Portfolio: -11.39%
S&P 500: -0.93%
NASDAQ: -5.28%

October was a tough month for the majority of tech stocks. The S&P 500 and Nasdaq both finished lower for the month but the Nasdaq was hit particularly hard, meaning it is now underperforming the S&P 500 for the year.

In fact, the tech sector is now at the lowest level since the bear-market low 4 years ago. The forward P/E is now just 0.1 pts above the S&P 500.

One of the primary reasons was the renewed scrutiny of AI spending. For me, the question is how much of the value created by AI will ultimately reach shareholders, and how much capital businesses must invest to earn it.

Disappointing results from Amazon and Intel added to concerns about AI spending and contributed to a sharp technology sell-off on the final trading day.

In September, we discussed how a relatively small group of technology companies had supported the indexes while much of the market declined. October was a reminder that those leaders can come under pressure too.

Portfolio

October was another difficult month for the portfolio. Following September’s 9.2% decline, we saw a further drawdown of approximately 11.9%, bringing the portfolio’s YTD return to -12.95%.

The portfolio has materially underperformed both indexes this year, and October has widened that gap. We will discuss more about the portfolio and individual businesses below.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +158.73%

S&P 500: +83.22% (Outperformance: +75.51%)

NASDAQ: +106.27% (Outperformance: +52.46%)

Portfolio Performance (YTD)

Portfolio: -12.95%

S&P 500: +10.73% (Underperformance: -23.68%)

NASDAQ: +9.96% (Underperformance: -22.91%)

Personal Thoughts on the Market & Performance

In the September review, I discussed my concerns around oil prices, inflation and treasury yields. Most of those concerns carried into October and no doubt affected my portfolio.

However, I do not think it is useful to attribute all of the portfolio’s underperformance to macro.

When the portfolio is down almost 13% while the S&P 500 is up almost 11%, there are also questions to ask about stock selection, valuations and position sizing. However, I do not believe that a difficult period invalidates every thesis.

Much like how I described my ridiculous outperformance in October last year to be too early to tell, I do not believe this underperformance is indicative of how the businesses within the portfolio are doing.

In my mind, the key questions to ask are:

• Has the competitive position weakened?

• Are future earnings likely to be lower than I previously expected?

• Will the business require more capital?

• Or has the price investors are willing to pay simply fallen?

The answer will differ by company. I do not think “the market is wrong” is a sufficient explanation for every stock that declines.

Much of my research this month has focused on exactly these questions, particularly around Sea, AppLovin and AI infrastructure.

My Positioning

Read more

I’ve already discussed my thoughts on the Monee IPO, including why I believe there are legitimate strategic and capital-related reasons for pursuing a public listing.

In this article, I want to focus on the second announcement, Sea Cloud.

As many have pointed out, the reason Sea fell 9.1% on the day of announcement was likely related to this announcement rather than that of the potential Monee IPO.

At first glance, it is understandable why investors might be concerned. After all, Sea is an e-commerce, gaming and fintech conglomerate. Why is it suddenly entering the cloud computing business, an industry dominated by some of the world’s most capital-intensive technology companies?

The immediate comparison is naturally with Amazon Web Services (AWS), Microsoft Azure and Alibaba Cloud.

These businesses have spent tens of billions of dollars building their infrastructure, and the idea of Sea entering the same market, after a period where it has embarked on multiple capital intensive ventures may be worrying to investors.

I think there is more nuance to the story.

The announcement is not quite as surprising as it initially sounds, particularly when we examine what Sea has already built internally over the past decade.

More importantly, I think investors need to distinguish between Sea attempting to become the next AWS and Sea commercialising infrastructure and capabilities it already owns.

These are two very different propositions with vastly different implications for capital requirements, profitability and ultimately shareholder value. Since we have very little information available at the moment, the best we can do is to speculate through scenario analysis.

Subscribe now

Sea already has a cloud business that it has operated for over a decade.

One of the more interesting developments preceding the Sea Cloud announcement was a presentation given by Sea at NVIDIA AI Day Singapore in September 2026.

During the presentation, Sea described its transformation from a digital-native company into what it calls a full-stack AI company.

The most revealing part was the infrastructure disclosure that it shared.

Sea currently operates:

• Over 150,000 servers and 25,000 GPUs.

