to capture this. They already build, launch, and operate the world’s largest satellite constellation, and also operate the largest coherent AI training cluster on Earth. They are building a chip fab (Terafab) with Tesla and Intel, while owning one of the four frontier models (Grok), and a major distribution channel (X). Recently, their purchase of Cursor, which is the most engaged developer workflow also fits seamlessly in this.

5. Product Offerings

A. Launch Vehicles

Falcon 9 (Fully Operational)

• This is the workhorse. Two-stage system with a reusable first stage.

• 23 metric tons to LEO when fully expendable, ~17.5 tons reusable.

• ~620 launches as of March 31, 2026 with >99% mission success.

• Boosters now demonstrated to re-fly 34 times.

Falcon Heavy (Fully Operational)

• Partially reusable super heavy-lift.

• 64 metric tons to LEO. 11 launches with 100% success rate.

• Used for the most demanding missions (NSSL heavy, Europa Clipper, Psyche).

Starship (Active Testing & Development Phase)

• Fully reusable super heavy-lift.

• Designed for 100+ metric tons to LEO in reusable mode (vs. ~150 tons fully expendable).

• 33 Raptor engines on the Super Heavy booster + 6 Raptors on the upper stage.

• This is the biggest and most important bet. Starship is in active testing and has done 12 test flights so far. The most recent was just last week, the first of its V3 iteration that was launched from Starbase.

• It ⁠deployed a clutch of mock satellites and executed a controlled splashdown of the spacecraft in the Indian Ocean. But it failed to achieve a controlled landing ​of the Super Heavy booster, which tumbled into the Gulf of Mexico.

If Starship works as intended (full reusability + airline-like turnaround), SpaceX will be targeting multiple launches per day per pad. Currently, Starship is the largest and most powerful of the heavy-lift rockets, easily towering over competitors.

B. Spacecraft

Dragon

• Sends crew and cargo to LEO and the ISS.

• 78 crew members from 20 countries flown since 2020.

V1 Mobile

• Direct-to-cell satellites (~650 in orbit).

• Provides SMS, light data, and OTT voice today.

V2 Mini

• Current broadband satellites.

V3

• Next-gen broadband, 1 Tbps downlink per satellite, 60 deployable per Starship launch (vs. ~22 V2 Minis per Falcon 9).

• Expected to be deployed in H2 2026.

V2 Mobile

• Next-gen direct-to-cell with full 5G NR-NTN compliance.

• Targeting 2027 deployment on Starship.

• Requires the EchoStar spectrum to close (Nov 2027 expected) and handset manufacturer adoption of new RF front-ends.

C. AI Compute

COLOSSUS I

• 130 MW initial cluster brought online in 122 days, ~100,000 H100s. Currently, Colossus I has over 300 MW of capacity, with 220,000 NVIDIA GPUs, consisting of a mix of H100, H200 and GB200 accelerators.

• This is currently rented out to Anthropic for $1.25B per month.

COLOSSUS II

• Memphis + Southaven, MS. Combined ~870 MW. Initial 110,000 GB200 cluster brought online in 91 days with a follow-on 110,000 GB300 cluster in 64 days.

• Future expansion: +220,000 GB300s and +400 MW. This is currently training Grok-5.

Combined: ~1 GW of coherent compute today, with line of sight to multi-GW.

Terafab

• Chip manufacturing JV with Tesla and Intel.

• xAI is targeting 1 terawatt of compute hardware per year.

• This however, is extremely aspirational and is a far stretch today.

D. Software & Models

Grok

• Grok-1 was launched in Nov 2023 with Grok-4 currently in production. xAI is training Grok-5 using Colossus II.

• While Grok is a capable model, it is still behind the leading frontier labs: OpenAI and Anthropic.

SuperGrok, SuperGrok Heavy, SuperGrok Lite: Consumer Subscriptions

Grok Business, Grok Enterprise, xAI Gov: B2B Subscriptions

Grok API: Developer Access

Imagine: Image/Video Generation

Grok Voice: Real-time speech.

Macrohard: Early-stage AI workflow platform being co-developed with Tesla.

E. Future / Speculative

As with every Elon company, there is a long list of potential businesses in play. These are some that have been listed in the S-1.

• Orbital AI compute (target deployment from 2028)

• Lunar mass driver (electromagnetic launch from the Moon)

• Money Product on X (payments + banking)

• Asteroid mining, in-space manufacturing, lunar/Mars passenger transport

6. Moats & Differentiation

Read more

S&P 500: +5.15%
NASDAQ: +8.43%

May was a strong month for the indexes, not least led by semis and tech, as we have become accustomed to.

The MAG7 was up strongly, AAPL up 15%, TSLA up 14%, MSFT up 10%. Of course, the real strength was in the AI/Semi/Compute trade (AMD +45%, MU +87%, DELL +101%, QCOM +39%). Interestingly, we also saw some rotation to software (!), who would’ve guessed it?

CRWD and FTNT were up 64% each, NOW up 40% and DDOG up 87%. Incredible numbers for stocks that got sold off just a couple months back as the entire market seemingly believed Anthropic was eating all of software up.

However, this did hide a lot of weakness in other sectors of the market. Financials, Consumer Defensive, Energy and Utilities were notably weak. It was by no means a perfect month for the index. However, due to the relatively larger weighting of tech and semis in particular, we of course saw the index continue climbing.

SPY is now up nearly 11% for the year while QQQ is up over 20%. This is of course following a weak March and April where stocks tanked and the indexes were down 7-8% respectively for the year. This is a remarkable turnaround that I cannot claim to have foreseen.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +188.08%

S&P 500: +83.22% (Outperformance: +104.86%)

NASDAQ: +118.23% (Outperformance: +69.85%)

Portfolio Performance (YTD)

Portfolio: -3.08%

S&P 500: +10.73% (Underperformance: -13.81%)

NASDAQ: +16.34% (Underperformance: -19.42%)

Portfolio reviews are exclusively for paid subscribers. If you would like to gain access to my portfolio holdings, deep dives, and tons of other articles, do consider subscribing.

I post several articles per month, ranging from extensive 50-60 pager deep dives on the most interesting businesses I can find, to shorter articles on valuations, market musings and specific verticals/geographies.

In the past month, I’ve shared these articles with subscribers:

Free subscribers get access to posts too, of course, with free earnings reviews and occasional articles.

In the next few segments, I discuss my general thoughts on the market, portfolio companies, performance, portfolio allocation and portfolio changes.

General Thoughts on the Market & Performance

Read more

EPS: $0.70 v $0.77 🔴

Sea Limited reporting earnings on 12th May before the market open.

The stock closed the day up 13%.

Subscribe now

Selected Key Metrics

Total Gross Profit: $3.145B (+40.7% YoY)
Total Net Income: $438.22M (+6.7% YoY)
Total Adjusted EBITDA: $1.034B (+9.3% YoY)
Cash & Cash Equivalents: $11.1B (up from $10.3B in Q1 2025)

E-Commerce (Shopee)

Gross Orders: 4.0B (+29.3% YoY)
GMV: $37.3B (+30.2% YoY)
Revenue: $5.1B (+45.1% YoY)
Adjusted EBITDA: $223.2M (-15.5% YoY)

Digital Financial Services (Monee)

Consumer & SME Loans Outstanding: $9.9B (+71.3% YoY)
Revenue: $1.2B (+57.8% YoY)
Adjusted EBITDA: $275.2M (+14.0% YoY)

Digital Entertainment (Garena)

Bookings: $931.4M (+20.1% YoY)
Revenue: $696.6M (+40.6% YoY)
Adjusted EBITDA: $573.6M (+25.2% YoY)
Quarterly Active Users: 666.5M (+0.7% YoY)
Quarterly Paying Users: 72.6M (+12.4% YoY)
Quarterly Paying User Ratio: 10.9% (vs 9.8% in Q1 2025)

Share Repurchase Program

Repurchased 1.8M shares for an aggregate of $168.4M = $93.55 per share

Sea Limited saw a massive top-line beat, beating estimates by over 10%. The bottom-line missed, but again, this was to be expected as the focus is on investing in logistics and Shopee VIP to maintain its lead. More on this later.

As I’ve discussed often, my base case is >20% growth for the decade to come. Growth has only accelerated since. All 3 segments are growing top-line over 40%, which is absolutely incredible to see. I continue to think this is one of the most asymmetric investments in the market.

Subscribe now

Table of Contents

1. Shopee (E-Commerce)

2. Monee (Digital Financial Services)

3. Garena (Digital Entertainment)

4. Management Commentary

5. Concluding Thoughts

1. Shopee (E-Commerce)

Shopee’s GMV hit a new all-time high this quarter, making it the 8th consecutive quarter of growth. Sequential growth of $700M is especially impressive considering seasonality of Q4 to Q1. In addition, monthly active buyers were up 16% YoY and monthly average purchase frequency grew around 12% YoY. So before even considering average order value, Shopee’s buyer activity implies roughly 30% YoY growth in purchase occasions. This shows habit formation is increasing.

Management reiterated guidance of 25% GMV growth, which I expect to be hit relatively easily. As we can see, GMV growth has maintained over 20% for the past 10 quarters, a huge feat in itself.

Shopee’s GAAP take rate continued to climb, up to 13.7% this quarter.

A large reason for the 13% stock price increase was this slide. Adj. EBITDA margin as % of GMV appears to have stabilised at 0.6% following the previous quarters’ drops.

Management reiterated their guide for “full year adjusted EBITDA no lower than that of 2025 in absolute dollar terms”.

The reason for the slowdown in Adj. EBITDA growth is of course due to the Shopee VIP program that has been a huge success. So far, member growth has been exponential:

• Q1 2025: 1M members

• Q2 2025: 2M members (+100% QoQ)

• Q3 2025: 3.5M members (+75% QoQ)

• Q4 2025: 7M members (+100% QoQ)

• Q1 2026: 10M members (+43% QoQ)

Despite the rapid growth in VIP members, it represents just 2.5% of the entire active user base of Shopee (400M people). Incredibly, these 2.5% of users make up ~20% of total GMV, as revealed by management. This implies that VIP members spend on average 9.75x more than non-VIP members.

Shopee Ads

In Q1 2026, ad revenue grew 80% YoY and ad take rate increased by over 90bps. Ad-paying sellers and their average ad spend both increased by ~35% YoY.

This was an inflection in growth from Q4 2025 which I mentioned in the previous earnings report was something I expected. Ads continue to be a key contributor to the bottom-line and will get larger over time.

Shopee Logistics (SPX Express)

In Indonesia, Shopee’s instant delivery order volumes grew over 35% YoY with cost per order reducing ~20% YoY in Q1 2026. This is the shift towards quick commerce that I have been discussing. and I wrote a full length article (FREE) on this that discussed how China’s quick commerce landscape was moving to Southeast Asia. Shopee happens to be at the bleeding edge of it:

Management also noted that fulfilment order volumes grew ~25% QoQ in Q1 2026. In Asia, over one third of parcels fulfilled by SPX Express were delivered within the next day in March.

Shopee VIP Program

As we discussed earlier, the Shopee VIP program is front and center of focus now. Management shared that VIP members total over 10 million, growing over 40% QoQ, with program retention averaging over 80%.

This is extremely impressive and only possible due to Monee’s capabilities. Management discussed that for a paid VIP programme, the hard part is not only convincing users to subscribe. It is getting the subscription to renew successfully every month.

In markets with high credit-card penetration, this is easier because the card just gets charged automatically. But in Shopee’s markets, many users do not have credit cards, so subscriptions can fail when moving from one month to the next.

Chris Feng said Shopee solved this by working closely with Monee to create a smoother payment process for Shopee VIP. As a result, Indonesia’s VIP renewal rate improved from around 40% to 70% over the past few quarters.

2. Monee (Digital Financial Services)

Monee continued with yet another record quarter of loans principal outstanding. Its loan book grew ~70% YoY to $9.9B while the NPL ratio held steady at 1.1%.

There are now over 38 million active credit users on the platform, up more than 35% YoY. Average loan outstanding per user also grew to ~$250 at the end of the quarter, up 25% YoY.

Brazil has become the 4th market to cross $1B in loan book size, growing over 250% YoY. As we can see from Mercado Pago’s numbers in the region, Brazil is certainly a massive market for growth.

Brazil is still very early in Shopee’s FinTech monetisation curve. SPayLater GMV penetration is only ~10% in Brazil, well below mature Shopee markets. This suggests that Brazil has much room to grow. SPayLater usually increases conversion rate + basket size + purchase frequency + retention, which is a natural flywheel for Shopee.

GAAP revenue for the quarter came in at $1.24B. Adjusted EBITDA continued climbing, to $275M for the quarter.

The key to watch here is off-Shopee momentum. Management shared that following strong momentum in Malaysia, they are also seeing good traction in Thailand and Indonesia, with off-Shopee SPayLater loans in both markets exceeding 20% as of end-March.

There is also strong growth in higher-value categories such as electronics and two-wheelers in Indonesia, where instalment credit plays a meaningful role in enabling such purchases.

Again, the major worry I have here is off-book loans. They are growing very slowly compared to on-book loans. This will take time, as 3rd party financial institutions are difficult to convince, typically requiring default history across economic cycles, regulatory approvals in each individual market, and a minimum track record on the specific product.

Subscribe now

3. Garena (Digital Entertainment)

Garena had a stellar quarter in Q1, with the strong quarter of results since 2021. Bookings came in at $931M, against a significantly high comparable last year. This was driven by a huge increase in the Quarterly Paying User Ratio, up from 9.8% in Q1 last year to 10.9% this quarter.

Management cited strong response to Free Fire’s collaboration with Jujutsu Kaisen, which generated over 700 million official content views. It appears that Garena has mastered IP collaboration, with last year’s NARUTO SHIPPUDEN also being a massive success. They also pointed towards Arena of Valor which delivered record high quarterly bookings this quarter, its tenth year of operation.

Again, the continued success of Free Fire, which is in its 9th year of operation proves that many of these hit games are certainly evergreen.

4. Management Commentary

This section focuses mainly on questions during the earnings call, with key highlights of what I believe are important points to takeaway. I may adjust a few words here and there for clarity purposes, (English is not the first language for the majority of Sea’s management team) but I will not change the meaning of these quotes.

On Quick Commerce:

“In Indonesia, our instant delivery service can deliver orders in as little as 2 hours in urban areas. Order volumes for this service grew over 35% in the first quarter with cost per order reducing by around 20% year-on-year.

Building this service has enabled us to expand our product assortment into higher frequency categories. We expanded partnerships with major convenience stores and the pharmacy chains such as Indomaret.

At the end of March, we had around 7,000 offline stores available on our instant services. This has shifted more offline purchasing behavior online and into the Shopee ecosystem.”

On Shopee Taiwan:

“In Taiwan, our collection point network expanded to over 3,100 locations at the end of the fourth quarter, nearly 50% more locations compared to just a year ago.

We leveraged our growing fulfilment capability to scale initiatives such as shipping directly to lockers without additional packaging, improving speed while reducing costs. With these efforts, average buyer waiting time improved 12% in the first quarter year-on-year.

We reported double-digit GMV growth year-on-year in the first quarter in Taiwan”

On Shopee’s Content Ecosystem:

“Our content ecosystem continues to grow healthily. In the first quarter, orders from live streaming and short-form video grew more than 50% year-on-year.

These orders accounted for more than 25% of total physical goods orders in Southeast Asia. To further strengthen our content ecosystem, we continue to deepen our content partnerships.

Orders driven by YouTube more than doubled year-on-year. Our collaboration with Meta is scaling well with over 4.5 million affiliates across our market, up nearly 30% quarter-on-quarter. In Indonesia, we have extended our Meta collaboration to enable seamless product promotion and checkout, not just on Facebook, but also on Instagram.”

On Shopee Brazil:

“We continued to improve delivery time by more than 1 day in the first quarter compared to last year. We opened 3 new fulfilment centers, bringing our total to 5.

This effort allowed us to onboard more merchants, especially to Shopee Mall, supporting stronger spending among buyers. In the first quarter, GMV from Shopee Mall sellers more than doubled year-on-year and now contributes around 15% of GMV.

We remain confident in Brazil’s long-term growth potential and in our ability to further strengthen our competitive position in this market.”

On Shopee Brazil’s Profitability:

“We have been profitable in Brazil for the last few consecutive quarters. I don’t foresee any change towards that at this point in time.

We will continue to grow healthily in Brazil, likely with the profitable kind of margins we see right now.”

On Shopee Brazil’s Fulfilment Potential & Plans:

“On the fulfilment business, especially in Brazil, we do expect fulfilment to become a larger percentage of the business as we continue to build it out.

We only started not too long ago, so we are still in the early stages of scaling our fulfilment operations. Typically, we do not overbuild too aggressively. Our fulfilment centre capacity utilisation is relatively high because we are able to forecast fulfilment volumes well ahead of time, and then build our fulfilment centres according to that timetable.

So it is unlikely that we will build a lot this year, stop next year, wait for demand to catch up, and then start building again. Instead, we expect this to be a continuous process as we scale our fulfilment network.

Ultimately, we would like our fulfilment network to be larger than our closest competitors in the market, in terms of absolute volume.

In terms of the return on investment, if you look at individual fulfilment centers, typically, the CAPEX is actually not that high as we don’t own the fulfilment center itself.

We typically rent a fulfilment center. The CAPEX essentially is to make sure the fulfilment center is well equipped. So if you look at that particular part of investment, the return on investment is pretty fast.

It’s not that long ahead of the time. The other product investment we’re doing for the fulfilment businesses is more about moving the seller to be a part of the fulfilment center and advocating the buyers to understand the fulfilment business that we have. So that’s part of the ongoing investment we use to drive business growth.”

On the Impact of Oil:

“The first-order impact is the absolute oil price, which does affect our operating costs.

The positive is that, in many of the countries we operate in, we benefit from government subsidies that help absorb part of the cost increase, especially in last-mile delivery, which is the largest component of delivery cost. We also work closely with our partners, such as our line-haul and airline partners, to manage costs together.

So overall, if you look at actual costs, higher oil prices do have an impact. But we believe we can manage this within the guidance we have provided. And from a timing perspective, you are right that Q2 will likely see a larger cost impact than Q1.

The second-order impact is that higher oil prices could affect consumer spending power in some countries, especially if people have to spend more money at the gas station. That said, we are generally seeing only a moderate impact on our platform.

The main reason is that our platform is one of the cheapest places for consumers to find the products they need. So when people are looking to save money, they actually tend to come to us more. Our platform is also more focused on essential products, rather than luxury or highly discretionary spending. Compared to offline spending and other discretionary categories, our platform is less exposed to that kind of pullback.

All of this helps reduce the second-order impact we see from higher oil prices.”

On Monee/SPayLater’s Expansion Beyond Shopee:

“On the split between On-Shopee and Off-Shopee, SPayLater usage on Shopee was the majority of the business when we first started. Today, it is already less than half of the business.

Even when you compare SPayLater usage on Shopee versus off Shopee, SPayLater on Shopee now represents only around 20% of total expenditure across both On-Shopee and Off-Shopee channels. This is a significant milestone for us.

It proves that we are not only able to drive credit adoption within the Shopee ecosystem, but that we have also successfully expanded it beyond Shopee. In fact, we are seeing higher growth in the Off-Shopee ecosystem compared to the On-Shopee part of the business.”

On Factors Driving Monee’s Growth:

“There are three key factors driving this growth.

First, within our existing user base, we still see significant room to drive more credit adoption. This will come from more product rollouts, better credit assessment as we accumulate more data over time, deeper integration with Shopee, and the expansion of more non-Shopee use cases for these users. Essentially, even within the same user base, there is still a lot of room for us to deepen credit penetration.

Second, we are expanding into new usage scenarios beyond what we have today. This includes partnering with more online merchants that can accept SPayLater, as well as more offline merchants that can accept SPayLater.

In some markets where credit card usage is more developed, we have also rolled out a debit card product that leverages the user’s existing credit limit. This allows users to spend their SPayLater credit through the card network as well. All of this expands the addressable market for our existing user base.

Third, we will continue to expand into new user segments. This is also very important for us. As Forrest mentioned in the opening remarks, we initially started more with the subprime segment. But as we accumulate more risk data and improve our risk models, we are able to expand into more prime user segments with slightly different products across different markets.

These users may have slightly lower ROA, but they give us access to a much larger outstanding loan pool.

Overall, we believe these factors will continue to drive the growth of our lending business across our markets in the coming years.”

5. Concluding Thoughts

As I write this, the stock is down over 3% today, following a 2% down day yesterday, wiping out a large portion of gains.

We are clearly in an environment where the market is focused on companies on the AI train. That, combined with the uncertainty of oil, in my view is what has led to the weakness in Sea Limited stock in the past couple of months. Of course, there is also the overhang of TikTok Shop, Mercado Libre and other competitors, and what it does to Shopee’s margins.

Personally, I think these set of results that Sea has displayed, shows that reinvesting in the business today, both in logistics and the VIP program are the right decisions. Much of my conviction in continuing to buy Sea is trust in the management team. They have proven their ability to out-hustle and out-compete their closest rivals in the past. I believe this is no different.

From a simple SOTP or a very conservative discounted cash flow, Sea is a steal at these prices. Frankly, I have been flabbergasted by the reaction of the market to its extremely strong earnings results. However, this is simply how the market reacts in the short term.

As long-time followers will know, this does not faze me, and I hope it doesn’t faze you either. If there was a deterioration in the quality of the business, I would certainly highlight it. Yet, I believe this was possibly one of the best quarters the business has put up. All 3 segments are firing on all cylinders.

Should the stock revisit the $70s, I will certainly be a buyer again.

Paid Subscription Upgrade

If you’d like to support the work I do, consider becoming a paid subscriber. Your support will allow me to spend more time finding asymmetric opportunities in the market, writing and analysing various businesses.

As a reminder, paid subscribers get access to:

• Monthly Portfolio Updates

• Earnings Reviews on Portfolio Companies (SE, GRAB, DLO, MELI etc)

• Archive of Deep Dives and Posts (16 Deep Dives and counting)

• Southeast Asian coverage of industries and companies

Share

Disclaimer: The content presented in this thesis is for informational and academic purposes only and does not constitute financial advice. The analysis and opinions expressed are based on research and should not be interpreted as a recommendation to buy, sell, or hold any security. Readers should conduct their own due diligence and consult with a qualified financial advisor before making any investment decisions.

Today, we’re looking at something slightly different. For the first time, I’ll be covering the S-1 (registration statement) of a business looking to IPO imminently.

The business we’ll dive into today is Cerebras ($CBRS). Cerebras is at its core, an AI chip and AI compute infrastructure company. And we know how high in demand these are today.

It will IPO on Thursday this week (14th May 2026) with Reuters reporting an exclusive earlier this week that it intends to raise its IPO price range to $150-$160 ($34B market cap) up from $115-$125 ($26.6B market cap). It also intends to increase its offering to 30M shares, up from 28M shares, meaning it will raise about $4.8B at the top end.

Against $510M in 2025 revenue, that implies a ~67x P/S, a steep number for sure.

But of course, that is a simplified manner of looking at things. In this piece, we will look at the growth prospects of Cerebras, determine if the business has a moat, and just how large it could become. Is this truly an NVIDIA competitor, or merely a pretender?

Subscribe now

Table of Contents

1. Introduction

2. Company History

3. Business Model

4. Value Proposition

5. Product Offerings

6. Moats & Differentiation

7. Market Context & Industry Positioning

8. Competitive Landscape

9. Financials

10. Ownership & Management

11. Valuation

12. Catalysts & Outlook

13. Bull and Bear Case

14. Concluding Thoughts (What I am personally doing)

1. Introduction

If you walk into any AI data center today, you will find racks full of NVIDIA GPUs. Each GPU is a small chip, about the size of a postage stamp. To train or run big AI models, you wire thousands of these little chips together with cables, switches, and networking gear. The model gets "chopped up" and spread across all of them.

This works, and is why NVIDIA is the largest company in the world. However, there is one problem, moving data between chips is slow and power-hungry. Every time one GPU needs to talk to another, the signal has to leave the chip, travel down a wire, through a switch, and into the next chip. AI models do this trillions of times. The chips end up spending most of their energy and time just shuffling data around, not actually computing.

Cerebras solves this exact problem. Instead of cutting a silicon wafer into hundreds of small chips, Cerebras decided to use the entire wafer as one chip.

The result is the Wafer-Scale Engine (WSE), the largest computer chip ever sold commercially, about the size of a dinner plate.

To give you a sense of scale, the current WSE-3 is:

• 58 times larger than NVIDIA’s flagship B200 chip

• Has 900,000 compute cores on one piece of silicon (a B200 has tens of thousands)

• Has 44 GB of memory built directly onto the chip (250x more than a B200)

• Moves data internally at 21 petabytes per second (about 2,600 times the memory bandwidth of a B200)

For 50+ years, the entire chip industry assumed wafer-scale chips couldn't be made commercially. For 3 reasons:

1. Yield. Every silicon wafer has tiny manufacturing defects. The bigger the chip, the more likely it contains a defect, and traditionally one defect ruins the whole chip. A wafer-sized chip should be un-manufacturable.

2. Power and Heat. A chip the size of a dinner plate draws enormous power and generates enormous heat. Cooling it without melting the system around it is hard.

3. Connecting the regions. Wafers are normally exposed to light in small rectangular sections called "reticles." Getting those sections to communicate as one chip required inventing a new manufacturing technique.

Cerebras spent roughly a decade solving all three. Their architecture is fault-tolerant. If a small region of the chip has a defect, the chip routes around it. They invented a custom water-cooling system and a proprietary cross-reticle interconnect built into TSMC's fabrication process.

When Cerebras first shipped chips in 2020, the market didn't really need them as AI was nascent, training was slow, and most people didn't care about speed.

The ChatGPT moment happened, and then “reasoning models” were released with o1 by OpenAI. Modern AI models think through questions, sometimes generating tens of thousands of internal tokens before responding. Speed became a bottleneck, and a coding agent that takes 30 seconds to respond is useless compared to one that responds in 2 seconds.

In January 2026, OpenAI signed a $20+ billion deal to run a chunk of its inference compute on Cerebras hardware. And in March 2026, AWS agreed to put Cerebras chips inside Amazon Bedrock, making Cerebras compute available to every AWS developer.

2. Company History

Cerebras was founded in 2015, 11 years ago by 5 founders: Andrew Feldman (CEO), Gary Lauterbach, Michael James, Sean Lie and Jean-Philippe Fricker. The 5 of them worked together at SeaMicro (a low-power server startup), started in 2007 by Andrew and Gary and was later sold to AMD in 2012 for $334M.

That exit gave the founding team operational credibility and the AMD network that would later prove instrumental.

The founders saw in 2015 that AI was on the horizon and knew it would consume vast amounts of compute. They made two fundamental bets.

First, existing general-purpose processors would not be sufficient for AI. They believed what PCs did for x86, graphics did for GPUs and mobile did for ARM, would prove true in AI too.

Second, they bet that modifying existing compute architectures would not realise AI’s potential. They would need to build a new computer architecture from first principles, optimised in every way for AI.

Both bets were contrarian at the time, and have proved to be right.

Timeline Highlights:

• 2016: Series A of $27M led by Benchmark, Foundation Capital, and Eclipse Ventures

• 2018: Elon Musk reportedly tried to acquire Cerebras through Tesla to support OpenAI — founders declined

• August 2019: First-generation WSE-1 announced (16nm, 400,000 cores)

• 2020: First CS-1 systems delivered to Argonne, NETL, GSK etc…

• April 2021: WSE-2/CS-2 launched at 7nm (850,000 cores, 40GB SRAM)

• November 2021: Series F at >$4B valuation, led by Alpha Wave Global with G42 participation.

• 2023: Condor Galaxy supercomputer network announced in partnership with G42

• 2024: WSE-3/CS-3 launched at 5nm. First S-1 filed in September. CFIUS opens review of G42’s stake within weeks; IPO indefinitely delayed

• May 2025: CEO Feldman publicly confirms CFIUS clearance obtained

• September 2025: Series G of $1.1B at $8.1B valuation

• October 2025: Original S-1 withdrawn (”out of date”)

• December 2025: $1B OpenAI loan + warrants secured. NVIDIA acquires Groq IP for ~$20B in the same month (Groq is the closest pure-play competitor to Cerebras)

• January 2026: Master Relationship Agreement with OpenAI announced — $20B+ over multi years, 750 MW base

• February 2026: Series H of $1B at $23B valuation, led by Tiger Global, with AMD, Coatue, Benchmark, and Fidelity participating

• March 2026: Binding term sheet signed with AWS for Bedrock integration

• April 17, 2026: New S-1 filed

3. Business Model

The best way to describe Cerebras is as a vertically integrated AI compute company. They have three revenue streams that are converging:

1. Hardware Sales (On-Prem Systems)

Customers buy CS-3 systems (a third-generation, wafer-scale AI accelerator system designed for training and inferencing massive AI models) from Cerebras outright. This was historically the entire business. In 2025, this accounted for 70% ($358.4M) of total revenues.

Customers in this bucket include Sovereign AI initiatives, foundation model labs that demand data isolation, and national labs/defence.

2. Cloud (Cerebras Cloud + Partner Clouds)

Customers consume tokens or rent compute by the month/year through Cerebras's own cloud, or through AWS, Microsoft Azure Marketplace, IBM watsonx, Vercel, OpenRouter, and Hugging Face. In 2025, this accounted for 30% ($151.6M) of total revenues.

The OpenAI Master Relationship Agreement is structured primarily as a multi-year compute commitment funnelled through this layer.

3. Co-development & AI-model Services

Cerebras's engineers help customers select architectures, prepare training data, build draft/speculative-decoding models, and tune deployments. Bundled with strategic accounts (explicitly cited in the S-1 as a reason top-10 customers grew spend ~80% within 12 months of initial purchase.)

This is not a revenue generating segment, but is inarguably a key piece of Cerebras’ competitive advantage.

Revenue Concentration

The S-1 discloses MBZUAI (world’s first graduate-level, research-based artificial intelligence university in UAE) at 62% of 2025 revenue and G42 (leading UAE-based AI and cloud computing company) at 24%. Together, they account for 86% of total revenues. However, OpenAI’s $20B+ commitment will reshape this mix dramatically from 2026 onward.

4. Value Proposition

The key to Cerebras’ pitch is all about speed.

In Cerebras’ S-1, they make the analogy that going from GPU inference to Cerebras inference is “like going from dial-up to broadband”.

Cerebras claims that they are up to 15x faster than leading GPU solutions on standard open-source benchmarks. They are 1,000x+ faster on certain exotic workloads and are >10x faster in training time-to-solution vs. same-generation GPUs.

This speed also unlocks several other advantages for Cerebras:

1. Speed unlocks new product categories. For instance real-time coding agents (Cursor, Codex, Claude Code), digital twins/voice avatars that require sub-second latency, and “deep research” agents that can scan thousands of documents in seconds.

2. Quality. GPU-based systems force a tradeoff between answer quality and latency (developers cap reasoning tokens to keep response times tolerable). Cerebras’ speed allows models to use more reasoning tokens within the same response time, producing better answers for users.

3. Cost. Power is the dominant operating expense in AI compute, and most of that power is spent moving data between chips. Keeping data on a single wafer reduces data movement and the elaborate networking stack that comes with GPU clusters.

In essence, Cerebras is betting that wafer-scale is what wins AI, much like how x86 won PCs and ARM won mobile.

Subscribe now

5. Product Offerings

Hardware

• Wafer-Scale Engine 3 (WSE-3)

  

  • This is the main product. WSE-3 uses a 5nm TSMC node with 900,000 AI cores, 44GB on-chip SRAM, 21 PB/s memory bandwidth, 4 trillion transistors, 46,225 mm² die.

  • It is 58x larger than NVIDIA’s B200, with 19x more transistors, 250x more on-chip memory, and 2,625x more memory bandwidth than a B200 package.

• CS-3 System

  

  • This is the data-center-ready chassis which houses one WSE-3 with proprietary power delivery, water cooling, networking and slots into standard racks.

• Cerebras AI Supercomputers

  

  • Clusters of CS-3 systems orchestrated as a single logical machine. The Condor Galaxy network with G42 represents some of the largest deployed.

Software

• CSoft (Cerebras Compiler)

  • Compiles PyTorch models directly to the WSE. No CUDA dependency.

• Inference Serving Stack

  • Manages hosting, scaling, and routing; exposes standard OpenAI-compatible API endpoints.

• Cluster Manager

  • Telemetry, scheduling, health monitoring. Lets customers flip between training and inference on the same systems.

Deployment Models

• On-premises

  • Full CS-3 supercomputers shipped to customer data centers (Sovereign AI, defense, regulated enterprises, foundation model labs).

• Cerebras Cloud

  • Managed inference and training, API-based.

• Partner Clouds

  • AWS Bedrock, Microsoft Azure Marketplace, IBM watsonx, Vercel, OpenRouter, Hugging Face.

• Hybrid

  • Unified software stack across on-prem and cloud.

Roadmap

Cerebras has shipped 3 generations of WSE, from 16nm to 7nm and now 5nm. The company intends to continue investing in expanding on-chip memory, memory bandwidth, interconnect density, and process-node migration.

Notably, the architecture is presented as forwards-compatible, meaning customers don't have to rewrite code as Cerebras moves between generations. Currently, when NVIDIA releases a new GPU generation (say H100 to B200), developers often have to re-write or re-optimise their code to take full advantage of the new hardware. The CUDA ecosystem mitigates this but doesn't eliminate it.

Cerebras claims that customers won’t have to do this, as a model that ran on WSE-2 will run on WSE-3 without code changes and will run on WSE-4 the same way.

6. Moats & Differentiation

I believe Cerebras’ moats are purely technological. That could turn into distributional or other types of moats in future, but I do view it to be rather limited at this point.

1. Wafer-Scale Integration IP and Process Know-How

Historically, there have always been two problems that made wafer-scale chips impossible to produce commercially.

Firstly, chip-making machines can only “print” patterns onto silicon in small rectangular sections (about the size of a postage stamp). Normally, those sections get cut apart into individual chips. To make one giant chip, Cerebras needed wires that cross between those sections. They worked with TSMC for years to develop a custom process that creates these connections. This is now baked into TSMC’s manufacturing flow specifically for Cerebras.

Secondly, defects. Every silicon wafer has tiny manufacturing flaws scattered randomly across it. A normal chip with one flaw gets thrown out. Each Cerebras chip, owing to its size, has 40-50 flaws. By normal logic, it should be unusable. However, Cerebras solved this by designing the chip with hundreds of thousands of small identical cores plus spare cores. When a defect is found, the chip automatically routes around it and switches in a spare.

Multiple companies have tried wafer-scale chips for 50 years but only Cerebras has commercialised one. Replicating this will require years of work with TSMC.

2. Full-Stack Co-Design

Running AI on NVIDIA GPUs is complicated as customers manage a stack of software from CUDA to distributed training frameworks, cluster management and inference servers. Major AI labs have to hire hundreds of engineers just to keep the stack working.

Cerebras’ pitch is that you are able to write your model in PyTorch, hand it to Cerebras’ compiler, and it runs smoothly with all the complexity hidden. Once a customer is committed, this becomes a workflow lock-in, similar to a switching cost moat that the likes of Microsoft possesses with its Microsoft Suite.

3. Manufacturing Relationship

Cerebras has spent years working with TSMC on the cross-reticle process. Even a well-funded competitor would need 3-5 years to catch up. However, this also means TSMC is a single point of failure. If something disrupts TSMC, Cerebras has no Plan B.

4. Strategic Partnerships

Cerebras has three anchor relationships today, which is probably why most are interested in the stock, and also proves its capability.

Firstly, it has a $20B+ multi-year deal with OpenAI, and a $1B loan plus warrants for ~10% of Cerebras stock if the full deal is exercised. OpenAI has a direct financial incentive in Cerebras’ success.

Secondly, Cerebras chips are being added to Amazon Bedrock, which means every AWS developer will be able to use them. This is a level of distribution that Cerebras could never build itself, but provides them with network effects.

Lastly, the legacy relationship in MBZUAI and G42 that today account for 86% of their revenue. Their close relationship has enabled the business to put up the strong numbers and validated their investment in wafer-scale chips.

While these may not be moats in the strict sense, they create demand visibility, customer validation, and a sales proof point.

7. Market Context & Industry Positioning

Bloomberg Intelligence estimates that the combined AI training infrastructure and inference market was $251B in 2025 and will grow to $672B by 2029 implying a 28% CAGR. Inference is also growing >2x faster than training infrastructure.

IDC projects AI investments will yield $22.3T in global GDP impact by 2030.

I don’t think any of us are doubting these numbers. They are quite clearly a large enough addressable market and even a small slice probably supports a $34B market cap. The question is, how large of a slice could Cerebras get and how many other players are vying for that slice?

The Inference Shift

One of the dominant stories in the past 6 months has been that inference, not training, drives the bulk of new compute demand. This has been corroborated by the performance of AMD compared to NVDA in the past year (+320% v +80%). The perception has always been that AMD is more focused on inference than NVDA.

The demand for inference is driven by 3 compounding forces:

1. More users

2. Higher frequency of use per user

3. More compute per use (reasoning models do “test-time compute” generating long internal chains of thought that consume many more tokens than a single Q&A)

Inference-time compute is driving the demand for Cerebras. Due to reasoning models, speed has become a binding constraint rather than a nice-to-have. A 15x speed advantage on a 30-second-thinking reasoning model is more economically meaningful than a 15x speed advantage on a 1-second chat reply.

HBM Bottleneck

A point worth flagging is that Cerebras is a natural hedge against HBM (High Bandwidth Memory) shortages. GPUs depend on HBM, which is supply-constrained by Samsung/SK Hynix/Micron capacity. Cerebras uses on-chip SRAM. As HBM remains tight through 2026-2027, this is a structural tailwind for Cerebras.

8. Competitive Landscape

This is the most important section to understand, because Cerebras's valuation depends almost entirely on whether they can hold a defensible position against this list.

18 months ago, the AI silicon market was a wild zoo with dozens of well-funded competitors (Cerebras, Groq, SambaNova, Graphcore, Tenstorrent, Mythic, Rain AI, Tachyum, Lightmatter, D-Matrix etc.), most of whom claimed to be the “NVIDIA killer”. The inference compute market has consolidated dramatically in the last six months.

Read more

Revenue: $955M v $921M est. (+24% YoY, 19% Constant Currency)

EBITDA: $154M v $142M est. (+46% YoY)

EPS: $0.04 v $0.02 est.

Selected Key Metrics

• On-Demand GMV: $6,131M (+24% YoY)

• On-Demand GMV per MTU: $130 (+4% YoY)

• Group MTUs: 51.6M (+16% YoY)

• Loan Portfolio: $1,438M (+130% YoY)

• Adjusted EBITDA: $154M (+46% YoY)

• Adjusted Free Cash Flow: $76M (+10% YoY)

Grab’s earnings for the quarter beat expectations on both top and bottom-line. This was quite surprising considering Q1 was expected to be a weak quarter due to fuel price spikes related to the Strait of Hormuz crisis.

Interestingly, it was not a case of deliveries/financial services cushioning the fall in mobility. The mobility business remained extremely strong, with volumes in April up 32% YoY despite the increased fuel prices.

In this piece, I will break down the earnings in full, highlight key points on management commentary during the call and discuss my personal thoughts in the last section.

Table of Contents

1. Deliveries & Mobility

2. GFin, Financial Services

3. Key Call Commentary

4. Positives & Negatives

5. Conclusion

1. Deliveries & Mobility

Deliveries is Grab’s largest segment by GMV and Revenue.

GMV grew 24% YoY (or 22% on a constant currency basis). Adj. EBITDA margin also grew from 2.0% to 2.3%, tracking closer towards management’s long-term target of 4+% margins.

This growth has been attributed to higher contributions from advertising and gains in operating leverage. Segment Adjusted EBITDA was up 40% YoY as a result.

Mobility GMV grew 23% YoY (or 19% YoY on a constant currency basis). Revenue growth was slower at 19% YoY (or 15% YoY on a CC basis) due to the continued focus on affordability.

Segment Adj. EBITDA continued expanding from 8.8% to 8.9%. This is more or less the steady state margin target that management has issued for mobility (9+%).

2. GFin, Financial Services

This quarter was yet again very strong for GFin, following on from last quarter. The net loan portfolio reached $1.44B, up from $1.18B just last quarter. That represents 22% QOQ growth and 130% YoY growth.

Management has set a FY2026 target of $2B, that appears to be a rather low bar considering the current numbers. Segment Adj. EBITDA margin improved from -41.1% to -15.71%. Grab tracks closer towards profitability, which it expects in the 2nd half of this year.

Customer deposits across GXS Singapore and GXBank Malaysia remained relatively stable at $1.63B.

As I’ve discussed in the past, I believe Grab has the potential to grow much faster than ~40%. For instance, Sea has 10x the loan book but is growing faster. Grab’s recent launch of cash loans for consumer could help spur this.

3. Key Call Commentary

Again, Grab hosted a live video earnings call with an IR head taking charge of questions.

On New Product Innovations

“Product innovations we have made have really targeted affordability and reliability. Group orders, for example, has GMV up 74% year-on-year, and we launched group rides at GrabX last month, which is a similar concept for sharing rides to reduce pricing for individual consumers. And that’s now available across all 6 of our core markets. GrabUnlimited, of course, is very good value for high-frequency customers, and it continues to account for 1/3 of our deliveries GMV. So all of these are highly affordable products, which keep the demand strong even when consumers are stretched.”

On Fuel Crisis

“We’re monitoring the fuel situation extremely closely. And of course, we will not hesitate to act further if needed. In the medium term, we are committed to accelerate the EV transition to reduce our driver partners exposure to fuel price volatility.

So for example, in Thailand and Philippines, we have a drive-to-own program that connects our drivers with OEMs like BYD and GAC, where we have deals of up to 70,000 vehicles available across 6 markets with accessing to financing so they can own those more easily. In Vietnam, we have secured preferential charging rates also through our charging network partners, EBOOST and Charge+, which helps our drivers in the transition also. And finally, in Thailand, I am pleased to say that our total fleet supply has crossed 30,000 EVs on the platform and demand for those from consumers is also strong, where they can select that EV option, and that demand has grown by over 35% year-on-year.

So this fuel crisis has become an opportunity in the sense that it helps us to accelerate that EV transition.”

On Indonesia’s 8% Commission Cap

“The recent announcements are explicitly focused on Ojek drivers, who are our 2-wheel ride hailing partners. The 4-wheel drivers earn well above the minimum wage, so we believe they're less of a concern for regulators in Indonesia.

That said, of course, we're engaging very proactively with the relevant ministries to try to seek absolutely clarity and the technical aspects of how the decree will be implemented.

It's worth noting that 2-wheel mobility, that the decree referred to in Indonesia, is less than 6% of our total mobility GMV. So we are, therefore, reiterating our expectations for mobility margins to stabilise within that historical range and not to go outside of that range."

On GrabMart

“GrabMart is an exciting segment. The TAM is very large, arguably larger than food delivery altogether. So we are doing a lot to accelerate the product innovation particularly the front end, the AI-powered shopping agent, which we think will transform the ease with which consumers can, for example, create a weekly shopping basket and then improve the targeting for Grab more cross-sell as well.

And by the way, GrabMore grew more than double-digits quarter-on-quarter. Overall, as a result, the grocery MTUs are growing at 2.6x the rate of food MTU growth on a year-on-year basis. So that shows you that it’s really expanding the top of our funnel, which is extra important in the age of AI in terms of generating data and deepening the long-term value relationships that we have with our consumers.