• More than 40 data centres worldwide with >100 MW of combined critical IT load.

• Connections to more than 120 tier-one internet service providers.

• A network carrying more than 1.1 exabytes of monthly traffic.

  • This equates to watching 700 million hours of high-definition video.

These are substantial figures for a company that operates outside of the cloud and compute space, and is primarily associated with e-commerce, gaming and financial services.

Importantly, Sea’s infrastructure team already manages the entire lifecycle of its infrastructure, from design and procurement to deployment, testing, operations and maintenance.

The company relies on colocation partners for data-centre facilities, meaning it does not necessarily have to own the underlying buildings or land. However, it manages much of the computing infrastructure operating within them.

This distinction matters because it demonstrates that Sea has already developed expertise in operating computing infrastructure at scale.

Sea has also been purchasing NVIDIA accelerators since 2018, was Southeast Asia’s first NVIDIA DGX A100 customer and recently announced that it would become the first enterprise in the region to adopt NVIDIA’s Vera Rubin platform.

In other words, Sea’s investment in computing infrastructure is hardly a recent development.

It has been building these capabilities for more than a decade, initially to support Garena, Shopee and Monee, and increasingly to support AI-related workloads.

The question is whether these existing capabilities can now be monetised externally.

Why would Sea want to enter the cloud computing market?

Read more

Pinduoduo has been aggressively marketing and advertising in Southeast Asia in recent months, and if you are based in Singapore, Malaysia or Thailand, you would probably have seen their ads on Instagram, Facebook etc…

For those unfamiliar, Pinduoduo is one of China’s largest e-commerce platforms, owned by PDD Holdings, which also operates Temu.

Both Pinduoduo and Temu are being rolled out across Southeast Asia, in 5 of the 6 major markets, with the exception of Indonesia, which continues to block cross-border e-commerce.

The natural question is: why is PDD Holdings allowing both their subsidiaries to compete with each other?

Momentum Works wrote an article recently discussing this:

Momentum Works Intelligence

Why is PDD letting Temu and Pinduoduo compete in Southeast Asia?

PDD Holdings is now rolling out Temu and Pinduoduo across five of Southeast Asia’s six major ecommerce markets (except cross-border-unwelcome Indonesia…

Read more

12 days ago · 4 likes · 1 comment

In short, it is because of PDD’s culture of internal competition.

The next question is, what is the difference between Pinduoduo and Temu?

Realistically, there is very little between them, as they both operate cross-border e-commerce services. The difference lies in their approach.

Pinduoduo essentially gives consumers access to China's domestic marketplace, offering extremely low prices but generally longer delivery times and a less localised shopping experience. Temu, on the other hand, was designed specifically for international markets, offering a more localised platform, easier payment options and generally faster delivery, albeit often at slightly higher prices.

Their competitive proposition is very straightforward. They sell extremely cheap products, often sourced directly from Chinese manufacturers, supported by aggressive pricing and promotional strategies.

It is understandable that Shopee investors may be concerned as Pinduoduo has built an enormous business in China, while Temu has demonstrated that the group is capable of rapidly attracting consumers in overseas markets.

However, I think investors are making a mistake by extrapolating Pinduoduo’s success in China and Temu’s in North America and Europe directly to Southeast Asia.

In this piece, I will argue the case for why I’m not particularly worried about Pinduoduo entering Southeast Asia:

Read more

1. Plans for an IPO of its digital financial services business, Monee

2. The launch of Sea Cloud, an AI-first enterprise platform

Both announcements came through an internal letter from CEO Forrest Li, commemorating the 9th anniversary of Sea’s IPO in 2017.

The market’s initial reaction was positive, with the stock up 2.5% in pre-market, before closing down 9.1% at the end of the market close.

At first glance, the two developments might seem unrelated. One involves potentially separating an existing business, while the other involves entering an entirely new market. However, I think both announcements are worth considering within the broader context of Sea’s capital allocation strategy.

Sea Limited has spent years building three increasingly profitable businesses, and management now appears to be considering how to fund the next phase of growth while exploring a fourth business opportunity.

In this piece, I will be discussing Monee’s IPO specifically, while leaving the Sea Cloud launch for another article.

Why IPO Monee in the first place?