And then the order frequency that we saw were 1.8x higher than the food-only users, which illustrates that long-term value enhancement that I was speaking about. So over the long term, the North Star is very clear. We’ve got global peers, who have achieved like 20% to 40% mart penetration as a percentage of their deliveries business overall. So it’s definitely the right model that we’re pursuing.”

On Grab’s Loan Book

“We actually don’t have any issue at all in raising deposits. We’ve been really gratified at the trust that consumers have in the Grab brand, the Grab ecosystem, our capabilities to protect their money. And if you look at the pricing of our deposits, we are never the most aggressive in the market. We’re able to actually gather sufficient deposits to create the right shape of balance sheet. So there’s no point having excess deposits, particularly in this yield curve environment. So what you’re seeing is that’s carefully managing the level of deposits to make sure that we optimise for P&L purposes. If we needed to raise more deposits, we’re very confident that we can do that.”

4. Positives & Negatives

Positives:

1. Growth accelerated despite seasonality and fuel price concerns

Coming into earnings, one of the biggest worries was the increase in fuel prices that would affect demand across the region. Yet, transport volumes were up +32% YoY in April.

GMV also came in much stronger than expected which led to a beat in top-line expectations.

2. Operating leverage in action

Grab’s Adj. EBITDA margin increased from 13.7% to 16.2% YoY, which combined with GMV increasing 24% YoY led to a 46% increase in Adj. EBITDA.

This is classic operating leverage in action, and I believe we will start to see this increase exponentially in the months and years to come, especially as GFin turns profitable in H2 2026.

3. Advertising is becoming a real profit lever

GrabAds is becoming a core profit driver for the business. Average spend of quarterly active advertisers on Grab’s self-serve platform had average spends grow 44% YoY.

GrabAds is vital for Grab because it can lift delivery margins without hurting consumers or drivers. It is similar to how advertising improved the economics of other marketplace businesses. (Think Amazon, DoorDash, Meituan)

4. New product innovations are contributing to top and bottom line growth

Group orders for instance, grew by 74% YoY and has contributed significantly to deliveries growth. I discussed my take on new product innovations that were announced at Grab’s 2026 product day.

5. The initial reaction to the 8% commission cap appears to be overblown

Grab discussed during the earnings call that it was only applicable to 2-wheelers, which represented just 6% of total mobility GMV. Using FY2025 numbers, and assuming a 12pp cut in take-rates, it would result in a $57M drop in annualised revenue or 1.7% haircut on the group’s top-line.

Negatives:

1. Partner Incentives

It is never good to see partner incentives up 54% YoY. In this case, it was a necessary action to an unexpected event. In time, this will moderate, but it certainly affected the bottom-line, which led to a QoQ fall in net income.

2. Fuel costs are likely to remain a headwind

Management repeatedly referred to elevated fuel costs and the regional fuel crisis. Grab is supporting drivers through fuel discounts, cashback, EV transition initiatives, and government subsidy coordination.

The risk is that if fuel costs stay high, Grab may eventually have to choose between three bad options: subsidise drivers more, raise consumer prices, or accept weaker supply. None are great for margins.

3. Indonesia commission regulation sets a precedent and potential spillover effect

As I discussed in this piece, the worry of Indonesia’s policy, is that drivers in other Southeast Asian countries may push for similar caps. If it spreads to Malaysia, Thailand, Vietnam, Philippines, or Singapore in modified form, the valuation impact could be much larger.

4. GFin growth could be faster

Again, I believe Grab should be much more aggressive in growing its Financial Services business. It has to be a core growth driver that brings the group’s revenue growth closer to 30% as Monee is doing for Sea.

Grab’s management ultimately remains relatively conservative which is one of the largest drawbacks. I believe they do have shareholders’ interest at heart and are long-term thinkers, but to succeed and stake a strong claim on the financial services segment in the region, I do think they have to be more aggressive.

5. Conclusion

Overall, this was a very strong quarter for the business amidst challenges that they couldn’t anticipate or prepare for. Grab stock is down over 25% YTD due to concerns over fuel prices and general weakness in emerging markets.

This quarter has put to bed many of these concerns.

That said, fuel will continue to be a headwind for the foreseeable future.

While fuel will continue to be a headwind for the foreseeable future, it is ultimately a short-term issue and will mean less to Grab as it transitions to EVs and AVs over the long-term. I remain optimistic and positive on Grab at these prices as I have reiterated many times.

Please feel free to read my 2026 thesis to understand why:

Thanks for reading! If you enjoyed this piece, please give it a like and consider subscribing to a tier of your choice.

Subscribe now

As a reminder, paid subscribers receive 6-8 articles a month covering:

• Earnings Reviews (Grab, Sea Limited, dLocal, Mercado Libre and more)

• 1-2 Deep Dives (always happy to receive suggestions)

• Portfolio Review (sharing all my positions & changes)

Just yesterday, I released my Circle Deep Dive which I would recommend checking out if you are interested!

Disclaimer: The content presented in this thesis is for informational and academic purposes only and does not constitute financial advice. The analysis and opinions expressed are based on research and should not be interpreted as a recommendation to buy, sell, or hold any security. Readers should conduct their own due diligence and consult with a qualified financial advisor before making any investment decisions.

Circle is undoubtedly the best public market proxy on stablecoins today. Its flagship product, USDC, is a fully reserved digital dollar that circulates across 30+ blockchain networks, effectively a protocol for dollars on the Internet.

At the end of 2025, there was $75.3 billion USDC in circulation (up 72% YoY), with on-chain transaction volume hitting nearly $12 trillion in Q4 alone (247% YoY growth).

While ~95% of Circle’s revenues today are derived from USDC, the business is diversifying its revenue streams into other areas such as Circle Payments Network, Arc, wallet infrastructure, developer services, tokenised money market products, and enterprise settlement tools.

In this piece (11,700 words), we will discuss Circle’s history, its value proposition, moat, future prospects and what I personally think of it as an investment.

Subscribe now

Table of Contents

1. Company History

2. Value Proposition

3. Business Model

4. Product Offerings

5. Moats & Differentiation

6. Market Context & Industry Positioning

7. Competitive Landscape

8. Financials

9. Ownership & Management

10. Valuation

11. Catalysts & Outlook

12. Bull and Bear Case

13. GabGrowth Quality Score

14. Concluding Thoughts (What I am personally doing)

Subscribe now

1. Company History

Origins and Vision (2013-2016)

Circle was founded in October 2013 by Jeremy Allaire and Sean Neville in Boston, Massachusetts, with $9 million in seed funding from Jim Breyer, Accel Partners and General Catalyst.

Allaire was already a serial internet entrepreneur. He'd co-founded Macromedia (maker of Flash) and had a front-row seat to the web, mobile, and cloud platform shifts. He also founded Brightcove in 2004, which became a major online video platform for media companies and brands, helping businesses publish and distribute video on the internet before YouTube and streaming became as dominant as they are today.

When Bitcoin emerged, he viewed it as a potential protocol for value on the internet, analogous to HTTP for information or SMTP for email.

Allaire believed that blockchain technology would eventually allow fully reserved dollars to move on the internet in an open, interoperable way, which would drive the cost of storing and moving value down to effectively zero.

However, in 2013, none of the conditions for that vision existed. Blockchain infrastructure was primitive, regulators was non-existent and banks wouldn't touch it. Consumers certainly weren't asking for internet-native money, or understood it at the time.

Therefore, Circle started with what was available, creating a consumer Bitcoin payment app called Circle Pay. The product let users buy, hold, and send Bitcoin and fiat currencies through a simple mobile interface. It was essentially a Venmo for crypto, and with Allaire’s credibility, Goldman Sachs led a $50M funding round alongside IDG Capital Partners.

In September 2015, Circle became the first company to receive New York's BitLicense, the state's newly created framework for virtual currency businesses. In April 2016, it became the first firm to receive virtual currency licensure from the British government. While competitors were moving fast and breaking things, Circle was collecting licenses. At the time, this was seen as a slow move, but looking back, was probably one of the most decisive moves.

By late 2016, Circle stopped supporting Bitcoin exchange functionality in its consumer app entirely. The company stated it was "now more than ever not a consumer bitcoin exchange" and would focus on global payments and "next-generation blockchain technology." It was the first clear signal that Allaire was pivoting toward what would become USDC.

Pivot & Costly Mistakes (2017-2019)

The next phase of Circle's history is messy, costly, and absolutely critical to understanding the company today.

In February 2018, riding the crypto bull market, Circle acquired Poloniex, one of the largest cryptocurrency exchanges at the time for $400 million. The logic was to build a broad digital asset marketplace. Circle simultaneously raised $110 million in a Series E led by Bitmain (the Chinese mining giant) at a nearly $3 billion valuation. The company was applying for banking and securities licences, envisioning itself as a regulated crypto-financial conglomerate.

However, the 2018 crypto winter crushed trading volumes. Poloniex's ranking plummeted. The SEC began investigating tokens traded on the exchange. Circle found itself managing a regulatory minefield that distracted from its core vision. By October 2019, Circle sold Poloniex (just 18 months after buying it) at an estimated $156 million loss. It subsequently sold its Circle Trade OTC business to Kraken and its Circle Invest retail trading app to Voyager Digital.

Circle made bad mistakes and these were expensive lessons, but by stripping away all their non-core business lines, Circle was left with one product, which embodied its founding vision: USDC.

USDC had launched quietly in September 2018, co-created through the Centre Consortium, a joint venture between Circle and Coinbase. Each USDC token was fully backed by U.S. dollar reserves, issued on the Ethereum blockchain, and designed to be the kind of boring, transparent, fully reserved digital dollar that institutions could eventually trust. In its first few months, over 100 crypto-focused businesses adopted USDC, but it was still tiny, mostly a niche settlement tool for the digital asset ecosystem.

COVID Growth and the SPAC Era (2020-2022)

COVID-19 was an accelerant for digital dollars just as it was for digital commerce. As the world moved online, demand for on-chain settlement surged. USDC circulation grew from roughly $4 billion at the start of 2021 to $52.5 billion by early 2022, a staggering 13x increase in barely a year. Circle went from interesting niche player to systemically important infrastructure almost overnight. Its market share surged during this period to as high as 35%.

In July 2021, Circle announced plans to go public via a SPAC merger with Concord Acquisition Corp, initially valued at $4.5 billion. By February 2022, with USDC's circulation having more than doubled since the deal was announced, the terms were renegotiated to $9 billion.

Then the crypto industry imploded. Terra/Luna collapsed in May 2022, wiping out over $60 billion. Celsius and Three Arrows Capital failed. FTX, once valued at $32 billion, went to zero in November 2022. The SEC, understandably spooked, delayed qualification of Circle's SPAC filing. In December 2022, Circle and Concord mutually terminated the merger agreement.

The path to public markets was dead. But the worst was yet to come. On March 10, 2023, Silicon Valley Bank collapsed. Circle disclosed that $3.3 billion of USDC's cash reserves, roughly 8% of the total backing, were trapped at the failed bank. USDC immediately broke its dollar peg, plunging to $0.87 over the weekend. In a single week, investors redeemed a net $6 billion out of USDC.

This was an existential moment. Had the FDIC not backstopped SVB deposits that Sunday night, Circle could have faced a permanent shortfall in its reserves. The peg recovered within days once the government guarantee was announced, but the damage to market confidence was severe. USDC circulation, which had already been declining from its $55 billion peak, continued falling to a low of roughly $24 billion by late 2023. Tether, which had no direct U.S. banking exposure, surged to over 70% stablecoin market share.

The SVB episode is essential context for understanding Circle today. It exposed a genuine vulnerability (counterparty risk in reserve custody) that Circle has since addressed by shifting the vast majority of reserves into short-term U.S. Treasuries managed by BlackRock.

However, it also demonstrated something important about the network. Even after the worst stress test imaginable, USDC didn't die. It de-pegged, bled assets, and then stabilised and began rebuilding. It proved the stickiness of the underlying network effects and that it had true resilience.

Refocusing, IPO, Platform Expansion (2023-Now)

The 2023-2024 period was Circle's rebuilding phase. The company cut staff, discontinued non-core investments like SeedInvest (sold to StartEngine), and narrowed its entire focus to stablecoin infrastructure. Revenue remained strong during this period driven by high interest rates on the reserve portfolio ($1.45B in 2023 and $1.68B in 2024). Circle started generating real profits from what was now a clean, focused business.

In April 2025, Circle filed its S-1 for a traditional IPO. By June 2025, the company listed on the NYSE under the ticker CRCL. After years of false starts and near-death experiences, Circle was finally public.

Circle stock was priced at $31 per share but was heavily oversubscribed and eventually opened at $69, closing the day at $83. In the subsequent few weeks, Circle traded as high as $289.99, an incredible 4.2x from the opening price. Circle has since been in a significant drawdown as crypto experiences a bear market again, falling to as low as $50 in February this year.

Today, the core layer remains USDC, which by year-end 2025 had rebounded to $75.3 billion in circulation, up 72% year-over-year, with nearly $12 trillion in on-chain transaction volume in Q4 alone. Circle also issues EURC (the largest regulated euro stablecoin at €310 million) and acquired Hashnote to gain USYC, a tokenised money market fund that grew to roughly $1.7 billion.

Below that, Circle launched Arc in Q4 2025, a new Layer 1 blockchain designed as an economic operating system for regulated financial activity, with over 100 major companies in its testnet including Goldman Sachs, Deutsche Bank, Visa, and Mastercard. Arc is pre-revenue and pre-mainnet. It's a bet on the future, not a contributor to today's financials.

Above the core, Circle built Circle Payments Network (CPN), an application layer for cross-border money movement. By February 2026, CPN had 55 financial institutions enrolled, $5.7 billion in annualised volume, and was live in 14 markets with 11 more planned. It also launched StableFX in production beta for on-chain foreign exchange.

The evolution from Bitcoin wallet to crypto exchange to stablecoin issuer to full-stack internet financial platform is what I believe most investors miss when they look at Circle's simple-looking P&L. The 13 years of regulatory groundwork, painful pivots, and infrastructure investment have formed the foundation of the business, and is the moat today.

2. Value Proposition

Before we start with Circle’s value proposition, it is important to discuss the value proposition of stablecoins and how they work, as it is crucially important to the Circle thesis.

Moving money on the internet is still shockingly slow. A wire transfer takes 1 to 3 business days. It costs $25 to $50. Cross-border is even worse due to correspondent banks, FX spreads, compliance checks, and 3 to 5 day settlement windows. The system runs on bank hours which closes on weekends and this hasn't meaningfully improved in decades.

Stablecoins close that gap. A stablecoin is a digital token on a blockchain, pegged 1:1 to a real currency like the US dollar. For every token in circulation, there's a matching dollar sitting in reserve, typically held in cash and short-term US Treasuries. You can think of it as a dollar that lives natively on the internet.

When you send a stablecoin from Singapore to Argentina at 2am on a Sunday, it settles in seconds, costs fractions of a cent with no banks in the middle. There is no waiting for business hours and the blockchain is the settlement layer which never closes.

This matters for 3 reasons:

1. Speed and Cost. Stablecoins do what the traditional banking system does, but faster and cheaper. A cross-border payment that used to take days and cost tens of dollars now takes seconds and costs nearly nothing. For businesses moving money globally, this is a step change improvement.

2. Access. Most of the world doesn't have easy access to dollar-denominated savings or payments. In countries with high inflation, capital controls, or weak banking infrastructure, a dollar-pegged stablecoin on a phone is transformative as it acts as a savings account, a remittance rail, and a payment method in one.

3. Programmability. Unlike a dollar in a bank account, a stablecoin can be programmed. For instance, it can be held in escrow and released automatically when goods are delivered. It can also be streamed as payroll by the second or used as collateral in lending protocols.

The stablecoin market is around $320 billion today, with projections to reach trillions of dollars in a few years. Transaction volumes could hit $100 to $200 trillion annually, while Visa and Mastercard are already integrating stablecoin settlement. Stablecoins are a game changer and have a very enticing value proposition.

Circle's value proposition is fundamentally different from most FinTech companies because it doesn't serve end consumers directly. Instead, it is infrastructure that enables services to be built on top of it. This distinction matters enormously for understanding who Circle actually creates value for and why the network keeps growing without Circle needing to acquire every user itself.

For Developers and Technology Companies:

If you're building any product or service that needs to move, store, or programme dollars on the internet, USDC is essentially the default choice. It operates across 32 blockchain networks, more than any other stablecoin, making it the most portable and interoperable digital dollar available. Circle provides SDKs, APIs, smart contract templates, and programmable wallets that allow developers to embed dollar-denominated financial functionality into their products without needing to build banking infrastructure from scratch.

This is a massive unlock. A startup in Southeast Asia building a remittance app doesn't need to negotiate with correspondent banks, obtain money transmission licences in every jurisdiction, or build settlement infrastructure. They can integrate USDC and plug into a globally available, near-instant, near-free payment rail that already has deep liquidity. The same applies to AI developers who need their agents to make and receive micro-payments, e-commerce platforms that want to offer stablecoin checkout, or capital markets firms building tokenised products that need a cash settlement layer.

Importantly, most of these integrations happen without any commercial relationship with Circle. Tens of thousands of products and services use USDC today, and the vast majority simply build on the open protocol. This organic, developer-driven adoption is what creates the utility network effect and makes USDC’s position so hard to displace.

For Financial Institutions:

This is the cohort that I believe will drive the next phase of Circle's growth, and where the value proposition is most compelling relative to the legacy system.

To understand this, we have to give an analogy of what cross-border money movement looks like today. A business in Portugal sending payment to a supplier in Brazil goes through multiple correspondent banks, each taking a cut, with settlement that can take days, requires pre-funded nostro/vostro accounts, and involves significant FX spread on top.

The capital tied up in these intermediary positions is enormous. For instance, JPMorgan alone moves north of $10 trillion per day in gross money settlement.

Stablecoins collapse this entire chain as settlement happens in seconds. There's no pre-funding requirement because the value moves atomically. The capital efficiency gain is what’s driving major institutions to Circle’s doorstep. Firms like Fiserv, FIS, Corpay, and Matera, infrastructure providers that serve tens of thousands of banks, have all announced integrations with Circle's technology. Visa and Mastercard have launched products around USDC settlement. Even G-SIBs are leaning in to improve their own international money movement using stablecoin rails.

What Circle offers these institutions is not just the stablecoin itself, but the trust infrastructure around it. Circle holds regulatory licenses across 49 U.S. states plus dozens of international jurisdictions. It also has elite reserve transparency (BlackRock-managed Treasuries, audited monthly) and conditional approval for a National Trust Bank charter. No other stablecoin issuer comes close on this dimension.

For Enterprises and Corporates:

Large enterprises that operate across geographies face constant friction in treasury management, collecting revenue in multiple currencies, repatriating funds, managing FX exposure, dealing with settlement windows and bank cut-off times. USDC offers a way to unify treasury operations onto a single, programmable rail. We're seeing adoption from firms like Brex, Deel, and Gusto for payroll and corporate payments, and from companies that want to move money across subsidiaries without waiting for SWIFT to process.

For Dollar-Hungry Populations:

This is perhaps the most under-appreciated use case. In countries with high inflation, capital controls, or unstable banking systems, holding dollars is not really an investment, but a priority for survival. USDC provides a way for individuals and businesses in these markets to access, hold, and transact in dollars without needing a U.S. bank account.

Allaire has spoken about this as a continuation of American monetary policy goals. For instance, the GENIUS Act itself was partly motivated by the strategic value of extending dollar reach globally through digital means.

3. Business Model

At a high level, Circle's business model today is very simple. The target is to grow the amount of USDC in circulation, earn yield on the reserve assets backing it, and re-investing a portion of that yield into distribution partnerships and platform expansion to grow circulation further. A perfect flywheel.

However, it isn’t that simple of course. There are important nuances around distribution economics, margin structure, and the emerging revenue lines that will determine whether Circle can evolve from a one-dimensional reserve income business into a multi-revenue platform.

Reserve Income (The Core Engine)

Circle holds cash-equivalent assets, primarily short-term US treasuries and repurchase agreements managed by BlackRock to back every USDC in circulation. It earns the prevailing short-term interest rate on these reserves. In Q4 2025, the reserve return rate was 3.81%, down 68 basis points year-over-year as SOFR declined.

This is the business in its simplest form. More USDC in circulation x the risk-free rate = more revenue. In Q4 2025, total revenue and reserve income was $770 million, up 77% year-over-year, driven overwhelmingly by the growth in average USDC circulation (which ended the year at $75.3 billion, up 72% YoY) partially offset by the lower reserve rate.

The obvious bull case here is volume growth. Circle's own internal model assumes a 40% CAGR in USDC circulation, which they describe as their "conservative" base case. Third-party research CAGRs range from 25% at the low end to 90%+ at the high end, with a median around 60%.

The equally obvious risk is interest rate sensitivity. If the Fed cuts aggressively and short-term rates fall to, say, 2%, Circle's revenue per dollar of USDC roughly halves. This is the single biggest vulnerability in the current model and the reason why diversifying into non-rate-dependent revenue is strategically critical.

Distribution Costs (The Coinbase Question)

This is where the P&L gets complicated and where I think most investors should spend their time.

Circle doesn't acquire most of its USDC circulation directly. It grows distribution through partnerships with exchanges, wallets, neobanks, payment companies, and other financial institutions. Many of these partnerships involve revenue-sharing arrangements where Circle pays a portion of reserve income to the partner as an incentive to hold and distribute USDC on their platform.

The largest and most scrutinised of these is the Coinbase relationship. Coinbase took an equity stake in Circle and is the co-creator of the original Centre Consortium. Under their agreement, Coinbase receives a share of reserve income on USDC held on its platform. As Coinbase holds a significant portion of total USDC circulation, this is a material cost line. In Q4 2025, total distribution, transaction, and other costs were $461 million (roughly 60% of gross revenue).

The resulting metric that matters is RLDC margin (Revenue Less Distribution Costs). In Q4 2025, RLDC margin was 40.1%, up modestly quarter-over-quarter. For FY 2025, RLDC margin was 39.4%, exceeding the company's guidance of approximately 38%.