In the letter, Forrest suggested that Monee is approaching the stage where it can operate as a more independent business, and that a public listing could give it greater resources to accelerate growth.

I think there are three plausible motivations behind the proposed IPO: unlocking shareholder value, accessing capital and establishing Monee as a standalone financial institution.

Subscribe now

Is this really a spin-off?

Before that, I wanted to discuss a common misconception that I’ve seen going around.

Many investors seem to be assuming that Sea intends to spin off Monee in a traditional sense, distributing shares to existing shareholders and separating the two businesses.

I don’t think this is the most likely outcome.

Instead, I suspect Sea will pursue a partial IPO, selling a minority stake in Monee while retaining majority ownership and control.

There are several possible structures:

My base case is a partial IPO involving newly issued shares.

This seems most consistent with Forrest’s emphasis on providing Monee with additional resources to expand.

For example, suppose Monee were valued at $20B before the IPO and issued $4B in new equity as part of the IPO. Its post-money valuation would rise to $24B, with new investors owning ~16.7% of the business while Sea retains ~83.3% of it.

Monee would receive the $4B in cash to fund further expansion, and Sea would continue to consolidate the business as long as it retains control.

This is quite different from a traditional spin-off, where the parent might distribute shares to existing investors without raising new capital.

Importantly, a partial IPO does not necessarily destroy the synergies between Shopee and Monee. Both businesses can continue operating within the same ecosystem, subject to regulatory requirements and appropriate commercial arrangements.

I would be surprised if management were willing to relinquish control of Monee entirely given the strategic importance of embedded financial services to Shopee.

In the next section, we’ll discuss the 3 plausible explanations for a Monee IPO.

1. Unlocking Shareholder Value

Monee has rapidly grown into a substantial business over the past several years. In Q2 2026, it generated $1.4B in revenue, representing 58.9% YoY growth, while its outstanding consumer and SME loan portfolio reached $11.1B, up 62.5% YoY.

In fact, revenue has grown at a 161% CAGR since 2019.

On the profitability end, it has had a remarkable turnaround over the past few years, now generating north of $1B in Adj. EBITDA on an annualised basis.

In other words, Monee is no longer a small financial services operation built to service Shopee. It is a meaningful contributor to Sea’s overall revenue and profitability profile, at 18% of group revenues and 31% of group Adj. EBITDA.

The key question, is whether the market is appropriately valuing this business within Sea.

DBS recently estimated that Monee could be worth between $18B to $22B, based on 12-15 times projected 2027 adjusted EBITDA of $1.5 billion. The bank argued that Sea’s prevailing share price implied a considerably lower valuation for the division. As I write this, Sea is trading at $34B in market cap with nearly $7B in net cash.

On a simplified sum-of-the-parts basis, assigning Monee a valuation of $18B-$22B would leave a surprisingly modest residual value for Shopee and Garena. Of course, this isn't a clean enterprise value calculation, particularly given Monee's financial balance sheet and the capital required to support its lending operations.

Nevertheless, I think there is a reasonable argument that Monee could receive a more appropriate standalone valuation as a listed company, particularly if investors can analyse its financial performance separately from Shopee and Garena.

2. Accessing Capital

The other plausible explanation for Monee’s IPO is its increasingly substantial capital requirements.

Historically, Monee operated primarily as a payments business, benefiting from Shopee’s enormous transaction volume and existing customer base. However, the business has increasingly shifted towards lending, which has fundamentally different economics.

Read more

In the past week, CoreWeave bonds sold off, with yields touching 15% amidst higher than expected inflation and recent news around the 2 massive AI giants: OpenAI and Anthropic.

On 8th October, the FT reported annualised revenue approaching $50 billion, versus the approximately $70 billion previously communicated. It turned out to be a gap in how revenue was calculated, but it was enough to show the fears among investors.

On 14th October, Anthropic reported Q3 revenue of $16.8 billion, up 45% QoQ. This remained exceptional growth, but represented a slowdown from the previous quarter and disappointed some investors’ expectations.

With these 2 labs being the harbinger for all other AI companies, the connection was great to CoreWeave, with its future revenue entirely dependent on the capability of these businesses to fulfil their contracted capacity.

For a business that relies heavily on external financing to build infrastructure, the 15% bond yields are hugely important. Higher funding costs can weaken project returns and make future growth more expensive.

Yet, the operating picture remains encouraging. CoreWeave continues to hold a substantial contracted backlog, while SemiAnalysis’s latest assessment supports its technical credentials. The tension is between strong demand for its infrastructure and growing concern about how that infrastructure is financed.

Below, I discuss the pros/cons of owning CoreWeave at these prices amidst recent happenings:

Read more

Portfolio: -9.20%
S&P 500: -0.45%
NASDAQ: +1.93%

As we can see from the chart above, September was a truly mixed month. Many technology names rallied, led by META, AMD, MU and INTC, while pretty much the rest of the market fell.

We discussed in the August review that September has historically been the weakest month for the markets over the past 20 years, with -0.6% monthly returns on average. It was only higher 45% of the time. Again, it was a negative month for the S&P 500 index, although the NASDAQ did close higher thanks to the technology rally we discussed earlier.

10/11 sectors were red in September, but Technology with the highest weighting, heading up 5.1%, meant the overall index was only down slightly at -0.4%.

Additionally, in September, only large cap growth was up for the month, no doubt led by the technology giants we mentioned earlier. All other categories were red on the month.

My decision to raise cash in late-August was the right one in hindsight. I had raised my cash position from 4.6% to 10.7%. As subscribers will know, I don’t make moves on a daily basis, and some times not even for weeks. In the past month, I have deployed the cash that was raised, along with some re-balancing of the portfolio that I discussed in the article below:

We will discuss more about the markets looking forward in the later segments.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +193.64%

S&P 500: +84.94% (Outperformance: +108.7%)

NASDAQ: +117.78% (Outperformance: +75.86%)

Portfolio Performance (YTD)

Portfolio: -1.22%

S&P 500: +11.77% (Underperformance: -12.99%)

NASDAQ: +16.09% (Underperformance: -17.31%)

Personal Thoughts on the Market & Performance

Read more

TikTok Shop entered Southeast Asia in 2021 and rapidly expanded across the region in 2022, with market share quadrupling in 2023 and 2024 from 1% → 4% → 16%. Today, with the acquisition of Tokopedia, it has at least 30% market share in Southeast Asia.

TikTok Shop is a serious competitive threat to Shopee. That much should be clear by now. As it stands, e-commerce in Southeast Asia is headed towards a duopoly.

Shopee and TikTok Shop collectively control over 90% of Indonesia's e-commerce market, with similar dominance patterns emerging across Vietnam (97% combined) and the Philippines (84% combined), signalling the end of the multi-player competitive era in Southeast Asia.

In this piece, we will discuss the competitive landscape today, their differing approaches, fulfilment, GenAI and the future outlook for these 2 giants.

Read more

Consumer lending is becoming a much larger priority for the group. In FY2028, Grab expects GFin to make up ~30% of total Adj. EBITDA. Groceries and retail are receiving more investment, and management increasingly sees financial services as the end-game.

Atome makes a lot of sense from a strategic fit perspective. In terms of pricing, I think we simply do not have enough information to make a determination on whether this acquisition is cheap or not.

I have some contrarian thoughts on what I think about Grab now given the amount of information that was shared in the Atome presentation.

After going through the investor update and management’s Q&A, here are my 8 main takeaways:

Subscribe now

1. Grab has decided Financial Services is the end-game

Grab specifically described financial services as the next unlock. I personally believe this acquisition, and Stash, proves Grab is not simply seeing it as the next unlock, but instead the primary bet.

Much of this is down to necessity. While Grab dominates mobility and to a smaller extent deliveries in the region, both are commodity-type businesses. They compete primarily on price and convenience, with low switching cost. In Singapore, most consumers will have at least 3 or even 4 ride-hailing apps downloaded for instance. The experience may differ slightly, but if the aim is to get from point A to B, the primary consideration is about price.

Financial Services gives Grab another way to monetise its ecosystem. Using existing customer relationships and proprietary transaction data, they are able to acquire and underwrite borrowers more efficiently. If those advantages translate into lower acquisition costs and better credit outcomes, it would then support durable lending profits that competitors cannot replicate simply by matching the product or interest rate.