There are two ways to think about this. The bear case and the bull case.

The bear case is that RLDC margins stay compressed or get worse as Circle has to pay more partners to drive growth, essentially giving away the economics to grow a network it doesn't fully control. This is a structural disadvantage that Circle faces due to its position of dependency.

The bull case is that three tailwinds provide structural support to RLDC margin over time. First, growing network effects mean that as USDC becomes more dominant, new partners integrate it organically without needing incentive deals. Many products are built and launched using USDC without any commercial relationship with Circle. The stronger the network, the less Circle has to pay to grow it. Second, growth in on-platform USDC (i.e., dollars held within Circle's own infrastructure rather than on partner platforms) that carries better economics. On-platform USDC grew 5.6x YoY to $12.5 billion at year-end, representing 17% of total circulation, up from roughly 3% a year prior. Third, growth in other revenue (covered next) adds high-margin dollars on top of the reserve income base, improving blended margins.

Other Revenue (The Emerging Layer)

Circle has begun monetising transaction flows and network infrastructure beyond pure reserve income. In FY 2025, other revenue reached $110 million, exceeding their guidance of $90-$100 million. These revenues come from two buckets.

Subscription and services revenue is primarily from blockchain network partnerships. When Circle deploys USDC onto a new blockchain, it often receives both upfront integration fees and ongoing recurring revenue. Circle added 12 new chains in 2025. This revenue stream is relatively “lumpy” because of the upfront component, although the recurring base is growing steadily underneath.

Transaction revenue comes from blockchain rewards (Circle runs validator infrastructure, including a super validator on the Canton Network), USYC redemption fees, and emerging payments-related fees. The Canton Coin trading launch drove an outsized Q4 contribution.

What matters most about this revenue segment is not the dollar amount ($110M). Rather, it is that these revenue streams are high margin and scale with network adoption and transaction volume rather than interest rates. For FY 2026, Circle guided other revenue of $140-$160 million, implying 27-45% growth. I think this is conservative given the pipeline of new products (CPN monetisation, Arc Mainnet fees, StableFX), but the base is still small relative to reserve income. This segment needs to scale meaningfully before it moves the needle.

4. Product Offerings

Circle's products span four core areas today, with two emerging platform layers that are pre-revenue but strategically important to the thesis.

USDC (The Core)

This is very much the core revenue driver for the business. USDC is a fully reserved digital dollar that operates across 32 blockchain networks. Every token is backed 1:1 by cash-equivalent assets, primarily short-term U.S. Treasuries and repurchase agreements managed by BlackRock. 96% of total revenues in 2025 came from reserve income, tied directly to stablecoins, of which USDC is the largest driver by far.

The question is, what gives USDC a moat? Why can’t other players simply enter with their own stablecoins and take share?

The beauty of USDC lies with the infrastructure stack built around it.

1. Circle Mint provides enterprise-grade minting and redemption, the primary market where institutions can create or destroy USDC by depositing or withdrawing dollars through Circle's banking infrastructure.

2. Circle has facilitated over $1 trillion in mint/redeem volume since 2018, and in Q4 2025 alone, mint/redeem volume was $163 billion. This primary liquidity, backed by relationships with systemically important banks globally, is one of Circle's deepest moats.

3. CCTP (Cross-Chain Transfer Protocol) is another critical piece of USDC infrastructure. It enables native burning and minting of USDC across blockchains, eliminating the need for "wrapped" or "bridged" versions that introduce counterparty risk. CCTP volume grew approximately 640% year-over-year in Q3 2025 to $31.3 billion. By Q4, CCTP represented over 50% of all bridged volume across all assets and all bridge providers, not just USDC.

EURC (Euro Stablecoin)

EURC is the largest regulated euro-denominated stablecoin, with €310 million in circulation at year-end 2025, up 3.8x year-over-year. This is still small in absolute terms, especially when compared to USDC.

As CPN scales cross-border payments, having native stablecoin pairs for on-chain FX (dollar-to-euro, dollar-to-other currencies) becomes a building block for payments infrastructure. Circle was the first stablecoin issuer to register as an Electronic Money Institution under MiCA in the EU, which gives EURC a regulatory head start in Europe.

I’m skeptical on the take-up rate of non-USD stablecoins in an environment where even European, Asian and South American users typically use USDC/USDT.

USYC (Tokenised Money Market Fund)

USYC is a tokenised U.S. Treasury and money market product that Circle acquired through its purchase of Hashnote in January 2025. It grew to approximately $1.7 billion and became the second-largest tokenised money market fund in the world. USYC is designed to serve as yield-bearing collateral in digital asset markets.

Institutions can hold USYC to earn yield and seamlessly swap into USDC when they need cash for trading or settlement. The integration with Binance to accept USYC as collateral in June 2025 was seen as a significant milestone. Circle's long-term vision is that the architecture of "yield-bearing collateral + digital cash" will become the standard for capital markets, on both crypto-native and traditional exchanges.

Personally, I’m not convinced that either EURC or USYC will be significant contributors. I think investors have to view both of these as call options and not ascribe any value to it. For the foreseeable future, USDC is and will remain the core driver for reserve income.

Circle Payments Network (CPN)

CPN is an application layer for cross-border money movement built on top of the stablecoin network. It is essentially the SWIFT of the stablecoin world.

As of February 2026, CPN had 55 enrolled financial institutions (up from 29 in Q3 2025), with 74 more in eligibility reviews and a pipeline of over 500. It was live in 14 markets with 11 more planned. Annualised payment processing volume (PPV) reached $5.7 billion, up 68% since the prior quarter.

The initial use cases for CPN are B2B cross-border merchant settlement (businesses exportign from Asia and importing into emerging and developed markets), along with south-to-south and north-to-south remittance applications.

What makes CPN architecturally different from Visa or SWIFT is that it's an open, extensible network. Third-party developers can deploy modules onto CPN. The first third-party apps are already launching, including a trade receivable credit module. Circle has built a refund protocol (transaction reversibility) and a companion insurance protocol for underwriting reversals. These are composable building blocks that don’t exist in closed payment networks.

CPN is still not monetised yet. Circle is currently focused on growing the network and its flows, with monetisation expected to follow as scale develops. I think this is the right call strategically.

Arc (Layer 1 Blockchain)

This is Circle’s most ambitious and speculative product. Arc is a new Layer 1 blockchain purpose-built for regulated financial activity. Layer 1s are notoriously competitive. Thousands of Layer 1s have been built in the past, but as of today, only a handful have succeeded. Arc’s angle here is to include sub-second deterministic settlement finality, low-cost USDC-denominated gas fees, and configurable privacy controls with opt-in confidential transfers. To put it in layman terms, it means to be very quick, very cheap, payable in USDC for zero complexity, and to allow for privacy.

Arc launched into public testnet in Q4 2025 with over 100 companies participating including Goldman Sachs, Deutsche Bank, Visa, Mastercard, Apollo, AWS, BlackRock, HSBC, and Standard Chartered. The testnet achieved near-100% uptime, half-second settlement finality, and over 166 million total transactions. Mainnet launch is planned for 2026.

Circle is also exploring a native token for Arc to drive utility, incentives, and governance. This hasn't been confirmed, but if launched, it could represent meaningful additional value for Circle, though it also introduces complexity around token economics and regulatory considerations.

The truth is that Arc remains entirely pre-revenue and is a speculative project. It could be a potentially transformative new business segment, but I do not expect it to be a needle-mover anytime soon. I will discuss the bull and bear case for this in later segments.

Developer Tools & Infrastructure

Apart from these consumer-facing products, Circle also provides a full stack of services for developers from programmable wallets, a smart contract platform, Circle Gateway for cross-chain USDC activities, and xReserve for supporting USDC expansion across blockchain ecosystems.

These are underrated tools that help to make the rest of the product stack usable for enterprises, especially those that have never touched blockchains before. These tools are not monetised and therefore won’t show up in financial statements, but are essential to the platform strategy.

5. Moats & Differentiation

Read more

S&P 500: +9.30%
NASDAQ: +14.30%

April was, in many ways, the mirror image of March. Both the S&P 500 and the Nasdaq just posted their best monthly performance since 2020. The S&P 500 closed the month at 7,209.01, a fresh all-time high and its first ever close above the 7,200 threshold. The Nasdaq notched 24,892.31, also a new record. The MAG7 gained over 16% on the month, semiconductors rallied an extraordinary 37%, and communication services led all sectors with an 18.4% gain.

The pivot point was April 7th-8th, when President Trump announced a two-week ceasefire with Iran. As I discussed in previous portfolio reviews, the silver lining in this crisis is the fact that these events could all be erased by a singular individual.

Interestingly, in the final two days of April, the war premium was creeping back in. Brent briefly touched $126 per barrel on April 30th, the highest level since 2022, as US-Iran negotiations stalled, the Strait of Hormuz remained functionally closed, and Trump signalled an extended blockade. Stocks and Oil are up at the same time, which makes it even more confusing. I would not be surprised to see another bout of volatility if the ceasefire formally breaks down.

But as I have done in previous months, my decision in geopolitical uncertainty is always to buy businesses that do well regardless.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +163.05%

S&P 500: +72.48% (Outperformance: +90.57%)

NASDAQ: +99.49% (Outperformance: +63.56%)

Portfolio Performance (YTD)

Portfolio: -11.50%

S&P 500: +4.24% (Underperformance: -15.74%)

NASDAQ: +6.34% (Underperformance: -17.84%)

Portfolio reviews are exclusively for paid subscribers. If you would like to gain access to my portfolio holdings, deep dives, and tons of other articles, do consider subscribing.

I post several articles per month, ranging from extensive 50-60 pager deep dives on the most interesting businesses I can find, to shorter articles on valuations, market musings and specific verticals/geographies.

In the past month, I’ve shared these articles with subscribers:

Free subscribers get access to posts too, of course, with free earnings reviews and occasional articles.

In the next few segments, I discuss my general thoughts on the market, portfolio companies, performance, portfolio allocation and portfolio changes.

General Thoughts on the Market & Performance

Read more

At the time, the stock traded at $77. The stock went on a strong run from then till September 2025, with the most recent high at $199.30. In the past 8 months, the stock has seen an incredible drawdown, trading as low as $77 (again) in late March, down over 61% from 52-week highs.

In this piece, I will give an update on my view of Sea Limited today and my plans for the stock moving forward.

Quick Summary of Initial Sea Limited Thesis

This was my initial thesis for Sea Limited in September 2024 when I published the write-up. Today, I believe the points remain.

Summary Update

In the past quarter, there have been several events and updates worth discussing. Let’s break it down below.

FY 2025 Financials:

This was, by almost all metrics, the best year in Sea’s history. FY 2025 revenue grew 36.4% YoY to $22.9B, while gross profit grew 42.2% to $10.2B, and net income reached $1.6B versus just $447.8M in 2024. Adjusted EBITDA was up 75.2% to $3.4B.

Shopee delivered $16.6B in revenue (+33.4% YoY) and $880.6M in Adjusted EBITDA, up from just $155.8M in 2024. Management shared that Shopee now serves 400 million active buyers and 20 million sellers, with FY 2025 GMV reaching $127B.

Monee was the standout segment in 2025. Revenue grew 60.1% YoY to $3.8B, with Adjusted EBITDA crossing $1B for the first time. The loan book reached $9.2B, up 80.4% YoY, while NPL90+ stayed disciplined at 1.1%. Over 20M unique first-time borrowers were added during the year, reaching a total of 37M active credit users.

Garena also had its best year since 2021, with bookings up 37% YoY to $2.9B and Adjusted EBITDA of $1.7B. There were over 100 million daily active users on average across the entire year.

Capital Allocation:

Management announced a $1B share buyback in November 2025, the first material capital return in Sea's history. I thought this was a significant signal that the company has crossed into a "mature compounder" phase rather than just a hyper-growth story. Cash and equivalents continued growing to over $11B.

2026 Guidance:

Management guided for Shopee GMV to grow ~25% YoY in 2026, which was generally a beat on consensus estimates. They also guided for Monee loan book to grow meaningfully faster than Shopee’s GMV. Garena bookings to grow double digits. Group-level adjusted EBITDA guided to be “at least flat” versus 2025.

Why Is The Stock Down?

Let’s address the elephant in the room. The sell-off, in my view, is a combination of three forces hitting Sea simultaneously.

1. 2026 EBITDA guidance disappointed

I believe the market had priced in continued operating leverage, which would have required growing Adj. EBITDA numbers in 2026. By guiding for “at least flat” EBITDA on top-line growth of 25%, it meant an EBITDA margin of ~0.7% of GMV this year, versus the Street’s prior expectations of ~1.2%.

2. Competition from TikTok Shop

Adding to the fear, TikTok Shop’s growth in the past few years have been nothing short of staggering, leading to fears that it could be wrestling market share from Shopee. J&T Express, the main logistics partner of TikTok Shop has seen an acceleration in growth over the past year or so. The market has largely viewed Shopee’s investment in its VIP program, logistics investment and AI capabilities as an act of desperation to defend market share.

3. Geopolitical Worries + Oil Concerns

The Iran-U.S. war which ultimately led to the SoH closure has hit Southeast Asia with over two-thirds of the region’s crude oil imports transiting through the SoH. About 90% of the oil and 83% of the liquefied natural gas that normally pass through the Strait of Hormuz are bound for Asia. This has led to fears of a slowdown in general consumption across the region.

These headwinds are real and could persist. However, none of these headwinds change the core thesis, as I will discuss below.

Addressing the Concerns

Let me address each of these one by one, because I believe the market is significantly mis-pricing the situation.

On 2026 EBITDA Guidance

The core bear thesis and fear that has shrouded the business due to the weak EBITDA guidance is that the margin compression is structural and permanent. It assumes Shopee is being forced into permanent reinvestment to fend off TikTok Shop and that terminal margins might be closer to 1% of GMV rather than the 3% or higher that bulls had hoped for.

This cannot be further from the truth. When we look at the e-commerce space and where terminal margins might be heading, it’s hard to avoid China, where the space is intensely competitive and probably 5-10 years ahead of Southeast Asia. There are 4 main players in the region, Alibaba, JD, PDD and ByteDance. Despite the fierce competition, it is notable that during the period of competition between 2022 to 2024, they ran at 2.5%, 1.1%* and 1.9% EBITDA/GMV.

*JD’s 1.1% here is an outlier due to its focus on 1P, electronics and home appliances focus where margins are materially lower.

Shopee as a 3P marketplace with a much higher mix of apparel, beauty and long-tail SKUs which carry far higher take rates and margins, should be compared to Alibaba and PDD. As long as TikTok Shop remains rational in pricing and competitiveness, we should see Shopee’s margins trend towards 2% or higher in the coming years. (More on this later)

On TikTok Shop Competition

This is probably the largest concern for investors today, and is directly tied to the above point. Let’s look at what the data says.

Alternative data (Yipit) suggests that TikTok Shop's GMV growth in Southeast Asia has decelerated materially over the past few quarters, from ~70% YoY in Q2 2025 down to ~30% YoY by late 2025. That's only modestly above Shopee's own ~25% YoY growth. More importantly, TikTok's relative market share versus Shopee in Indonesia, the region's largest market, has effectively flat-lined over the past year. (This can be seen in Momentum Works’ report too)

Of course, this data could be wrong. So let’s look at another key tell. TikTok Shop has been raising fees in the past year, following Shopee’s footsteps within days. This is a sign of stabilising competition, not one of an aggressive competitor trying to capture share at all costs. This was also corroborated by management’s comments in Q3 2025 and Q4 2025’s earnings calls where they specifically mentioned:

“What we see is a relatively stable competitive landscape. I think as you can probably observe as well from your own sources, we didn’t see any particular market behaving differently from another. The trends in competitive intensity and behaviour have been fairly consistent across Southeast Asia.”

- Sea Limited Q3 2025 Earnings Call

When we look at absolute numbers, the picture isn’t as scary as the headlines suggest too. TikTok Shop did ~$45B in GMV last year versus Shopee at ~$127B. Internal targets at TikTok Shop reportedly point to ~50% growth for TikTok Shop in 2026, while Shopee is targeting ~25% growth. If both of these targets play out, TikTok would add ~$22B in GMV as compared to Shopee’s ~$32B. Shopee is still growing faster in absolute terms and increasing their competitive lead in terms of logistics.

There is also a more nuanced point worth making here. Most of TikTok Shop's growth is coming from rural areas and Tier 2/3 cities, consumers who are entirely new to e-commerce. Many of these are impulse buys triggered by watching livestreams. Crucially, this is a segment of the population that Shopee had never been able to acquire due to the higher relative shipping costs, lower discretionary income and lower purchase frequency. In this way, TikTok Shop is likely expanding the overall e-commerce TAM and pie, rather than eating Shopee’s lunch directly. As these consumers mature into savvier shoppers, many will inevitably find their way onto Shopee too.

What about J&T’s growth in parcel volume share within SEA, with TikTok Shop being the major contributor?

The most common pushback, which I myself have highlighted, is around J&T Express’ reported parcel volume growth of ~74% YoY in Q4 2025. Let’s break down the growth here:

• Underlying SEA e-commerce parcel market growth: If J&T had simply held share flat at 27% (2024 levels), it would have grown 39% just by riding the tide.

• Per J&T's own 1H25 prospectus filings, its market share rose from 28.6% in 2024 to 34.4% in 2025. That's another ~28% of growth from share gains alone.

• Non-Platform Parcels: J&T's non-platform parcel business has grown materially in the past year and possibly accounts for ~5-10% of growth.

When we put all these together, you can see why J&T’s headline growth number is a misleading proxy for TikTok Shop’s underlying performance. Adding the numbers up, it reconciles pretty well, without requiring TikTok Shop to be growing at a scary rate that the headline implies.

The SEA logistics market is consolidating around J&T as smaller players exit, and J&T is leveraging its scale to expand into adjacent non-platform business lines. Neither of these is a Shopee bear signal.

If anything, the competitive backdrop appears to be improving, not deteriorating. And if that is correct, then Shopee’s current reinvestment cycle could be an offensive choice to widen the moat while competitors are seeing slowing growth, rather than a defensive scramble to stay alive.

On Geopolitics & Oil

The Iran-U.S. conflict and the SoH disruption are real, and it's affecting consumer sentiment across the region. With ~2/3 of SEA's crude oil imports transiting through the SoH, a slowdown in regional consumption is a legitimate near-term concern.

That said, two things are worth keeping in mind. Firstly, this is a macro headwind that affects all consumer-facing businesses across the region. It is not Sea-specific, and Sea is no more exposed than any other consumer business in SEA. If anything, e-commerce tends to be more resilient than offline retail in downturns due to its discount-led value proposition.

Secondly, my thesis on Sea is decade-long. I've said before that I believe Sea will be the first SEA-headquartered trillion dollar business. That trajectory is built on the structural digital adoption of 600M+ people, not on whether oil prices spike for a few quarters. If the macro environment remains weak, Sea will face short-term revenue and credit risks at Monee. But the long-term opportunity is unchanged, and a weaker macro environment historically favours the strongest players in any market, which Sea undeniably is in SEA.

Valuation Today

Read more

In this piece, I wanted to cover something slightly different from usual.

While I often share deep dive research reports, I thought it might be helpful for some of you if I were to document my own research process too.

Here’s my no-nonsense approach to research as a long-term minded investor.

Subscribe now

Step 1: Always start with the business

Before I look into valuation, stock prices, sentiment etc…

I always ask these 5 simple questions:

1. What problem does this company solve?

This sounds basic, but you'd be surprised how many investors can't articulate this clearly for stocks they own. If I can't explain the problem the company solves to a 10-year-old, I probably don't understand the business well enough to invest in it.

The best businesses solve painful, recurring problems that customers can't easily ignore. The more painful and unavoidable the problem, the more durable the business tends to be.

2. How does it truly make money?

Revenue composition is massively important for me. I want to understand the actual mechanics of monetisation. Is it a subscription model with predictable recurring revenue? A transactional model where revenue scales with volume? A platform model that takes a cut of every transaction? An advertising model dependent on eyeballs? Each model has very different characteristics in terms of predictability, scalability, and resilience during downturns.

I dig into the revenue breakdown by segment, geography, and customer type. I want to know if 40% of revenue comes from a single client (a major red flag) or if it's diversified across thousands of customers. I also look at how the company prices its product (per seat, per transaction, usage-based), because pricing power is one of the clearest indicators of a strong competitive position.

3. Why do customers choose this business over other alternatives?

This is where I start thinking about differentiation. I compare the business to its competitors, look through NPS scores, GlassDoor ratings, and more. I also look at churn rates and net revenue retention rates, if existing customers are spending more each year, it is a good indicator that they love the product.

4. What is the company’s moat?

The company’s moat determines how long it maintains its competitive advantage. As a long-term investor, this is possibly the most important question to ask.

I focus on the 5 main moats, network effects, switching costs, intangible assets, cost advantages and efficient scale. There is a 6th that isn’t a traditional moat, but one that I view to be the greatest moat of them all, and that is innovation. The 5 main moats are static moats that have been built in the past. Innovation makes a business a moving target.

The best businesses have multiple moats that are widening over time. I pay close attention to this distinction as a business with a large but narrowing moat is more often that not, a value trap.

5. What would kill this business permanently?

As Charlie Munger said: “Invert, always Invert”. Rather than focus on what can go right, it is sometimes more important to focus on what can go wrong. This question forces myself to imagine a scenario where the stock goes to zero and working backwards from there.

What would have to happen? Regulatory action that bans the product? A technological shift that makes it obsolete? A competitor that offers the same thing for free? Customer behaviour shifting away entirely?

The key word here is “permanently”. Every business goes through cycles of temporary uncertainty, be it recessions, bad quarters, PR crises etc. This question focuses on existential risks. If the only scenarios I can construct for permanent destruction are extremely low-probability events, that's a good sign. If I can easily imagine a realistic path to obsolescence within 5-10 years, that's a dealbreaker for me.