One more thing of note, is that within mobility and deliveries, we are seeing greater scrutiny by governments in the region. (Most recently, Indonesia with its commission caps)

Southeast Asia remains one of the most underbanked regions in the world and this is a massive opportunity for Grab.

2. Grab has identified Consumer Lending as the gap and they hope Atome is the solution

Grab shared the slide above that showed the weakness in consumer cash and consumer BNPL. Management also shared that only 1% of its 138M annual transacting users currently borrowed from Grab.

That is an astonishingly low number for a business of Grab’s size. For context, Sea’s Monee has over 40M active credit users, which if we were to compare to their annual transacting users on Shopee, equates to about 10%.

In other words, management's claims that they can 10x is certainly not hyperbole, it is very possible.

That said, it speaks to how much Grab has struggled to convert its delivery and mobility users into active credit borrowers.

In my view, there is a fundamental mismatch in the ticket sizes and type of transaction that has resulted in this situation. E-commerce has proven to be a fantastic customer acquisition tool for financial services, but not so much yet for Grab.

Atome should bridge that gap to an extent for Grab, but Grab has to also do much better at conversion.

3. Grab is deliberately investing more in affordability, groceries and retail

As part of the Atome acquisition, Grab increased its 2028 group Adj. EBITDA target from $1.5B to $1.7B. From the face of it, it would be easy to interpret the $200M increase as the profit Grab is buying through Atome.

However, the actual bridge is far different:

Firstly, the additions of FP Taiwan and Atome contribute $360M before the planned $160M investment in affordability, groceries and retail.

Secondly, Atome’s actual profit contribution was not shared as it was bundled with Superbank, which if you recall, was consolidated into Grab in May this year.

On the $160M investment, CFO Peter Oey discussed during the call that they intend to deploy this capital into deepening affordability across the platform, and specifically highlighted groceries (which is growing 2x faster than food), as a focus area.

The reason for the sudden change is the Atome acquisition. Peter mentioned that it will give them access to a new user base, enabling cross-sell opportunities, not to mention, the 30,000 merchants that Atome works with across travel, fashion, e-commerce etc.

Personally, I believe this will be a huge boost for their advertising segment.

The large majority of current advertisers are restaurants. Atome expands the advertiser base beyond restaurants, potentially attracting larger retail marketing budgets.

4. The merchant network on Atome could be as strategically valuable as the $1B loan book

Atome has over 30,000 unique brands across categories including travel, electronics, fashion, beauty and home products.

These merchants sit at different points on the spending spectrum from Grab’s core F&B network. Some sell higher-value items, some have less frequent purchases, and some benefit more directly from instalment financing.

Management shared that they see opportunities to bring more merchant relationships into Grab’s marketplace and offer additional services.

Personally, I see several potential avenues:

• Atome merchants becoming customers for Grab’s advertising and payment services.

• Suitable retailers joining Grab’s commerce offerings.

• Atome distributing credit products through Grab’s existing merchant network.

• More purchasing activity giving the group additional information about customer needs and repayment behaviour.

These are opportunities for Grab, but not necessarily automatic outcomes. There is lots to execute too.

5. Grab has been extremely unsuccessful in consumer lending

One startling fact from the presentation was Grab sharing that only 1% of its annual transacting users were borrowing from them. For comparison, Sea Limited’s Monee has 40M active borrowers, while Shopee has 400M active annual buyers. That equates to a 10% rate.

Two things we can learn from this:

1. Grab’s claim of >10x potential growth is not hyperbole, it is very possible

2. Grab has been incredibly poor at converting its users to credit borrowers

On the 2nd point, Grab’s BNPL product has been around since 2019, so it cannot be blamed on a lack of time to execute. My friend wrote about this in a post on X. I largely agree with his views, do check it out.

I think the fundamental reason behind this is that mobility/deliveries with its small ticket sizes are a mismatch for BNPL.

6. This feels like a desperate move from Grab, and not from a position of strength

I know this might be a controversial take, but it is how I feel.

Management sharing that only ~1% of its 138M annual transacting users borrowing from Grab after years of investment in broader consumer lending raises an uncomfortable question: If Grab’s ecosystem advantages are so powerful, why has converting its users into borrowers been so difficult?

The investment case has long been that Grab already has the customers, distribution and transaction data to build a substantial lending business. Yet broader consumer lending remains underdeveloped, and Grab is now paying heavily to acquire the capabilities needed to accelerate it.