Oftentimes, just answering these 5 questions, gives me a very good understanding of whether to proceed with my analysis. For instance, if a business has no moat, or makes money in a non-sustainable manner, it would make the decision very easy for me. I simply move onto the next.

Step 2: Company History

The next step for me would be to look deeper into the history of the business.

It is my belief that truly understanding the current context of the business and structure of the company is only possible after learning about how the business got to this stage.

In the past, this would include tons of google searches, reading biographies, wikipedia pages, company filings and even books.

However, with LLMs today, it all begins with a prompt. This is what I like to use:

“Give me a comprehensive, chronological company history of [Company Name] from its founding to the present day [Insert today’s date]. I want this to be extremely thorough and detailed, written at the level of a professional investment research report or business case study, not a short summary.

Include and clearly explain the founding story, key founders and early executives, evolution of each business segment, major strategic decisions and pivots, important product launches, capital markets events, crises and downturns, key partnerships and acquisitions, as well as recent strategic direction.

Do not omit uncomfortable details. Treat this as a complete history chapter for an investor who will challenge missing years, unclear leadership changes, and vague claims. Use high-quality primary sources wherever possible (SEC filings, annual reports, earnings call transcripts) and provide citations for important claims.”

Importantly, this is just the starting point for my research. Once I find a thread I can pull on, I then re-prompt to find out more details regarding that specific event and/or individual for instance.

What I’m actually looking for in the company history includes:

• Founder-Market Fit

  • Did the founders start this company because they deeply understood the problem, or was it a pivot from something unrelated? Companies born from genuine domain expertise tend to make better long-term strategic decisions.

• Capital Allocation Track Record

  • How has the company deployed capital over time? Were acquisitions value-accretive or dilutive? Did they buy back stock at good prices or at the top? Did they invest in R&D consistently or slash it during downturns? A 10-20 year track record of capital allocation tells me more about management quality than any single earnings call.

• Crisis Behaviour

  • How did the company navigate past recessions, competitive threats, or operational failures? Companies that emerge from crises stronger, through management decisions like cutting costs intelligently, gaining market share through short-term pain, or making opportunistic acquisitions for instance, tend to be run by exceptional management teams.

• Strategic Consistency

  • Has the company stayed true to its core mission, or has it chased trends? Some pivots have been brilliant (Netflix from DVDs to streaming), but most diversification efforts destroy value. Businesses must have a coherent strategy that ensures they stay the course.

I often feed this prompt into multiple LLMs as they tend to come up with differing approaches and results. Regarding deep research, my favourite* LLMs in order are ChatGPT, Claude, Grok, followed by Gemini.

*Note that this changes often as LLMs update their models and this is as of today’s post.

Step 3: Industry Landscape

Once I’ve grounded myself in the company’s core business and historical evolution, I shift my focus onto the broader ecosystem it operates within.

This is absolutely crucial in contextualising the company’s position.

After all, even a great business can falter if it operates in a shrinking market or gets out-competed by younger, hungrier competitors.

These are the key questions I ask:

1. What’s the total addressable market and the growth trajectory of this industry?

I want to understand both the current market size and, more importantly, the direction. A $50 billion market growing at 15% annually is a very different investment proposition than a $500 billion market growing at 2%. Both can produce great investments, but they require very different frameworks. Personally, as a growth investor, I look for large and growing markets. Another important point is penetration rate; specifically, how much of the TAM has been captured? Early-stage markets with low penetration rates offer the most explosive growth potential, but they also carry more uncertainty.

2. Who are the key players, and how do they stack up against each other?

Here, I map out the competitive landscape in detail. Market share data, growth rates relative to competitors, profitability comparisons, and product differentiation. I want to understand the industry structure. Is it a winner-takes-all market, a duopoly, or a fragmented market that is set for consolidation? Each structure implies different competitive dynamics and margin trajectories.

3. Are there regulatory tailwinds or headwinds?

Regulation can make or break an industry. Some regulations create barriers to entry that protect incumbents (banking, healthcare). Others can disrupt entire business models (data privacy laws impacting ad-tech). I pay particular attention to regulatory direction. Is the trend towards more or less regulation, and how does the company I’m researching stand to benefit or suffer? (Also take note that different geographies have very different regulations by sector)

4. What could potentially disrupt the industry?

I think about this from a first-principles perspective. What underlying technology, consumer behaviour, or economic shift could fundamentally change how this industry works? For example, AI is currently disrupting software per-seat businesses, electric vehicles are disrupting traditional ICE automakers, digital payments are disrupting traditional banks and legacy payments rails, and quick commerce is set to disrupt traditional grocery models and physical retail to a further extent. The key here is to understand whether the company I'm researching is the disruptor, the disrupted, or relatively insulated.

To gather this, I start with high-level overviews from sources like McKinsey reports, Grand View Research reports, industry associations, slide decks from the company or competitors. I also utilise LLMs, with a prompt like this:

“Provide a detailed analysis of the [Industry Name] as of [today’s date], including market size, growth projections (CAGR over 5-10 years), key trends, regulatory environment, and major competitive forces using the most suitable framework. Focus on [Company Name]’s position within it, comparing it to top competitors like [list 3-5 relevant peers]. Use data from primary sources such as Gartner, IDC, SEC filings, earnings reports, commentary and call transcripts, and recent analyst reports, with citations.”

With this data in hand, I begin to gain a much better understanding of the business model, competitive advantage of the particular company, and whether there is a case to be had for investing.

Step 4: Financials

Numbers aren’t everything, but they play a large part, especially with regard to valuation. Also, management and analysts may lie, but numbers don’t lie.

I dive into the financials. My favourite tool here is Fiscal AI. There are of course sources where you can access company financials for free such as Yahoo Finance, Morningstar, etc…

Personally, I use Fiscal AI as they provide up to 20 years of data in a very standardised and easy to use format. I highly recommend using them if you aren’t already doing so.

For those interested, I have an affiliate link that gives you 15% off the subscription pricing. Or, simply use the free version which works great too.

Here, I review the last 5-10 years of the 3 core financial statements (Income Statement, Balance Sheet, Cash Flow Statement). I focus particularly on top-line growth rate, which I utilise to judge which part of the business life cycle it is in.

My favourite businesses to invest in lie within self funding and operating leverage. These are typically businesses that are inflecting towards profitability or are at the early stages of operating leverage in the business.

Another key focus is with regard to margins. For industry disruptors, I would compare to industry heavyweights to get a gauge of where the optimal margin profile would be. This will be very helpful in the valuation stage of the analysis.

Other key areas of focus:

• Revenues: Are they recurring or one-off? Is there geographic diversity? Is there significant customer concentration?

• Profitability: Have margins expanded over time? What is the optimal margin profile for the business? Is there room for operating leverage?

• Efficiency & Returns: This goes more for businesses that are in Stage 4 and 5 of the life cycle chart above. I focus on ROIC/ROE, working capital cycles. For tech businesses, I look at R&D/CAPEX as % of revenue.

• Balance Sheet: How much cash does the business have? Are their debt levels elevated? Is it a business that requires enormous leverage?

• Cash Flows: Especially if the business is unprofitable, I want to ensure they are free cash flow positive or reaching there imminently.

• Stock Based Compensation: Is it egregious? How is management compensated, is it based on top or bottom-line or stock price performance?

This step takes the longest, but is non-negotiable for me. I’ve avoided many losses by spotting eroding fundamentals early.

Step 5: Management

Of course, numbers only tell one side of the story. Management is arguably the most important for me.

Think of all the successful businesses over the years. Each of them have been run either by extraordinary entrepreneurs or ruthless executioners. Think Apple, Microsoft, Tesla, Google, Nvidia. Each of these businesses are closely tied to one or two individuals who have been the face for decades.

Ultimately, businesses are people-driven, so I scrutinise the stewards of capital very closely.

I profile the C-suite and board:

• Is the founding team still in-charge?

• Read as many news reports, especially if they are first-hand interview transcripts.

• Watch as many interviews of the C-suite/Board as possible, going back 10-20 years if possible. Understand how their views have changed over time. Are they trustworthy people?

• Is the board filled with “Yes Men”? Or do they constantly challenge decisions? I want a board that is intellectually honest and willing to be contrarian.

• Does the C-suite have skin in the game? Is the company stock the majority of their net worth?

• How do they handle tough questions? Do they give direct, thoughtful answers, or do they deflect, ramble and redirect to pre-planned talking points?

• Do they under-promise and over-deliver, or vice-versa? I go back through old earnings call transcripts and compare forward guidance to actual results.

• How do they discuss failures? Every company has bad quarters, failed products, or strategic missteps. Leaders who own these mistakes openly and explain what they learned earn my trust. Those who blame macro conditions, one-time events, or "transitory" factors for repeated misses do not.

Step 6: Valuation & Scenario Modelling

The final decision to invest, often comes down to valuation. More specifically, investors should warrant a margin of safety when investing.

Valuation isn’t precise, but is often probabilistic. There are a huge variety of different methods to value a business, but I do believe that every business has an optimal way of being valued.

For instance, diversified conglomerates like Sea Limited and Mercado Libre, in my view, should be valued through a Sum-of-the-Parts valuation model. DCFs are generally one of my favourite methods and is considered by many to be the gold standard. Multiples is another method to value, through P/E, P/S, EV/EBITDA and other various ratios.

Valuation often determines the amount of size I deploy into a stock, and I think it is absolutely key to nail this down.

Choosing the Right Valuation Method:

Discounted Cash Flow (DCF):

The DCF is generally one of my favourite methods and is considered by many to be the gold standard. The core premise is simple: a business is worth the sum of all its future cash flows, discounted back to present value. However, this is a very subjective process and that is where numbers tend to differ.

My approach to building a DCF:

• Revenue projections: I start with the company’s own guidance, then cross-reference with industry growth rates, competitive dynamics, and historical trends. I build a 10-year explicit forecast period, with revenue growth rates that gradually decelerate towards the industry’s long-term growth rate. The key is to temper expectations and never assume hyper-growth persists indefinitely.

• Margin assumptions: I project margin expansion (or contraction) based on the operating leverage analysis from Step 4, benchmarked against mature peers. In certain cases, where I expect a disruption in the business model, terminal margins could be higher than

• Discount rate: I use WACC (weighted average cost of capital) as my discount rate. For most established companies, this falls in the 8-12% range. For higher-risk growth companies, I use a higher rate (12-15%) to account for the additional uncertainty. Some investors use a flat 10% as a simplification. I don’t disagree with this approach as long as you’re consistent.

• Terminal value: This is the most important and most dangerous part of the DCF, as it typically drives 60-80% of the total valuation. I use both an exit multiple approach and a perpetual growth rate approach, and I check them against each other. I never use a terminal growth rate above 3-4% (roughly in line with nominal GDP growth), and I sanity-check the implied exit multiple to make sure it’s reasonable relative to the company’s peer group.

Multiples-Based Valuation:

Multiples is another method to value businesses, through P/E, P/S, EV/EBITDA, and other various ratios. While simpler than a DCF, multiples can be misleading if used sloppily. The key is selecting the right multiple for the right type of business:

• P/E works well for mature, profitable companies with stable earnings.

• EV/EBITDA is better for companies with varying capital structures, as it strips out the effects of debt and taxes.

• P/S (price-to-sales) is useful for high-growth companies that aren’t yet profitable, but should always be used alongside a clear path-to-profitability analysis.

• P/FCF (price-to-free cash flow) is my preferred multiple for companies that are already generating meaningful cash flow as it cuts through accounting noise.

I personally always compare multiples to historical averages and peer comparisons. For instance, where has the stock traded over the last 5-10 years? How does this compare to similar companies in the space? A stock trading at 30x earnings might look expensive in a vacuum but could be cheap if the company is growing twice as fast as peers trading at 25x.

Sum-of-the-Parts (SOTP):

For diversified conglomerates like Sea Limited and Mercado Libre, I believe SOTP is the most appropriate approach. These companies have distinct business segments with very different growth profiles, margin structures, and comparable companies. Valuing them on a blended multiple obscures the true value of each piece. I value each segment independently using the most appropriate methodology (DCF or multiples), sum the values, subtract net debt, and divide by shares outstanding.

Valuation often determines the amount of size I deploy into a stock, and I think it is absolutely key to nail this down.

Step 7: GabGrowth Quality Score

I use a checklist scoring system to rate stocks. The best investments have common, durable traits. I’ve used what I found over many years to come up with this and enables me to separate the good from great businesses.

I’ve written a full piece (FREE) on this that you can view below.

Step 8: Portfolio Fit & Final Decision

Ultimately, the final decision comes down to this step. While I’m not against investing a huge portion of my portfolio into a particular sector if I believe there are significant tailwinds going for it, I do need to compare the new company I’m adding to others that are already in the portfolio.

Every dollar I allocate to a new position is a dollar I'm not adding to an existing holding. So the new company doesn't just need to be "good", it needs to be better than adding to my current highest-conviction positions.

Regarding position sizing, I use a tiered approach based on conviction level and risk profile. For a high-conviction idea with a high margin of safety, I’m willing to size it at over 5% of the portfolio to start and letting it compound from there. I also tend to size up over time as the business proves itself through strong earnings beats. Each quarterly report is an opportunity to test the thesis. If the company is executing as expected or better, I'll add to the position on pullbacks. If the thesis is deteriorating (for e.g., slowing growth, compressing margins, poor management decisions), I’ll reduce or exit regardless of what the stock price is doing.

I will let my positions grow over time and am not afraid for a single name to become a substantial portion of my holdings. However, I will typically try to limit any one holding to around 30% of total portfolio size. Beyond that, the position-specific risk becomes too large regardless of conviction. If a position naturally grows past 30% through price appreciation, that's when I start trimming.

Final Thoughts

I hope this sharing is helpful for some of you who would like to get into the habit of researching stocks. I’ve always believed in the Peter Lynch saying:

“The person that turns over the most rocks wins the game”

This process is tedious and extremely tiring, but also very rewarding. A thorough analysis takes me anywhere from a few days to several weeks, depending on the complexity of the business. But rushing through it defeats the purpose entirely. The whole point is to build conviction, as it is the only way to hold through volatility, add on weakness and ignore the noise.

A large part of my investing success has been putting in the hours to study businesses. Less than 10% are investable ideas, but every single one has lessons to be learned. The companies I pass on sharpen my pattern recognition for the companies I eventually invest in.

Thank you for reading!

-Gab

These are not banks, nor PayPal clones, and definitely not crypto projects.

Instead, they are ecosystems that began as ride-hailing apps or e-commerce platforms with the ultimate goal of becoming the primary financial institution for hundreds of millions of people who have never had a bank account to start with.

The scale of the opportunity is massive. More than 70% of Southeast Asia’s adult population remains either unbanked or underbanked, lacking access to credit, insurance, investments and even basic savings products.

The companies who recognised this earliest have now built substantial businesses off the back of this, compounding at extraordinary rates.

Monee, Sea Limited’s FinTech arm, grew its loan book by over 80% in 2025 to reach $9.2 billion. GrabFin, Grab’s FinTech arm ended the year with a loan portfolio larger than $1 billion. GCash has over 90 million registered users in the Philippines, a country of 115 million people. MoMo in Vietnam has over 40 million users while Indonesia has a flourishing FinTech landscape with Monee (owned by Sea Limited), GoPay (owned by Gojek), Dana and OVO (owned by Grab) sharing the spoils.

In this piece, we will discuss the Southeast Asian FinTech landscape. Specifically:

• Who the players are

• How they differ structurally

• Where the competitive edge is being built

We will then move from the macro context down to individual company profiles, and close with a framework for thinking about this space as an investor.

Why This Market Is Different

Most Western fintech narratives are about convenience. The winners are typically players that make it slightly easier to move money that people already have access to.

However, in Southeast Asia, it’s a different story. It is more about access than convenience. For a delivery driver in Jakarta or a market vendor in Ho Chi Minh, a digital wallet is not just a better version of what they already have, but rather the only formal financial product they have ever used, or have access to.

To make this clear, here’s an example. In the West, companies like PayPal, Block or Apple have succeeded by making it slightly easier or cheaper to move money that people already had sitting inside the banking system.

In Southeast Asia, most consumers did not have bank accounts, credit histories, or even reliable ways to store money digitally.

That is why companies like Sea Limited, Grab and GCash did not start by competing with banks. They started by solving everyday problems such as ride-hailing, food delivery or e-commerce. Only after they had millions of users transacting on their platforms did they begin layering financial services on top.

Why does this distinction matter so much?

When a company acquires a customer in an underserved market, the switching costs are high because there is no alternative. There is often no incumbent to displace (the incumbent is cash).

Because these platforms often provide the first credit record a user has ever generated, the data moat compounds with every transaction. Every loan taken and repaid, wallet top-up, and transaction taken slowly builds a credit profile that traditional banks have no access to and competitors find difficult to replicate.

The regulatory environment has been much more constructive than many probably expect. While Southeast Asia is much less developed than most countries globally, many of the governments (Indonesia, the Philippines, Singapore, Vietnam, Malaysia) have all introduced digital banking license frameworks in recent years that give ecosystem players like Grab and Sea a regulated pathway into deposits and lending without having to acquire legacy banks.

This has allowed SuperApps like Grab and Sea to formalise their financial operations, fund their loan books at lower cost, and access interbank liquidity, dramatically improving the unit economics of lending over time.

The Players

There are several players in this space, some of which operate in individual countries while others dominate across the region’s key markets. Each of these companies have a different anchor product, value proposition and competitive moat. In this section, I intend to breakdown the differences between each of them.

Monee (Singapore, Malaysia, Indonesia, Philippines, Thailand, Vietnam, Taiwan, Brazil)

We have to start with Monee. It is by almost any measure, the most advanced digital financial services operation in the region. Monee began as AirPay, a simple top-up feature for Garena and subsequently ShopeePay, a wallet bolted onto Shopee’s e-commerce platform. However, it has since evolved into a full-fledged consumer and SME credit business, which is now Sea’s fastest-growing division and perhaps the most strategically important one.

Sea runs 3 main divisions. The largest, Shopee is the leading e-commerce business in the region with over 400 million active buyers. The most profitable division is Garena, their gaming unit that is responsible for the most played mobile game in the world, Free Fire. Monee is the newest segment and the fastest growing.

Since 2016, Monee has grown its revenues at a 105% CAGR, roughly doubling each year. Last year, it brought in $3.8B in revenues, with a 25.6% operating margin on a segment basis. This meant it accounted for nearly half of Sea’s entire EBIT for 2025.

At the end of 2025, Sea disclosed that it had over 37 million credit users on its platform. Importantly, despite the massive user base, there is still much room for growth. Shopee has 400 million active buyers and 20 million sellers that generate a continuous stream of target users for the business.

Apart from this, this massive user base also generates a constant stream of behavioural data that is likely to be immensely useful for Monee. For instance, purchase frequency, basket size, repayment history on Shopee PayLater, merchant cash flow patterns, which Monee can use to underwrite credit at a scale and accuracy that no traditional lender can match.

Sea now has a loan book in excess of $9.2 billion, while growing at 80% YoY. Despite the high growth rates, Sea maintains a very healthy 90-day non-performing loan (NPL) ratio of just 1.1%. This is a signal that the credit model is working even as the platform pushes into broader applicant pools and non-Shopee channels. It also has a deposit base of $3.8B, up 40% YoY from the previous year’s $2.7B.

A notable strategic shift in 2025 was the move from an invite-only lending model to an "all can apply" approach. Management opened the doors to everyone, and default rates did not move. They attribute this to Shopee's transaction data feeding AI underwriting models that process purchase patterns, return behaviour, payment timing, and browsing history through transformer-based models trained on long-sequence behavioural data.

This represents a 46% deposit-to-loan ratio which means nearly half of the on-book loan portfolio is being funded by deposits that probably cost Monee somewhere in the 1-3% range. The other 54% is funded out of Sea’s $11B cash pile or other sources, capital that could otherwise be earning a return elsewhere.

Management also noted on the most recent Q4 2025 earnings call that AI-assisted underwriting is enabling expansion to borrower profiles that were previously considered too risky.

Of note, in terms of risk, is that the $9.2B is largely on-book loans, which means a relatively capital-intensive structure. Just $1B of that $9.2B reflects a capital-light structure where Monee originates and services loans while offloading credit risk to third-party investors.

The off-Shopee expansion is worth watching closely. Off-Shopee SPayLater loans surged over 300% year-on-year in Q4 2025 and now make up more than 15% of the portfolio, with roughly 30% usage penetration in Malaysia. This is the key metric for whether Monee eventually gets valued as an independent fintech or remains tethered to Shopee's GMV growth rate. The BCL (Buy-Consume-Later) product for offline retail and direct consumer financial products are the vectors for this expansion.

One area that deserves attention is credit loss provisions, which surged 76.7% to $1.4 billion in FY2025. This outpaced even the 80% loan book growth, which is why Adjusted EBITDA grew only 40% on 60% revenue growth. The NPL ratio is stable, but the provision amount has gone up massively. This has not been stress-tested by a severe regional macroeconomic downturn or rapid FX deterioration. The loan book is largely unsecured consumer credit across emerging markets (Southeast Asia and Brazil), and it is approaching $10 billion in scale. As long as the NPL ratio holds, the provision build is just the cost of growth. If it does not hold, the capital intensity of the on-book model could become a vulnerability.

Long-term, the ideal scenario would be for Monee to become a marketplace for credit, growing the off-book ratio (currently just $1.0 billion of the $9.2 billion total) and enabling it to improve return on equity as the book scales. Management has guided for Shopee GMV growth of 25% in 2026, with Adj EBITDA no lower than 2025 in absolute terms. This implies Shopee margin compression as subsidies ramp, but Monee's continued scaling should partially offset that at the group level.

GrabFin (Singapore, Malaysia, Indonesia, Philippines, Thailand, Vietnam)

The second largest player that also operates on a regional basis is GrabFin, a subsidiary of Grab. Grab is the leading ride-hailing and deliveries player in the region with over 50 million monthly transacting users on its platform. It commands nearly 60% market share in both categories and is a household name.

Grab's financial services strategy is structurally different from Monee's in that it is deliberately multi-entity. Rather than building everything under a single fintech brand, Grab operates a layered system. GrabFin handles lending to its ecosystem partners (drivers, merchants, users), while GXS Bank (Singapore, with Singtel) and GX Bank (Malaysia) operate as fully licensed digital banks, and Superbank (Indonesia, 37.88% effective equity stake) handles deposits and credit locally.

Grab ended 2025 with a gross loan portfolio of $1.3 billion, with a net loan portfolio of $1.18 billion after provisions, up 120% year-on-year. Customer deposits across GXS Bank and GX Bank grew to an all-time high of $1.6 billion at end-Q4, up from $1.2 billion a year earlier. Notably, deposits dipped to $1.31 billion in Q3 2025 from $1.54 billion in Q2 as Grab proactively cut promotional deposit rates to manage funding costs. The Q4 recovery to $1.6 billion on the back of customer growth rather than rate competition is a healthier trajectory.

The deposit base now serves 7.4 million customers across the three banks. Importantly, there is a very clear cross-sell efficiency here. 76% of GXS Singapore depositors, 93% of GX Bank Malaysia depositors, and 60% of Superbank Indonesia depositors were already existing Grab users. This means Grab is converting super-app engagement into banking relationships with minimal incremental customer acquisition cost.

Revenues have grown at a 75.6% CAGR since 2022. However, Grab’s financial services segment remains unprofitable with a FY2025 adjusted EBITDA of -$110M. This is in comparison to Monee’s FY2025 adjusted EBITDA of $1.02B.

There are several reasons for this disparity.

Firstly, Monee had a five-year head start. Sea started building AirPay in 2014 for Garena top-ups. By 2019, they rebranded to ShopeePay and began embedding credit directly into checkout flows. By the time Grab got its Singapore banking license in 2022, Monee already had a multi-billion dollar loan book across seven markets.

Secondly, Monee is not just a bank that happens to be connected to an app. It has designed itself to be a financial layer embedded inside Shopee where credit exists to drive more Shopee transactions. SPayLater increases the basket size and purchase frequency, and seller loans keep inventory flowing. Therefore, every Shopee user (buyer and seller) is a potential Monee customer without any incremental customer acquisition cost.

GXS is structurally a standalone digital bank that happens to be owned by Grab and Singtel. It can be accessed through the Grab app and uses Grab’s data for underwriting. However, the banking product doesn’t exactly drive more rides or food deliveries the way that SPayLater for instance drives more Shopee orders. There is no closed-loop commerce flywheel to drive deposits and loans. Hence, GXS has to compete for deposits against the likes of traditional banks like DBS, UOB and OCBC in Singapore by offering promotional rates which is a traditional banking acquisition strategy, not an ecosystem play like Monee.

Thirdly, Shopee processes 13.9 billion orders a year across 400 million active buyers and 20 million sellers. Each of these transactions generate SKU-level purchasing data, repayment behaviour on SPayLater, seller cash flow patterns etc… that provide granular data that is then used to map to creditworthiness.

Grab also has extremely useful data such as how often rides are taken, food order sizes, discretionary spend, but it is fundamentally consumption data on a much smaller scale. This can be seen in GMV. Shopee generated $127B in GMV compared to Grab’s total on-demand GMV of $22B in 2025. That is nearly 6x difference and explains why the FinTech gap is so wide.

GCash (Philippines)

GCash is the leading digital wallet in the Philippines. It is backed by Globe Telecom and Ant Group (Alibaba’s Financial Arm), and has over 90 million registered users, in a country of 115 million people. They are effectively a national payments infrastructure.

It was first launched in 2004 as a SMS-based money transfer service under Globe Telecom. This was a critical service at the time with ~80% of Filipinos being unbanked. They employed a very simple model where users could convert cash to e-money via sari-sari stores (think small, neighbourhood retail stores) for a transaction fee of about ₱1 (about 1-2 cents in USD).

It expanded into online payments in 2006, relaunched as a fuller mobile wallet in 2009, and partnered with Western Union in 2010 for inbound international remittances (a big deal given the massive Overseas Filipino Worker economy).

The strategic inflection came in 2015 when Mynt was established as a joint venture between Globe, Ayala Corporation, and Ant Financial (now Ant Group). The Ant partnership brought the Alipay playbook to the Philippines. By 2017, GCash launched the country's first mobile money QR payment system, and by 2018 it had surpassed 10 million users. In 2019, GCash evolved from a payments tool into a nascent SuperApp with GSave (savings via CIMB Bank), GCredit (revolving credit), and GInvest (investments).

The COVID pandemic was GCash’s defining moment. Usage surged as millions of Filipinos used GCash to pay for goods and services online and to send/receive money. GCash also partnered with the Philippine government to distribute pandemic financial aid which was massive for user acquisition.

In November 2021, Mynt raised $300M at a $2B valuation, making it the first double unicorn in the Philippines. In August 2024, GCash secured additional funding from Ayala Corporation and MUFG pushing its valuation up to $5B.

Today, 8 in 10 Filipinos use GCash, with 6 million merchants and social sellers accepting payments. In 2025, GCash disbursed ₱323 billion (USD $5.4B) in loans to 10.2 million unique borrowers, a 73% increase year-on-year. GSave serves 15.3 million users, GFunds has 8.6 million, and GCrypto enables 4.4 million to trade digital assets.

GCash is targeting an IPO in the second half of 2026, delayed from an earlier timeline due to the Philippine stock market's weakness and the company is aiming to raise $1 to $1.5 billion, which would make it the largest IPO in Philippine history.

GCash is basically the closest analog to Alipay in Southeast Asia, though it is only present in the Philippines. The real challenge facing GCash now, is whether it can sustain its lending growth and ARPU expansion as it moves from payments (low-margin, high-volume) to financial services (higher-margin, credit risk).

Maya (Philippines)

Maya is perhaps the most under-appreciated player in this landscape. It is backed by PLDT (the Philippines’ largest telco), Tencent, KKR and the IFC (Private arm of World Bank Group). Rather than building financial services on top of an existing SuperApp, Maya obtained a full digital banking license from the central bank of the Philippines in Sept 2021 and converted from a payments wallet (PayMaya) into a regulated digital bank in 2022. It is perhaps the only player in this piece that is a bank first and a platform second.

What makes Maya worth watching closely is the pace of their financial ramp in recent years. Maya Bank launched in 2022 and disbursed just ₱3.1 billion (USD $52M) in loans that year. In 2023, disbursements jumped to ₱21.5 billion (USD $358M). In 2024, ₱68 billion (USD $1.1B). In 2025, approximately ₱164 billion (USD $2.7B), bringing cumulative disbursements since launch to ₱256 billion (USD $4.3B) across 3 million borrowers. That is roughly 141% annual growth in disbursement volume, to a borrower base that grew 95% year-on-year.

The deposit base has tracked a similar trajectory: ₱25 billion (USD $417M) at end-2023, ₱39 billion (USD $650M) at end-2024, and ₱68 billion (USD $1.1B) at end-2025, a 72% increase over the year. The bank doubled its customer base to 10.7 million in 2025. The loan-to-deposit ratio reached 51.1% in Q1 2025, which signals disciplined capital deployment, and the net interest margin widened to 18.9% from 13.9% a year earlier (as of September 2025). That NIM figure is extraordinary by any global standard, and speaks to the spread available in a market where the borrower base has historically had no access to formal credit at all.

It is important to distinguish between disbursements and the outstanding loan book. Maya's cumulative disbursements of ₱256 billion (USD $4.3B) reflect total loans ever made. The outstanding loan book, the actual on-book credit exposure at any point in time, stood at ₱27 billion (USD $450M) as of September 2025, up 59% year-on-year. Most of Maya's lending is short-duration consumer credit (Pay in 4, Flexi Loan, Personal Loan up to ₱250,000 / ~USD $4,200) that turns over quickly, which is why the outstanding balance is a fraction of cumulative volume.

Credit quality is the one area that needs honest scrutiny. In Q1 2025, Maya reported an NPL ratio of 3.8%, comfortably below the Philippine digital banking industry average. By end-2025, that ratio had climbed to 6.1%. That is a meaningful deterioration over the course of a single year and reflects the reality of pushing into broader, riskier borrower pools. It is still not alarming in context. 59% of Maya's borrowers had never accessed formal credit before using the platform, which means there is no prior credit history to underwrite against. The company is building credit profiles from scratch using AI-driven scoring on transaction data. But the direction of the NPL trend matters, and if it continues rising as lending scales into auto loans, housing loans, and supply chain financing (all of which management has flagged as 2026 priorities), that will be the metric to watch.

Maya posted ₱1.7 billion (~USD $28M) in net income in 2025, its first full year of profitability. It booked ₱1.6 billion (~USD $27M) of that in the first nine months. Revenue growth exceeded 100% year-on-year in 2024, and the Q1 2025 top-line accelerated further. This came just three years after the bank launched, which is a remarkably fast path to profitability for a digital bank operating in an emerging market.

On the payments side, Maya processed over ₱1 trillion (~USD $16.7B) in merchant payments in 2024, making it the largest merchant acquirer among Philippine digital banks. Visa has recognised Maya as the top acquirer for merchant transaction volume in the Philippines. The merchant acceptance infrastructure, built initially through the Maya Center agent network (formerly Smart Padala, which reaches deep into provincial and rural areas), gives Maya physical distribution that pure digital players struggle to replicate. 70% of Maya's customers live outside Metro Manila, and regional loan drawdowns surged 137% year-on-year in 2024.

The competitive dynamic between Maya and GCash is extremely tight, no doubt fuelled by the backing of Ant Group and Tencent. GCash has the user base and dominates peer-to-peer transfers and bill payments, while Maya has the banking license and deposit franchise, winning on deposit growth and credit product depth. Each product launched by one forces a response from the other, which has resulted in the Philippines leading the region on digital financial inclusion metrics.

Maya intends to IPO in 2026, targeting a dual listing, first in the US and then on the PSE in the second half of the year. If Maya lists at anything close to the multiples that Southeast Asian FinTech names have attracted in recent years, it would be the largest Philippine-origin tech IPO in history and would create a direct public market comp for GCash ahead of its own expected listing.

MoMo (Vietnam)

MoMo is Vietnam's dominant digital wallet, with over 40 million registered users in a country of 98 million people. It is a rare case of a Southeast Asian fintech that built scale without a SuperApp anchor.

MoMo grew as a standalone financial super-app, aggregating payments, lending, insurance, and investment services under a single brand without the benefit of a ride-hailing or e-commerce parent.

The company was founded in 2007 (as M_Service) and launched as an e-wallet in 2010. It has raised $434 million across five rounds, with the most recent being a $200 million Series E in December 2021 led by Mizuho, which valued the company at $2 billion. The cap table includes Warburg Pincus, Mizuho, Standard Chartered, Goldman Sachs, Goodwater Capital, and Kora Management.

MoMo’s current chairman is Anthony Thomas, the former CEO of GCash, which gives MoMo a direct connection to the operational playbook of one of the region's most successful digital wallets.

MoMo turned profitable for the first time on a full-year basis in 2024, with revenue growing 27% year-on-year. Revenue had grown from approximately VND 6 trillion (~USD $240M) in 2020 to VND 8.5 trillion (~USD $340M) by 2022, and the 2024 figure is likely north of VND 11 trillion (~USD $440M) based on the growth trajectory.

The company partners with over 50,000 businesses and more than 70 banks and financial institutions, with 300,000+ payment acceptance points nationwide. Its QR code infrastructure reaches into rural markets and traditional wet markets that international platforms have not penetrated.

In October 2024, MoMo repositioned itself from an e-wallet to an "AI-Powered Financial Assistant," signalling a strategic shift toward higher-margin financial services. The platform now offers micro-loans (through bank partnerships), insurance, investment products, stock trading, and BNPL with up to 45 days deferred payment. However, MoMo does not hold a banking licence and does not take deposits directly, which is a meaningful structural limitation compared to Maya or GXS. Its lending is intermediated through partner banks, meaning MoMo earns origination and servicing fees rather than net interest income.

The competitive landscape in Vietnam is more fragmented than in the Philippines or Indonesia. MoMo leads with an estimated 47-56% share of e-wallet users depending on the measure, but faces competition from ZaloPay (backed by VNG, Vietnam's largest tech company), ShopeePay (Sea Limited), VNPay (which dominates QR merchant acquisition for banks), and ViettelPay (backed by the military-owned telecom Viettel). The government's VietQR interoperability mandate, which has unified QR payments across 2.1 million merchants, has been a structural tailwind for adoption but also reduces the switching cost advantage that any single wallet player can build.

OVO (Indonesia)

OVO is arguably the most structurally interesting e-wallet in Indonesia. It was originally launched by the Lippo group (one of the largest real-estate conglomerates in Indonesia). It was then adopted by Grab as its Indonesian payments wallet in 2018 after Grab was unable to secure its own e-wallet license from Bank Indonesia.

In 2018, Grab invested in OVO, garnering roughly 40% share in the business. In 2021, it increased its stake to over 90% by purchasing shares from PT Tokopedia and Lippo Group.

OVO’s strategic significance lies in its integration with Superbank, the Indonesian digital bank backed by Grab, Emtek, Singtel and KakaoBank. In May 2025, OVO launched OVO Nabung by Superbank, a product that embeds a regulated bank savings account inside the OVO e-wallet. Users earn 5% annual interest on their balances with no minimum or fees, and can seamlessly use those balances for daily transactions. The product crossed one million users within five months of launch. Users who upgraded to OVO Nabung saw their average balances double and transaction frequency increase by 60%. One in three Grab driver-partners now uses OVO Nabung as their primary wallet and savings tool.

OVO Nabung creates a closed loop where OVO wallet users earn bank-level yield on their balances while Superbank gets sticky deposits with minimal acquisition cost, which it can then lend against. Superbank itself turned profitable for the first time in FY2025 (net income of 99.7 billion rupiah / ~$5.9M), with loans growing 50%, NIM of 10.64%, and gross NPL of just 2.6%. The cost-to-income ratio collapsed from 139% to 70.5% in a single year.

By end-2025, OVO had reached 121 million downloads and 3 million QRIS merchants across more than 800 cities and regencies. Over 40 million QRIS transactions were processed through the Grab ecosystem in 2025, with QRIS adoption growing 61% year-on-year across OVO's merchant base. More than 700,000 SME merchants saw transaction volumes increase by 35% through digital adoption.

Indonesia's QRIS system has been a structural tailwind for the entire wallet ecosystem. Bank Indonesia data shows QRIS is now used by nearly 60 million users across more than 40 million merchants (mostly SMEs), with transaction volumes growing 162.7% year-on-year as of July 2025. Gen Z accounts for 28% and millennials for 26% of total QRIS users.

The key question for OVO is whether the Superbank integration can scale fast enough to establish a deposit-funded lending flywheel before competitors catch up. GoPay is integrated with Bank Jago (GoTo's digital banking partner). ShopeePay has Monee behind it. Dana has the Sinar Mas banking relationship.

Indonesia's 277-million person market is large enough to support multiple winners, but the wallet that converts the most users into depositors and borrowers first will have the most defensible position. OVO's early traction with OVO Nabung suggests it is ahead on this metric but I believe this race is far from over.

GoPay (Indonesia)

GoPay is the fintech arm of GoTo Group, formed from the merger of Gojek and Tokopedia in 2021. It is the most direct competitor to OVO in Indonesia, and alongside ShopeePay, one of three wallets with a captive SuperApp ecosystem in the country.

GoPay launched in 2016 as a payments solution for Gojek users and driver-partners. It has since spun out into a standalone app (launched 2023) designed to reduce data usage and reach mass-market users with less powerful phones.

The standalone GoPay app has seen a growth inflection point, monthly transacting users reached 24.2 million in Q3 2025, up 29% year-on-year, and GoTo surpassed 500 million transactions in a single month for the first time in September 2025. By Q4 2025, digital payments volume hit IDR 120 trillion (~$7.5B) for the quarter.

It also has a relatively large lending business. GoTo Financial's consumer loan book expanded rapidly through 2025, reaching IDR 7.6 trillion (~USD $457M) by Q3 2025, up 76% year-on-year, with management guiding to exceed IDR 8 trillion (~USD $480M) by year-end. For the full year, GoPay's Fintech segment achieved Adjusted EBITDA of IDR 497 billion (~USD $30M), its first full year of profitability. Lending revenue grew 130% year-on-year to IDR 879 billion (~USD $53M) in Q2, and 84% year-on-year to IDR 1.0 trillion (~USD $60M) in Q3. Roughly 60% of loans outstanding originate through the GoPay and Gojek apps, with those in-ecosystem channels growing faster than third-party platforms.

The TikTok partnership has added an unexpected distribution channel. After TikTok acquired 75.01% of Tokopedia in late 2023, GoPay became embedded in the TikTok Shop checkout flow. GoTo launched GoPay Pinjam (instant cash loans) directly within TikTok Shop in 2025, making it the first platform in Indonesia to offer cash lending inside the TikTok ecosystem. BNPL on TikTok Shop and Tokopedia has shown strong momentum, and the partnership with Telkomsel (Telkomsel Wallet by GoPay, embedded in the MyTelkomsel app) further extends distribution.

GoTo holds a 22% stake in Bank Jago, an Indonesian digital bank that serves as GoPay's banking partner. The GoPay Tabungan by Jago product, launched in late 2023, embeds a Bank Jago savings account inside the GoPay and Gojek apps, mirroring the OVO Nabung by Superbank model. Users can open an account in two minutes, earn interest on balances, and use them seamlessly for payments across the GoTo ecosystem and all QRIS merchants. GoTo has also launched GoPay Asuransi (insurance), GoPay Simpanan (investments), and GoPay Pinjam BPKB (collateral-backed loans) on the consumer app, progressively building a full financial services stack.

GoPay turned profitable on an Adjusted EBITDA basis for the first time in FY2025, reaching IDR 497 billion (~USD $30M) for the full year. Financial services revenue nearly doubled in 2024 to IDR 3.7 trillion (~USD $230M). GoTo itself posted its first pre-tax profit in Q3 2025 (IDR 62 billion / ~USD $3.4M) and raised its FY2025 Adjusted EBITDA guidance to IDR 1.8-1.9 trillion.

GoTo's fintech segment is arguably the most under-followed listed fintech asset in the region, given that most international investor attention goes to Sea Limited and Grab. Unfortunately, I do not think GoTo as a whole is investable, due to the declining ride-hailing and food delivery business that makes up a large majority of group revenues.

DANA (Indonesia)

DANA is a digital financial services business based in Indonesia that was founded in 2017 as a joint venture between Emtek (one of Indonesia’s largest media conglomerates) and API Investment Limited, an investment arm of Ant Group. It launched in 2018, later than competitors such as GoPay and OVO but has since built scale.

It first got its start through being deeply integrated with Bukalapak (Emtek-backed Indonesian e-commerce player) and Lazada (Alibaba-owned). In 2022, Lazada purchased $304.5M of DANA shares from Emtek and Sinar Mas Group invested a further $225M to make it one of the largest shareholders.

DANA claims to have over 200M registered users today and as of 2023, processed an average of 10M transactions each day. DANA’s structural challenge is that it lacks a captive ecosystem. GoPay has GoTo (Gojek + Tokopedia) while OVO has Grab, and ShopeePay has Shopee. DANA’s e-commerce partners have either shut down or are not exclusively tied to it. This makes DANA the most exposed of each of Indonesia’s major wallets to the commoditising effect of Indonesia’s QRIS interoperability mandate, which requires all wallets to accept a unified QR standard across over 40 million merchants. (When every wallet works everywhere, wallets with the weakest ecosystem lock-in will be the most vulnerable)

That said, the Ant Group and Sinar Mas backing gives DANA optionality. For instance, if DANA can integrate with Bank Sinarmas for deposit-funded lending (similar to the OVO-Superbank model), it could transition from a pure-play payments wallet into an embedded finance player.

Competitive Landscape

Looking across these players, I believe there is a clear distinction between some of them. The most useful lens to view them is not by country or product vertical as these can easily change overnight, rather I believe it is best viewed through structural origin.

Where these companies originated and their backers, determines the data they have access to, the distribution they can leverage and ultimately what kind of moat they can build.

When comparing across this list, there are a few insights that I believe we can takeaway.

Read more

However, the stock has since been hit hard from a combination of factors including a Crypto bear market and general drawdown in the AI/Datacenter space over concerns of an overbuild.

It now trades at $18, seeing a 60% drawdown from all-time highs over the past 6 months. It also means my position is now down 25%.

In this piece, I will give an update of my thoughts on Galaxy as an investment and my plans moving forward.

Read more

S&P 500: -5.25%
NASDAQ: -5.08%

March was one of the most consequential months for markets in recent memory. The US-Israeli operation against Iran, launched in the final days of February and escalated throughout March. Stocks opened the month in free fall as futures plunged on the first Monday of March. By month end, the S&P 500 had posted its fifth consecutive weekly loss, the longest losing streak since 2022, and was sitting nearly 9% below its January all-time high. The Nasdaq entered correction territory, falling more than 12% from its high.

The result of the operation was an oil surge while almost everything else sold off. Brent crude topped $108 per barrel and WTI breached $94 by the final week of the month, as concerns over the Strait of Hormuz disrupting global tanker flows sent energy stocks ripping higher. Energy is the only positive S&P 500 sector this year, up over 30% year to date. Meanwhile, tech heavyweights have been hit hard. Nvidia is down 12% YTD, Microsoft down 24% YTD. Google, seen as the AI leader, is down over 13% YTD. Meta had an especially rough stretch, dropping 12% in three days on the back of layoffs across Reality Labs and Facebook, combined with two separate jury verdicts finding Meta legally responsible for harm to minors through its social media platforms.

The macro picture has darkened considerably. The OECD raised its US inflation forecast to 4.2% for 2026, well above the Fed’s own 2.7% projection. The word “stagflation” has returned to mainstream financial commentary, and could be a persistent worry for the rest of the year.