I see Atome, at least partly, as an expensive response to disappointing internal progress. Once the deal is completed, it will most likely come in the $2.5-$3B range for Grab, which is a substantial part of its cash pile. This means Grab is essentially betting half of its cash on this acquisition.

Management argues that building Atome’s capabilities internally would take years. Yet, Atome was started in the same year Grab launched its BNPL service. This is therefore an indictment on management’s failure to execute on its consumer lending segment in general.

On Atome’s side, my suspicion is that joining a larger platform was always an attractive potential outcome. A standalone lender has to secure funding, acquire customers and maintain engagement, while a broader ecosystem can support those activities through existing relationships. In that way, joining Grab was always going to be the ideal outcome.

Yet, the disclosed terms do not make it obvious that Grab secured a bargain from a seller with limited alternatives. Grab is paying a substantial initial amount, with further consideration linked to Atome’s subsequent performance. The remaining 40% appears to be a deal protecting the sellers, with a floor of $2B and ceiling of $4.5B.

I believe this is an acquisition driven partly by impatience with the pace of Grab’s own progress. It could still work, and the strategic fit is evident. However, the burden is on management to prove that this is the right decision.

Conclusion

Overall, I see the strategic logic behind Atome. It gives Grab the consumer lending capabilities, merchant relationships and higher-ticket transaction data that it has struggled to build internally. If management executes well, this could become a major growth driver across lending, advertising and commerce.

However, this is also an expensive admission that Grab’s existing ecosystem has not translated into consumer lending success. It is difficult to ascertain with the information on hand whether the purchase price is cheap or expensive, but Grab is committing a substantial portion of its cash to make this work. It is a massive bet.

I remain constructive on $GRAB, and it is still one of my higher-conviction businesses. However, the facts have changed, and my view has to change with them. The burden is now on management to prove that Atome can generate returns commensurate with the price paid, rather than simply buying the growth that Grab was unable to build itself.

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 (20 Deep Dives and counting)

You can get 30% off your first month with the link below.

30% OFF FOREVER

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.

The reason behind this, is that the most you can lose on a stock is 100%, while the upside is theoretically unlimited. A stock that compounds at 20% a year for 20 years becomes almost 40 times larger. Selling a winner too early, could therefore equate to losing money on multiple stocks.

Many great investors talk about this, Peter Lynch, who averaged 29.2% annual returns from 1977 to 1990 for example:

What Constitutes the Long Term?

My counter-argument is that this assumes you are an exceptional stock picker, with the unique capability to pick stocks that outperform in the long term.

Now what constitutes the long term?

Everyone’s answer will differ, but I think a useful data point is to look at this chart:

Over a 10 year period, the S&P500 was higher 93% of the time.

Over a 20 year period, the S&P500 was higher 100% of the time.

To be extremely safe, let’s pick 20 years as the long term benchmark. Holding stocks for 20 years, would therefore mean stock pickers shouldn’t lose money, right?

Not quite.

The reason is pretty simple. The index constitutes 500 stocks, that are adjusted on a quarterly basis. Therefore, these 500 stocks today are not the same 500 stocks from 1, 3, or 5 years ago. When you buy the S&P 500, you are buying a portfolio that constantly evolves. Companies shrink, disappear, get acquired or are removed from the index. New winners enter. Existing winners become larger weights.

The index does the selection for you over time, while a stock picker doesn’t quite have that luxury.

The Market’s Odds are NOT our Odds

It is perfectly reasonable to look at the chart above and conclude that time is the friend of the investor. However, the truth is that time is not necessarily the friend of every individual stock.

Hendrik Bessembinder's research on individual companies makes this distinction very clear.

Earlier work by Bessembinder found that just 4% of listed US companies accounted for the entire net wealth creation of the US stock market since 1926, while the remaining 96% collectively did no better than one-month Treasury bills.

Globally, the evidence is similar. Research covering more than 64,000 stocks found that just 2.4% of companies accounted for all $75.7 trillion of net global stock-market wealth creation between 1990 and 2020.

The advice to "let your winners run" exists precisely because stock-market returns are so positively skewed. Missing one Amazon, Microsoft or Nvidia can matter far more than avoiding dozens of mediocre investments.