There were moments of reprieve as markets reacted to various moments of Trump’s infamous TACO (Trump always chickens out). Trump suggested in a Truth Social post that the US and Iran had engaged in “productive” talks, which sparked a sharp relief rally mid-month. However, the optimism was short-lived. By the final days of March, oil was climbing again, the Strait of Hormuz remained a flashpoint, and China opened a trade probe against the US in retaliation to tariffs, adding another layer of uncertainty.

In my view, each subsequent TACO has diminishing returns, and the markets are certainly showing it.

March was yet another brutal month for investors, and I will be discussing my plans moving forward in the following article.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +143.63%

S&P 500: +53.33% (Outperformance: +90.30%)

NASDAQ: +68.09% (Outperformance: +75.54%)

Portfolio Performance (YTD)

Portfolio: -18.07%

S&P 500: -7.33% (Underperformance: -10.74%)

NASDAQ: -10.40% (Underperformance: -7.67%)

Portfolio reviews are exclusively for paid subscribers. If you would like to gain access to my portfolio holdings, deep dives, and tons of other articles, do consider subscribing.

I post several articles per month, ranging from extensive 50-60 pager deep dives on the most interesting businesses I can find, to shorter articles on valuations, market musings and specific verticals/geographies.

In the past month, I’ve shared these articles with subscribers:

Free subscribers get access to posts too, of course, with free earnings reviews and occasional articles.

In the next few segments, I discuss my general thoughts on the market, portfolio companies, performance, portfolio allocation and portfolio changes.

General Thoughts on the Market & Performance

Read more

One of the key criticisms I've had of the business was the $6B cash pile that they sat on for over a year. This, while the business continues to generate excess free cash flow each quarter (expecting about $450M in Adj. FCF in FY26).

However, in the past 6 months, they have started deploying this capital aggressively. In late 2025, they went on an AV investment spree, a $60M initial investment into Vay (with potential follow-on of up to $350M), and undisclosed investment dollars into WeRide, May Mobility and Momenta. In Feb this year, they announced the purchase of Stash Financial for $425M, and just last week, an all-cash $600M deal for FP Taiwan.

Even after all of this, Grab still sits on $5.4B in net cash as of December 2025. They're deploying aggressively, but from a position of overwhelming balance sheet strength, not levering up.

We also found out that the FP Taiwan deal was in discussions for 9 months. That means conversations likely started around June/July 2025, roughly the same time the AV investment spree kicked off. The capital allocation engine has clearly been turned on.

I would not rule out a potential deal for GoTo in the near future.

Let’s also forecast the effect of a GoTo deal. Currently, GoTo has a ~31% market share in the food delivery market in Indonesia. Hence, it would give Grab a 77% market share post-deal, dwarfing the 23% held by ShopeeFood. Gojek also has a ~40-50% market share in Indonesia. A deal would therefore give Grab an absolute monopoly in the largest SEA market.

Read more

Revenue: $337.9M v $296.1M est. (+65% YoY, +69% FX-Neutral) 🟢

Adj. EBITDA: $78.4M v $76.9M est. (+38% YoY, +9% QoQ) 🟢

GAAP EPS: $0.18 v $0.18 est. (+87% YoY, +7% QoQ) 🟢

Selected Key Metrics

Again, dLocal saw YoY and QoQ growth across all 6 key metrics. Management also added a new metric that they will be disclosing (operating profit). On that basis, operating profit grew by 48% YoY.

• TPV: $13.1B (+70% YoY) ✅

• Revenue: $338M (+65% YoY) ✅

• Gross Profit: $116M (+38% YoY) ✅

• Adj. EBITDA: $78M (+38% YoY) ✅

• Net Income: $56M (+87% YoY) ✅

• Adj. FCF: $65M (+100% YoY) ✅

Subscribe now

2025 as a whole

dLocal accelerated growth across its key metrics, and have grown operationally too. They now hold 37 licenses across 6 markets, adding 4 in 2025, with a further 16 applications in process.

APMs continue to grow and now represent the majority of EM e-commerce volumes. dLocal also now offers a full suite of stablecoin solutions to merchants, while it is collaborating with all the major agentic payment protocol builders to ensure its merchants are able to process agentic payment mandates.

dLocal has a TPV retention rate of 158% with existing merchants clearly seeing the value in the business. However, it is notable that new merchants accounted for a negligible amount of TPV. It appears that dLocal has largely acquired all the merchants it possibly can.

Still, management has called out that it is still only at <2% of its TAM and ~10% of its current merchants. It also expects this TAM to double in 5 years.

Market and merchant concentration has also fallen slightly year over year, although the numbers remain elevated. Ultimately, this is likely to continue being a concern as large merchants are likely to grow faster than smaller ones.

Table of Contents

1. Introduction

2. Financials

3. Guidance

4. Product Development

5. Management Commentary

6. Concluding Thoughts

1. Introduction

dLocal is a Latin American-based company that facilitates payment flows for businesses ranging from small startups to large conglomerates on a B2B basis, specifically targeting underserved emerging markets.

It thrives in complex payment environments, which are currently a huge tailwind for the business. Effectively, dLocal is a capital-light “toll bridge” business.

dLocal was first pitched to subscribers in July 2025 at $10.98. It has seen volatile stock price performance since but remains a positive returner at this stage.

2. Financials

There are 2 primary metrics to track for dLocal.

The first is total payment volume (TPV). This tracks the total monetary value of all transactions processed through a company’s platform.

The second is take rates. This determines the amount of revenue generated from TPV, which ultimately affects profitability of the business.

Summing it up, we get gross profit, which I perceive as the true top-line of the business. TPV and take rates viewed in isolation are pretty much useless.

Total Payment Volume

• TPV was up 70% YoY and 26% QoQ.

• This marked the 6th successive quarter of acceleration in TPV. At the current scale of $13.1B in TPV, this is not an easy feat. IMO, this is very very impressive and shows the clear inflection since Pedro joined.

• In Q4 2025 alone, TPV on a QoQ basis grew by more than the prior 3 quarters combined.

Gross Take Rates (Revenue/TPV)

• Gross take rates for the quarter were 2.58%, a 13bp decrease QoQ and 7bp decrease YoY.

• Again, this is part of parcel of running a payments business. As TPV increases and customer relationships deepen, factors such as custom pricing negotiations, increased competition, shift in payment mix, and operational efficiencies will affect take rates.

• Most importantly, we have to look at the net effect of rising TPV and decreasing take rates.

Net Take Rates (Gross Profit/TPV)

• Net take rates are arguably more important than Gross take rates as they take into account other costs that dLocal has to pay, such as network fees, partner fees and revenue sharing fees.

• It details exactly how much dLocal keeps after all these fees.

• Net take rates came down yet again this quarter, falling to 0.88%, an all-time low.

• Again, this should not be a surprise. It is clear that dLocal is prioritising TPV growth and that will come with margin compression.

Gross Profit

• dLocal’s gross profit continued to grow, up 38.4% YoY, reflecting the increase in TPV, despite the reduction in net take rates.

• Ultimately, this will determine dLocal’s long-term profitability and sustainability of its business model. Increasing gross profit dollars is the number 1 priority of the company, and the only way to do so is by increasing TPV or increasing take-rates.

• The latter is near impossible, but prioritising TPV will enable higher-margin services in future that could supplement the drop in margins.

3. Guidance

dLocal has guided for 50-60% YoY growth in TPV for 2026, compared to 59.1% in 2025. Pedro and Co. have sandbagged massively in previous quarters and I expect TPV to come in at the higher range of this guidance.

Gross Profit growth is slightly light, at just 25% at the midpoint. This is compared to 36.7% in 2025. Management noted that this is due to dLocal’s scale and long-term merchant relationships that will inevitably lead to more favourable terms for merchants.

Operating Profit, which is a new metric used by dLocal, is expected to grow 30% at the midpoint.

dLocal also expects new countries to contribute more to the growth of its TPV in % terms in 2026, with significant upside in Africa and Asia.

It also foresees more new merchants in 2026 across high-growth verticals such as Crypto, Travel and Gaming. New products are expected to contribute slightly towards TPV growth, in areas such as BNPL, merchant of record, virtual accounts and card-present.

Buyback and Dividend

dLocal also announced a $300M buyback plan supplementing the previous dividend plan that was shared. Last year, dLocal announced that 30% of the previous year’s Adj. FCF will be returned to shareholders.

The confirmed number is $57M which equates to $0.1939 per share (subject to adjustment according to number of shares outstanding as of record date). This equates to about a 1.6% dividend yield. Furthermore, the $300M buyback will equate to over 8% of shares outstanding.

This is a business that is highly cash generative with massive operating leverage to come.

4. Product Development

dLocal has prioritised its stablecoin infrastructure in 2025, and has made collaborations with some key players in the market such as Circle, BVNK, Fireblocks and Felix. Stablecoins are clearly a new innovation that will disrupt traditional payments, but dLocal appears well-positioned to participate at the bleeding edge.

It also announced BNPL Fuse which is now operational in 6 countries and scaling rapidly, with dLocal reporting 88% QoQ growth in 4Q25. Importantly, 93% of BNPL users are new customers, which means little to no cannibalisation of existing payment methods.

dLocal has also prioritised APMs through partnerships with DHL Express Brazil to automate Pix payments, Open English in Colombia, and Yape in Peru as some key highlights.

5. Management Commentary

On New Products

“Buy Now Pay Later fused products are now live across 6 countries with solid merchant adoption.

We've completed the launch of our full-service stablecoin suite, enabling merchants to on- and off-ramp fiat to stablecoins, settle and be settled in stablecoins and collect at checkout in stablecoins.

And we continue to add an ever-growing portfolio of APMs, a SmartAPM platform.”

On Stablecoins and AI Agents

“On stablecoins, we've offered a full suite of stablecoin solutions for merchants.

And on AI agents, a possible new frontier for commerce, we are collaborating with Google on the AP2 open standard for interoperable AI engine payments to ensure local payment methods across emerging markets are part of that infrastructure for the ground up.”

On Future Prospects and Take Rates

“Long term, we guarantee that we continue to be one of the scale leaders across emerging markets.

And we feel fairly confident that if you have the merchant relationship and you're processing for them, we will figure out ways to monetize those relationships and all that TPV.

So focused on TPV growth, focused on gross profit dollar growth and being scale leaders across emerging markets is what's implied in the guidance.”

On Prioritising TPV and Future Profitability

“If we have the TPV, we have the merchant relationships as our product portfolio widens, we have that TPV and those relationships to cross-sell new products and also to figure out different ways that we can help our merchants across the markets where we operate with them. So we continue to see take rate as an output metric.

The metrics we manage to our TPV growth which reflect market share, share of wallet and how our merchants choose.

Remember that TPV for us is revenue for our merchants and then be able to drive gross profit growth operating profit growth and earnings growth as a consequence of that sustained high level of compounding TPV growth.”

On 2026 Guidance

“Following a 2025 investment cycle, which has overhang into early 2026 as salaries and wages spend from '25 hirings gets annualized, we expect operating leverage acceleration to become evident more towards the second half of the year and then flow into the following year.”

6. Conclusion

Overall, I thought this was yet another fantastic quarter. Management has proven its ability to spur TPV growth.

Margins will continue to be an overhang, but I am optimistic that the market will change its mind once it realises that gross profit dollars are the true north star metric for the business.

Currently, dLocal is growing gross profit at 37% while trading at just 5.8x NTM EV/GP and 13.6x NTM P/E.

I believe the stock is very undervalued, and the market continues to underestimate the long-term sustainability of its business model and ability to generate operating leverage.

Several months back, I posted a full valuation model on dLocal, detailing my assumptions and stock price output.

The assumptions that I made for FY2025 have been surpassed:

TPV: $37.08B (est) v $40.8B (actual)

Operating Income: $200M (est) v $220M (actual)

Net Income: $164M (est) v $196M (actual)

At these prices, I foresee lots of upside to come. I’d encourage you to check it out.

To conclude, I have no idea when the market will re-price dLocal, but I do believe patience and a huge margin of safety will be rewarded eventually. dLocal is building a payments behemoth that is spitting out cash hand over fist and is entirely shareholder-friendly.

This is a business that I am happy to back.

Thanks for reading!

- Gab

Paid Subscription Upgrade

If you’d like to support the work I do, consider becoming a paid subscriber. Your support will allow me to spend more time finding asymmetric opportunities in the market, writing and analysing various businesses.

Till the end of March, new paid subscribers will get 25% off the paid plan.

As a reminder, paid subscribers get access to:

• Monthly Portfolio Updates (+98% in 2024, +26% in 2025)

• Earnings Reviews on Portfolio Companies (SE, GRAB, DLO, MELI etc)

• Archive of Deep Dives and Posts (13 Deep Dives and counting)

• Southeast Asian coverage of industries and companies

25% OFF for Limited-Time

Fiscal AI

Many of the charts and numbers I use are courtesy of Fiscal AI, which is in my view, the best platform for financial charting and market intelligence.

I have an affiliate link where you get access to a free 2-week pro trial and 15% off the pro plan. If you are interested, consider signing up. This will greatly help the Substack and help me to continue producing content like this. Thank you!

15% OFF Fiscal AI

Revenue: $377.7M v $342.3M est. (+46.3% YoY, +7.9% QoQ) 🟢

ARR: $1,460M v $1,441.5M est. (+33.6% YoY) 🟢

EPS (Non-GAAP): $0.04 v -$0.11 est. 🟢

Subscribe now

*Note that Rubrik’s fiscal year ends on January 21. In essence:

• FY2025 = Feb 2024 → Jan 2025

• FY2026 = Feb 2025 → Jan 2026

• FY2027 = Feb 2026 → Jan 2027

Selected Key Metrics

• 2,805 customers with $100K or more in subscription ARR, up 25% year-over-year

• GAAP Gross Margin: 82%, 6th consecutive quarter of increase

• NPS remains in the top 1% of SaaS businesses

Table of Contents

1. Introduction

2. Financials

3. Guidance

4. Management Commentary

5. Concluding Thoughts

1. Introduction

Rubrik is an American cloud data management and data security company based in California. It started as a challenger to traditional storage and disaster-recovery vendors, but has since evolved into a cybersecurity-driven data protection platform.

Rubrik helps companies to stay resilient against attacks and recover fast when disaster strikes, protecting enterprise workloads across on-premise, cloud, and SaaS environments.

Rubrik was first pitched to subscribers on 22 September at $80 following the earnings drop of 18% in Q2’26. This is the second earnings report since owning the business. Since the initial pitch, I’ve added to the position multiple times.

It is now my 6th largest position, just outside my 5 core positions in the portfolio. Let’s break down the earnings results.

2. Financials

Subscription ARR

This refers to the annual run rate from subscription contracts. For Rubrik this is the core recurring business coming from Security Cloud subscriptions, SaaS protections, and legacy subscription contracts tied to appliances or software. It excludes one-time services and hardware sales, so it is the cleanest signal of recurring scale.

Rubrik’s subscription ARR for Q4 2026 came in at $1.46B, up 34% YoY. Growth has understandably slowed in recent quarters as per the law of large numbers.

That said, it is good to see that growth has somewhat stabilised. If Rubrik maintains its subscription growth rate above 30%, it would be massively encouraging.

Cloud ARR

Cloud ARR isolates the portion of subscription ARR coming from Rubrik’s cloud native products, principally Security Cloud and other SaaS protections. This is the revenue that validates Rubrik’s move from on-prem backup into cloud first security services.

Cloud ARR this quarter came in at $1.293B and grew faster than total ARR again (48% v 34%). This is a strong sign for Rubrik that customers are buying the cloud offering at scale and that the cloud portfolio is driving a majority of the company’s growth.

Again, this is key as cloud ARR typically has better land-and-expand mechanics and lower cost-to-serve over time. This leads to higher margins moving forward.

Average Subscription Dollar-Based NRR

Rubrik has not disclosed its exact NRR but has maintained >120% NRR for the nth consecutive time. Since disclosing this metric from the S-1, it has never went below 120%. This is a key indicator of strength for the business as it implies existing customers are expanding well beyond what is lost to churn.

Customers with $100K+ in Subscription ARR

Another key indicator is the number of high-value customers where Rubrik is most likely to land strategic, high-value contracts. For Rubrik these are the accounts most likely to be mission critical and to adopt multiple modules.

As of Q4, Rubrik has 2,805 customers with $100k or more in subscription ARR, up from 2,085 last year. This represents 25% YoY growth, which is a slight drop from prior quarters numbers but is to be expected with the law of large numbers.

Subscribe now

3. Guidance

Rubrik’s management are notorious sandbaggers. They’ve guided for a midpoint revenue of $366M in Q1 2027, which implies 31.4% YoY growth. I expect the actual numbers to come in materially higher.

For FY2027, management has guided for subscription ARR of $1.834B and total revenue of $1.602B which would imply 45% and 21.7% growth respectively. While these numbers are sandbagged, it still feels rather low to me at 21.7% (normalising for material rights, ~27-28%), considering FY26 growth came in at 48.4%.

Q4 saw a weighted-average shares outstanding of 200.8M, and it is expected to continue climbing to 204M in Q1’27. For FY27, Rubrik is guiding for weighted average shares outstanding of 232M. That would represent 15.4% growth in shares outstanding.

That represents massive dilution incoming, albeit largely priced in by the market.

*The stock compensation plan first set out by management was planned pre-IPO and with many stock price targets hit in 2024/2025. This accounts for ~15.5M additional shares. If we consider just new SBC alone, the number is closer to 6M shares. At a base of ~200M, that is about 3% dilution, which is rather palatable.

This remains one of the key risks for shareholders.

4. Management Commentary

On Rubrik’s Core Value Proposition

“Rubrik is the system of record of last resort around data and identity when a large bank or hospital faces a ransomware attack. Data infrastructure and recovery capabilities are foundational to enterprise resilience.

Our platform is built through more than a decade of engineering and enterprise feedback. This technology foundation allows organizations to restore business operations quickly when cyber incidents occur.”

On Vibe Coding & LLMs

“You can close the windows and put on the locks, but ultimately, you need a bunker underneath your house to survive the doomsday, so your cloud and AI transformation journey continues uninterrupted. Rubrik Security Cloud is that bunker for enterprises. When ransomware inevitably hits, a LLM or Vibe code will not recover your business, Rubrik will.”

On Rubrik’s Identity Protection Product

“We have been rapidly disrupting the identity protection market. In just a few quarters this business grew to more than 900 customers, making it the fastest-growing product in our company’s history. In Q3, we had reported that we had crossed 400 customers.

Protection for Okta Identity allows Rubrik to span Okta, Active Directory, and Entra ID. Identity infrastructure has become a primary target for cyber attackers.”

On Rubrik’s Competitive Positioning

“Our competitive win rates exceeded 90% in the fourth quarter. Customers are increasingly turning to Rubrik to accelerate AI transformation while strengthening cyber resilience. Organizations can begin with data protection, cloud security, or identity resilience and expand across the platform over time.”

On Rubrik Agent Cloud

“Last quarter, we shared that Rubrik Agent Cloud was in beta. Just a few weeks ago, we made Rubrik Agent Cloud generally available. And we have a number of POCs ongoing across early AI adopters as well as Fortune 500 companies. While we are still in the early innings of multiyear effort to scale our Rubrik Agent cloud suite, we believe that we are building the most consequential security and AI operations company for the AI era.”

Bipul on the 3 Key Takeaways

“First, Rubrik is winning the cyber resilience market across data and identity. Second, we have accelerated our business growth while the competition has stalled. And third, we are defining the enterprise AI market with our unique and differentiated agent control and guardrail solutions.”

On Rubrik’s Competitive Advantage over Legacy Vendors

“My belief is that the traditional cybersecurity companies, or people in those cybersecurity companies will not be suitable for this market because traditional cybersecurity is all about rule-based platform, and they’re not in the real-time control of action. This market is about dynamic control. And you need to bring an AI to control agentic actions. And AI requires model engineers. And most of the cybersecurity company, probably none of the cybersecurity company or these start-ups have any model engineering.

We brought Predibase to solve this problem. We believe we have a unique solution to control agentic action with AI to really drive intent-based understanding of action and stopping it.”

5. Concluding Thoughts

Rubrik delivered a strong Q4, outperforming estimates as expected. Revenue grew 46% YoY and 8% QoQ, a slight deceleration, but nonetheless very strong results.

Whether Rubrik can maintain these high growth rates will depend on the adoption of its new offerings such as Identity Resilience and Agent Cloud. It will take time for these to move the needle, but they do have the potential to be much larger contributors to revenue over the long term.

In the past 3-4 months, Rubrik stock has seen a ~50% drawdown following the impressive earnings beat in Q3. This can be attributed to 2 main reasons: firstly, the fear of software disruption that has wiped out trillions in market cap and secondly, the broader market drawdown in high-growth stocks amidst worries of stagflation and recent Iran-Israel-US conflict.

On the fear of software disruption, I’ve had my say in recent months on where I foresee this going. I believe there is a very real risk of replacement across hundreds to thousands of software businesses who provide non-mission critical software services. This does not include Rubrik.

In fact, Rubrik appears likely to benefit from the proliferation of AI, LLMs, and vibe coding. In an Agentic era, Rubrik’s services will be in greater demand. What gives me confidence, is that Rubrik is a founder-led, young business that is hungry for success. In my view, this is the key determinant in which a winner is to be picked for the coming decade.

There are certainly risks to be had with regard to Rubrik. Firstly, stock-based compensation and the resulting share dilution is a real dampener on shareholder returns. That said, the pre-IPO compensation plans should come to a head by the end of the new fiscal year. While SBC remains 22% of revenue today, it has come down massively since IPO.

The second major risk in my view, is the ability for Rubrik to become a truly differentiated platform business in cybersecurity. Currently, it remains a backup-focused business. While their new products such as Identity Resilience and Agent Cloud are exciting, they still remain niche products and not a needle-mover.

The third major risk is the ability for these new products to support the growth trajectory that will inevitably falter as Rubrik takes a larger stake of the backup market.

At this point, it is clear to see that Rubrik is a market leader in this rather niche space. That means it is a big fish in a small pond. Now as it graduates into a larger pond, it will have to ensure it remains a big fish.

This quarter reinforces my confidence in the business. Bipul and Co. are clearly moving in the right direction. My thesis remains that Rubrik can be a massive player in multiple growing markets within cybersecurity. Despite fears by the market, cybersecurity remains rooted in a multi-decade tailwind and there is no other sector in SaaS that I would rather be invested in.

Thanks for reading!

- Gab

Paid Subscription Upgrade

If you’d like to support the work I do, consider becoming a paid subscriber. Your support will allow me to spend more time finding asymmetric opportunities in the market, writing and analysing various businesses.

Till the end of March, new paid subscribers will get 25% off the paid plan.

As a reminder, paid subscribers get access to:

• Monthly Portfolio Updates (+98% in 2024, +26% in 2025)

• Earnings Reviews on Portfolio Companies (SE, GRAB, DLO, MELI etc)

• Archive of Deep Dives and Posts (13 Deep Dives and counting)

• Southeast Asian coverage of industries and companies

25% OFF (Limited-Time)

Fiscal AI

Many of the charts I use are courtesy of Fiscal AI, which is in my view, the best platform for financial charting and market intelligence.

I have an affiliate link where you get access to a free 2-week pro trial and 15% off the pro plan.

15% off Fiscal AI

Food delivery is the foundation of it all, and the largest by far at the moment. It is a $23B market today which will grow to ~$35B by 2030. On-demand grocery and daily essentials is a largely untapped opportunity at the moment that could grow to ~$18B by 2030. Non-food instant retail will add a further $12B or so.

Together, these three layers represent a $65B quick commerce market by 2030, roughly 3x the size of the food delivery market today.

Grab is the only platform with an existing operational footprint across all three layers, but its current share of the grocery and non-food prize is negligible. To capture the opportunity, it will require a major supply-side buildout.

This report will examine each layer of quick commerce, Grab’s current position, the investments required to make it happen, and what the economics look like if Grab executes to a tee.

Subscribe now

In case you missed the deep dive on the quick commerce landscape in China and Southeast Asia, do check out the free article linked below:

It is a 15,000 word article written in collaboration with where we go through exactly why quick commerce is being competed over fiercely by major players across China and Southeast Asia.

1. The $65 Billion Market

The defining structural insight about quick commerce in Southeast Asia is that it is not a single market with a single competitive dynamic. Instead, it is three distinct demand pools, each at a different stage of maturity, different unit economics and each requiring a different set of merchant and infrastructure capabilities.

Yet, all three can, in principle, be served by the same on-demand rider network.

This is the strategic logic behind quick commerce as a platform play. Achieving sufficient density in food delivery means solving the hardest infrastructure problem. Specifically, a platform will need a highly-liquid, city-level rider network with excellent demand forecasting and merchant relationships. Both groceries/daily essentials and non-food instant retail rides on that existing network at no marginal cost. That is the beauty of this quick commerce opportunity.

For Grab, it is crucial that they treat food delivery as the foundation of this journey, not the destination.

These are the 3 layers and how I define them:

• Layer 1: Food Delivery

  • This refers to prepared meals (food & beverage) delivery, and as of 2025, GMV is projected to be ~$23B according to Momentum Works. Grab owns ~55% of the market.

• Layer 2: Grocery & Daily Essentials

  • This refers to services like GrabMart, PandaMart and RedMart that offers same-day delivery of groceries and daily essentials. GrabMart is ~10% of GrabFood as of Q4 2025, implying a bottom-up estimate of ~$3B in market size for this business. (Noting that RedMart for instance, has no food delivery segment and therefore is incremental to this number)

• Layer 3: Non-Food Instant Retail

  • Non-food instant retail is the smallest layer today and encompasses FMCG products, personal care and beauty, OTC pharmaceuticals etc…

  • This segment currently accounts for over 40% of China’s total quick commerce market today. With how nascent this space is in Southeast Asia, I would give an estimate of ~$2B for this segment, although there is massive room for growth.

2. Layer 1: Food Delivery

Food delivery is the most mature of the three layers and the one where Grab maintains dominance. Grab currently has ~55% market share across Southeast Asia and is the closest equivalent to Meituan in the region, a platform in China that has leveraged food delivery to build a massive logistics and consumer internet business that is now fuelling their quick commerce charge.

For Grab, food delivery is not primarily about market share gain. Rather, it is about continued category expansion with more restaurants, more cuisines, more price tiers, and deepening the habit across its user base.

In terms of the financial profile for this segment, Grab has a take rate of ~20-30% on restaurant GMV with delivery fees partially passed onto consumers. Grab’s model is therefore to monetise the movement of goods and skim a small take from each transaction. Grab also offers sponsored listings, promotional campaigns and other merchant services that provide visibility. In return for this, restaurants on GrabFood get access to millions of active users with intent to order. Grab also handles payment processing, fraud risk, chargebacks and cash float.

Overall, this is not a high-margin business in isolation. But it is the very foundation on which quick commerce will be built and the engine that keeps riders deployed, demand predictable and the platform within the consumer’s daily habit loop.

3. Layer 2: Grocery & Daily Essentials

Online grocery in Southeast Asia is one of the most structurally under-penetrated categories in digital commerce. Across most SEA markets, online grocery accounts for low single digits of total grocery retail spend, compared to 10-15% in mature markets like South Korea, Japan and the UK. In fact, it is ~25% in China.

I believe the gap between current penetration and mature market benchmarks represent the bulk of the layer 2 opportunity.

The reason behind the low penetration rates really boils down to a few key reasons. One, consumers in Southeast Asia who purchase groceries are typically the elderly who cook for their large multi-generational families. They are very skeptical over fresh produce quality and two, have the habit of frequenting their usual wet market spot where they can select items individually. Thirdly, the low basket sizes make delivery economics rather challenging without subsidies or subscription programmes.

However, these are challenges that can be overcome easily. For one, these constraints are likely to erode over time as consumer familiarity with on-demand delivery increases and as platforms invest in fulfilment consistency.

We’ve seen a similar story in China where food delivery became sufficiently habitual, and consumers began applying the same convenience expectation to adjacent categories. For example, a user who orders GrabFood daily may start asking why their shampoo, rice, or cooking oil is taking two days to arrive.

In 2022, the total grocery retail market in Southeast Asia was estimated at about $475B, growing 7% annually. That would imply a $622B market in 2026.

Subscribe now

By 2030, if online grocery penetration reaches just 8% of total grocery retail (which still implies a level lower than most developed Asian markets), that would imply approximately $60B in online grocery GMV.

The majority of this would likely still be fulfilled through next-day delivery or longer. In terms of instant/same-day delivery which would qualify as quick commerce, I believe a realistic number here could be 30%. If so, that would equate to this segment being worth $18B in 2030.

GrabMart is positioned to compete for a meaningful share of this through its layer 1 capabilities.

4. Layer 3: Non-Food Instant Retail

Non-food instant retail is the smallest layer today but the fastest growing. It encompasses FMCG products, personal care and beauty, over-the-counter pharmaceuticals, baby and pet care, consumer electronics accessories, and eventually a broader range of lifestyle and home products.

In China, Meituan’s Flash Shopping non-food service grew to over 10 million daily orders by Q3 2024, with peak daily volumes reaching 25 million by August 2025. Non-food now accounts for over 40% of China’s total quick commerce market by value.

In SEA, the non-food quick commerce market is still very nascent. Consumer behaviour has not yet shifted to treating on-demand non-food delivery as a default channel. I believe there are a few constraints for this. Firstly, on the supply-side, brands and retailers have not yet positioned inventory close enough to consumers to make 30-60 minute non-food delivery economically viable across a broad SKU range.

As a reminder, the SKU range for non-food retail is 100x or more the size of food delivery.

The e-commerce market is estimated to be worth $359B in 2030. I believe it is fair to assume that about 3-5% of that could become eligible for same-day or instant fulfilment in dense urban areas within the region.

Assuming we do reach the higher end of that range by 2030, it would represent about $18B in GMV.

5. Grab’s Position Across All 3 Layers

Read more

EPS: $0.63 v $0.62 (+61.5% YoY) 🟢

Subscribe now

Selected Key Metrics

Total Gross Profit: $3.0B (+36% YoY)
Total Net Income: $410.88M (+72.9% YoY)
Total Adjusted EBITDA: $787.14M (+33.2% YoY)
Cash & Cash Equivalents: $11.1B (up from $10.4B in Q4 2024)

E-Commerce (Shopee)

Gross Orders: 4.0B (+30.5% YoY)
GMV: $36.7B (+28.6% YoY)
Revenue: $5.0B (+35.8% YoY)
Adjusted EBITDA: $202.5M (+33% YoY)

Digital Financial Services (Monee)

Consumer & SME Loans Outstanding: $9.2B (+80.4% YoY)
Revenue: $1.13B (+54.3% YoY)
Adjusted EBITDA: $263.1M (+24.7% YoY)

Digital Entertainment (Garena)

Bookings: $672.4M (+23.8% YoY)
Revenue: $701.0M (+35.1% YoY)
Adjusted EBITDA: $363.8M (+25.6% YoY)
Quarterly Active Users: 633.3M (+2.5% YoY)
Quarterly Paying Users: 58.0M (+15.0% YoY)

Sea Limited beat on the top and bottom-line, though consensus estimates for the bottom-line range widely due to the lack of sell-side coverage on the stock.

As I’ve discussed often, my base case is >20% growth for the decade to come. I don’t see any signs of it slowing so far. In Monee, Sea has a business that is now reaching scale to move the needle. Shopee is not slowing down either with 25% GMV growth guidance and Garena guiding for double-digit growth again.

In this piece, I will break down the full report, earnings call and discuss my personal thoughts on the quarter.

Subscribe now

Table of Contents

1. Shopee (E-Commerce)

2. Monee (Digital Financial Services)

3. Garena (Digital Entertainment)

4. Management Commentary

5. Concluding Thoughts

1. Shopee (E-Commerce)

Shopee’s GMV continued climbing to record highs, with GMV now at an annualised $146.8B run-rate. Sequentially, GMV grew 14% QoQ, compared to 4% QoQ in Q2 2025 and 8% QoQ in Q3 2025, signalling yet more acceleration in the business.

For FY2026, Shopee is guiding for 25% GMV growth. This was above sell-side consensus of 19%, which is obviously great news. Personally, I thought management would guide for 20% GMV growth as they did last year (eventually raising to 25% mid-year and then closing the year at 26.8%), so I am very happy with 25%.

Shopee’s GAAP take rate continued to climb, up to 13.5% this quarter. This brings FY2025 take rate to 13%, up 60bp from FY2024. Again, this was expected, as I predicted in last quarter’s review.

There is still much more room for take rate to go up and I expect this to continue, albeit with some quarter to quarter fluctuation.

Of course, what concerned the market in particular in Q3 2025 was Adj. EBITDA margins for Shopee. This continued into Q4 2025, with 0.6% margins, compared to Q1 highs of 0.9%.

Management guided for “full year adjusted EBITDA no lower than that of 2025 in absolute dollar terms”. This spooked the market as that would mean 0.55% margins for FY2026, compared to sell-side analysts consensus at 0.9%.

Personally, I am of the belief that management is sandbagging guidance here. For instance, when we look back to Q4 2024 where management shared guidance for FY2025, here is what they said:

“We remain confident about our ability to continue delivering profitable growth in 2025 and expect Shopee’s full year 2025 GMV growth to be around 20%, with improving profitability.”

In actuality, top-line beat massively (28.6% v 20%) while Adj. EBITDA for Shopee actually grew 465% (slightly more than improving profitability).

Yet, a lot of this will depend on the rollout of the Shopee VIP program, the equivalent of Amazon Prime. Shopee has been re-investing heavily into their logistics network and anecdotally, I have noticed a very aggressive push for the VIP program. This could be leading to short-term margin pressure, which will continue for a quarter or two, hence the cautious guide.

Shopee Ads

In Q4 2025, ad revenue grew 70% YoY and ad take rate increased by over 80bps. Ad-paying sellers increased by over 20% YoY and their average ad spend increased by over 45% YoY.

Ads will be an increasingly large contributor to the bottom-line and I believe next year should be a key inflection point.

Shopee Logistics (SPX Express)

Management shared that SPX Express now processes on average over 30 million parcels a day. This ranks it 2nd in the region, behind J&T Express which is the key supplier for Shopee’s main rival: TikTok Shop.

Shopee has begun expanding instant delivery into additional use cases, including partnering with local supermarkets and suppliers to deliver fresh groceries in Thailand in as little as 1 hour.

Management shared that buyers using instant and same-day delivery spend around 15% more on average after adoption, this is to be expected as I have shared in the quick commerce deep dive.

“We scaled economical shipping to serve buyers seeking affordability. In Indonesia, orders using economical shipping more than doubled year-over-year in the fourth quarter. With our delivery capability well-scaled, we started to roll out fulfilment service in various markets across 2025. We are seeing encouraging adoption trends from both buyers and sellers, with double-digit order penetration in some markets. In 2026, we plan to expand fulfilment further across all our markets and aim to double our fulfilment order penetration by the end of the year.”

-Forrest Li, CEO of Sea Limited

Shopee VIP

In Q3 2025, management revealed that there were over 3.5 million Shopee VIP members, growing 75% QoQ. In Q4 2025, this went up to over 7 million members, implying 100% QoQ growth.

It is clear that management is prioritising Shopee VIP as a means of access to stable revenue/earnings and a sticky user base. I am fully in support of this, though it will likely lead to continued weakness in margins.

Management shared that across every market it has launched, the program has consistently produced double-digit spending uplift by members after joining and in Indonesia, VIP members have been spending 30-40% more than before joining. In fact, in some markets, VIP members already contribute to more than 15% of total GMV in Q4. This is massive when you consider it is just 1.75% of users.

There is still much room for growth re: the Shopee VIP program. Management shared that in 2025, Shopee served around 400 million active buyers. This implies that Shopee VIP is only at a 1.75% penetration rate. For reference, here are the take-up rates of other popular e-commerce subscription services:

• Amazon Prime: 65%

• Coupang Rocket WOW: 65-68%

• Walmart+: 43%

• Alibaba 88VIP: 5%

• JD Plus: 4-5%

There is huge variability here, with the likes of Amazon Prime, Coupang Rocket WOW offering many more services than just free shipping. However, it is clear that Shopee has huge runway to go.

Interesting to note too, is the synergy between Monee and Shopee in enabling a successful subscription business here. Chris Feng (President of Sea) noted during the call that historically, one of the core challenges for similar programs in the region, is the payment success rate month to month due to the lack of credit card availability. However, due to Monee, there is a smooth payment process for the VIP program, which has led to increase in subscription retention rate from 40% to 70% in Indonesia for instance in the past few quarters.

YouTube & Meta Partnership

In Q4 2025, orders driven by YouTube content more than tripled YoY. Since the launch in October, the collaboration with Meta has seen huge success, with more than 3 million affiliates linking their Shopee and Facebook accounts.

2. Monee (Digital Financial Services)

Monee continued with yet another record quarter of loans principal outstanding. Its loan book grew ~80% YoY to $9.2B while the NPL ratio held steady at 1.1%.

Management shared that they added 5.8 million unique first-time borrowers in the quarter, up from 5 million adds in the previous quarter. Monee now has 37 million active credit users, up more than 40% YoY. Average loan outstanding per user was ~$240, up 27% YoY.

GAAP revenue for the quarter came in above $1B for the first-time. Adjusted EBITDA continued climbing, albeit at a low margin, owing to interest rate declines in SEA markets.

Importantly, Sea is continuing to grow its off-Shopee SPayLater loan book, citing over 300% YoY growth, accounting for over 15% of the total SPayLater portfolio. For FY2025, Adj. EBITDA is now over $1B. Quite a feat.

There is a concerning aspect to Monee though. Off-book loans have barely grown for a year now. While on-book loans have grown from $5.1B to $8.2B YoY, off-book loans have only grown from $0.9B to $1B. (Off-book loans principal outstanding mainly refers to channeling arrangements, which is lending by other financial institutions on Monee’s platform)

Building off-book loans takes time of course, as 3rd party financial institutions will need a few things before they’ll commit capital again. For instance, a long enough default history across economic cycles, regulatory approvals in each individual market, and typically a minimum track record on the specific product.

Right now, Sea is essentially funding most of its loan growth off its own $11B cash pile. That works while the cash is abundant, but it's capital-intensive and creates a ceiling on how fast Monee can scale without either diluting shareholders or constraining other investments.

Long-term however, off-book loans will be key. It will enable Monee to keep growing without cannibalising the rest of Sea’s strategic optionality. Distributing credit risk to 3rd party institutions also acts as a structural hedge. The market will also likely re-rate Monee meaningfully the moment off-book becomes a material portion of the funding mix.

3. Garena (Digital Entertainment)

Garena had probably the weakest results of the 3 segments. Q3 2025 was certainly an anomaly in terms of bookings numbers. Bookings for that quarter came in at $841M, which was a massive increase over Q2. Q4 was more moderated, with a slight reversion to the mean as bookings came in at $672M. GAAP revenue hit a multi-year high of $701M.

Garena remains a very lucrative cash cow for the business.

QAUs fell to 633.3M which was the lowest for the year, albeit with a modest 2.4% increase over last year. QPU ratio fell to 9.2%, also the lowest for the year, but a full point above last year’s Q4 numbers.

4. Management Commentary

This section focuses mainly on questions during the earnings call, with key highlights of what I believe are important points to takeaway. I may adjust a few words here and there for clarity purposes, (English is not the first language for the majority of Sea’s management team) but I will not change the meaning of these quotes.

On Shopee’s GMV Growth:

“If you look back, at the start of 2025 we guided for around 20% top-line growth. As the year progressed, we realised we could grow the business faster and ended up delivering well above that. In Q4, year-on-year growth was much higher than 20%.

Over the year, we identified areas where we could further drive market growth and learned which levers we can pull to accelerate it.”

On Shopee’s EBITDA margins:

“I believe we can continue expanding profitability and margins, and that this trend will persist over the years. We have previously guided toward a 2–3% margin for the e-commerce business over time, and that conviction remains. We will continue to demonstrate this to the market.

Over the medium to long term, we see a 2–3% EBITDA margin as achievable, based on what we have observed so far.”

On the Competitive Landscape:

“Regarding the competitive landscape, what we have observed is a relatively stable environment across most markets. We have not seen anything materially different from what we observed in previous quarters.”

On Shopee Brazil:

“On AOV, we believe it will grow over time. The gap versus peers may remain, but we expect it to narrow gradually.

For SPayLater penetration in Brazil, it is still at a very early stage. While we have seen strong growth, we entered Brazil much later than other markets. Current penetration levels are comparable to the early stages we observed in our initial markets. We expect penetration to continue increasing in 2026, following a trajectory similar to what we saw in our Asian markets.”

On Garena:

“Regarding Garena’s outlook, we continue to expect double-digit growth in 2026.

On the collaboration pipeline, we are very encouraged by the strong performance of our partnership with IPs such as Naruto. We plan to extend this collaboration this year, with delivery currently targeted around Q3. We are also actively exploring additional IP partnerships.

In addition, this is a major football year with the FIFA World Cup. We see significant overlap between the global football community and our global gamer base, and we plan to launch a series of football-related promotions during the tournament period.”

5. Concluding Thoughts

This was far from a perfect quarter from Sea. As I write, the stock is down 23%.

My belief is that the guidance for Shopee Adj. EBITDA being flat for the year is the main culprit. That would imply Shopee Adj. EBITDA margin falling from 0.7% in FY2025 to 0.55% in FY2026. The market clearly doesn’t like it.

But I think this misreads what management is actually doing. At 1.75% penetration and already contributing over 15% of GMV in some markets, Shopee VIP is the priority. Investing in the infrastructure and incentives to make that program work costs money in the short term. Sacrificing profitability in the short-term is a clear choice to build the sticky, recurring revenue base that makes the long-term margin target more durable.

The harder question is whether this is an offensive investment or a defensive one. Shopee is aggressively building out logistics and VIP at the same time TikTok Shop is scaling across the region. If the answer is "both," that's not necessarily a bad thing as the best moats are often built during periods of competitive pressure. It is a fair debate, and one worth monitoring over the next few quarters.

One thing that I guess some investors may be spooked about was the optics of the margin trajectory. Going from Adj. EBITDA margins of 0.2% in FY2024 to 0.7% in FY2025 was the narrative that brought many investors back to the stock. Guiding that back down to 0.55% disrupts that story, even if the underlying business is in better shape than ever. Management arguably could have communicated this better.

Personally, I’ve long said that I intend to hold Sea Limited for a decade. The path to 2-3% EBITDA margins, which I actually think is conservative, was never going to be linear. There will be quarters like this one, where the market punishes short-term optics while the business continues compounding in the background.

The top-line growth in Shopee of 25% is very pleasing to hear (ahead of 19% consensus), and shows a clear push toward growth over profitability for the year ahead. As management reiterated during the call, they have levers to pull on either side, and they can ultimately decide which to prioritise. The fact they are prioritising growth despite knowing the market will not like it in the short-term, reinforces my conviction that this is a long-term minded management team.

The main concern over the next year is that the stock lacks a clean catalyst to move higher while the margin overhang persists. In short, I believe this is a fantastic buying opportunity for the long-term investor.

Paid Subscription Upgrade

If you’d like to support the work I do, consider becoming a paid subscriber. Your support will allow me to spend more time finding asymmetric opportunities in the market, writing and analysing various businesses.

As a reminder, paid subscribers get access to:

• Monthly Portfolio Updates (+98% in 2024, +26% in 2025)

• Earnings Reviews on Portfolio Companies (SE, GRAB, DLO, MELI etc)

• Archive of Deep Dives and Posts (13 Deep Dives and counting)

• Southeast Asian coverage of industries and companies

Subscribe now