However, this creates another problem:

How do you know if you own an Amazon, or a mediocre business?

We Identify Winners Retrospectively

If a stock goes up 3x in 5 years, it is easy to tell ourselves that we were right and that the correct decision is simply to keep holding.

However, a rising share price does not necessarily mean the company will continue compounding for the next decade. Likewise, a stock that has fallen 50% is not automatically a loser.

Sometimes the business is progressing exactly as expected while the market has simply changed what it is willing to pay for it. (Multiple compression)

This is where survivorship bias creeps in. We study companies like Amazon, Apple Nvidia, Microsoft and conclude that the lesson was to never sell your winners.

The issue is that we are focusing on a small number of companies that survived, adapted and compounded for decades. There were many other winners who won for 10 years and suddenly stopped winning, either by stagnating, being disrupted or simply becoming too expensive.

In fact, the duration of winners that have continued winning have shrank considerably in the past century, no doubt due to the advancement of technology, which rapidly accelerates both the upward and downward trajectory of businesses.

In 1958, the average tenure of a company in the S&P 500 index was 61 years. Today, it is just 15 years. Interestingly, this is not even as long as the long term we determined earlier on.

We spend enormous amounts of time studying the businesses that successfully navigated multiple technological cycles, competitive threats, recessions, management transitions and valuation bubbles.

We spend far less time studying the companies that looked equally dominant at some point and then stopped winning.

There have been plenty of these, as judged by the stats above. This can be due to market leadership changes, industries maturing, technology shifts, regulatory changes and management teams changing or simply losing their edge.

In just 26 years, the top 10 has evolved massively. Of the 10 largest US companies in the world in 2000, only 1 remains in the list. That is Microsoft, which has been overtaken by 3 businesses who were not even near the top of the list 26 years ago.

Today’s dominant companies are not guaranteed to remain tomorrow’s dominant companies.

These are not just any 10 random businesses, they were the most dominant businesses in their time, and yet, they have fallen behind.

The counterpoint is that, while these businesses have not remained at the top, perhaps they have still done incredibly well, but been outdone by recent winners.

I looked at the top 10 US companies in 2000, mapped out their annual returns and compared it to the S&P 500. Here are the results:

Even the top 10 US companies, who would be deemed as “winners”, failed to beat the the overall index. Of the 10 on the list, only Microsoft and ExxonMobil have had market-beating returns, while the likes of General Electric, Citigroup and Cisco have struggled.

Key Takeaways

The data we just looked at tells us two things:

1. We want to own the tiny number of extraordinary winners

Since stock returns are so positively skewed, selling a genuine long-term compounder too early can be one of the most expensive mistakes an investor makes.

Yet,

2. Most individual companies will NOT become those extraordinary winners

This means that simply holding onto winners and hoping they remain winners is not a suitable strategy.

That is the paradox of the “letting your winners” run advice.

We want to hold onto great companies long enough to capture the extreme right tail of returns, but in real time, we will not know with certainty which companies are going to remain in that right tail.

It sounds obvious when phrased this way, however in practice, this is incredibly difficult.

Let’s walk through a practical example.

Imagine you bought a company at $10.

5 years later, it trades at $50.

The business performed brilliantly, your thesis was right and you made 5x your money.

Now, imagine that I discovered that same company for the first time today.

The stock trades at $50. Would I buy it?

This is actually the same question existing shareholders should ask themselves, but often avoid.

The only thing that matters at this point is:

What return can this business plausibly generate from $50 onwards?

I think this thought experiment is incredibly useful because ownership creates psychological baggage. When we have made a lot of money in something, we naturally become attached to the investment.

It becomes our winner, and selling feels like abandoning something that has worked. There is also the fear that it keeps going up after we sell, which creates one of the most painful forms of investment regret.

The correct benchmark we should be gauging it against, is opportunity cost.

Capital has an opportunity cost.

Suppose you own a fantastic business trading at a valuation that you believe implies a 7% annual return over the next five years. You also identify another business where you believe the expected return is 15%.

If you decide to keep the existing winner, you are effectively choosing 7% over 15% expected returns. Of course, expected returns are uncertain, and the forecasts may simply be wrong. Incumbent businesses often continue to surprise positively too.

Re-underwriting the winner.

If you had to take away just one thing from this article, I think it would be to continually ask if your stocks are still winners (not from a stock price standpoint, but from a fundamental business standpoint)

The real skill is not holding our winners blindly, but consistently evaluating these businesses, re-underwriting them as they become larger positions and being ready to take profits when the situation changes.

Thanks for reading!

-Gab

P.S. This is a very different piece from my typical article, but I felt compelled to write this, and I hope this is helpful for you if you read through this short piece.

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 (20 Deep Dives and counting)

You can get 30% off your first month with the link below.

30% OFF FOREVER

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.

There have been some worries that this is similar to the investment cycle in late-2023 which resulted in profitability going negative after an aggressive cost-cutting cycle in 2022 and early 2023 that saw it make its first ever profit.

So far, there hasn’t been any evidence of it. Sea has been consistently profitable for 9 quarters so far with no signs that it would go back into negative territory.

A major factor for why Sea have embarked on yet another investment cycle is related to ShopeeVIP, their own membership program. This is similar to Amazon Prime, Coupang WOW, MELI+ and Alibaba’s 88VIP program.

Hence, this launch was not necessarily a surprise. Rather, an inevitability. What has been surprising is the progress and traction of ShopeeVIP just over a year into the rollout.

ShopeeVIP first rolled out to Indonesia in Q1 2025, reaching 2M members by Q2 2025 as it rolled out across other SEA countries throughout the rest of the year. By Q4 2025, it reached 7M members and by Q2 2026, it has reached 15M members with Brazil the latest market it has rolled out to.

However, this does not necessarily imply it was a success, as it could very easily have been heavily subsidised and operated at a loss. (More on this later)

There are 2 reasons why I believe this has been a success:

1. ShopeeVIP has managed extremely high retention rates thanks to the presence of Monee’s credit cards. In Q2 2026, management revealed that average monthly retention remained stable at ~80%.

2. In Q2 2026, VIP members contributed 24% of total GMV across Asia. This is despite VIP members making up ~3.75% of Shopee’s total active user base (400M). This implies that ShopeeVIP members are spending over ~8x more than their non-VIP counterparts.

Given these figures, it would be in the shareholders’ best interest for Shopee to continue doubling down on its efforts to acquire as many ShopeeVIP members as possible.

ShopeeVIP Membership

Every country has slightly different membership perks and costs. Some screenshots below:

ShopeeVIP Unit Economics

To get the weighted average cost of each membership and therefore determine Shopee’s revenue from ShopeeVIP, I’ll take the contribution of each country to total GMV and multiply by the US$ price after converting through a fixed FX rate.

Doing so, we get to a weighted average price of $1.44 per month, which would imply ShopeeVIP currently generates $21.6M in monthly membership fees, or $260M yearly.

ShopeeVIP generated 24% of Shopee’s GMV in Q2 2026, which implies ~$9.2B from members alone. If we assume a cost-to-serve of ~2-3%, it equates to between $184-$276M quarterly. Offsetting membership fees, it comes down to $119-211M. On an annualised basis, Shopee is spending between $480M to $840M on this program. In 2025, Shopee generated Adj. EBITDA of $880M. Therefore, this is a substantial investment by management.

Unit economics may seem bleak if we were to view it from this lens, but it must be noted that this treats the entire cost of serving ShopeeVIP members as an incremental expense. In reality, the purpose of ShopeeVIP is to change user behaviour by increasing purchase frequency and improving retention. It also serves to consolidate Monee wallet share and ultimately increases the lifetime value of its users.

Hiking Membership Fees

This is the obvious area that would dramatically reduce cost-to-serve and generate more revenue for the business. Currently a weighted average of $1.44 ranks among the cheapest subscription programmes for e-commerce across the world.

A good comparison would be the Coupang WOW program in South Korea. Coupang WOW was first launched in early 2019 and membership fees were ₩2,900 per month ($2.12). They then hiked prices about 3 years in when penetration reached ~50% of active users by 72% to ₩4,990 ($3.65) in Dec 2021. Another price hike was introduced in April 2024, with prices raised by 58% to ₩7,890 ($5.77). This equates to a cumulative price hike of 172% while it also expanded its subscriber base.

Shopee is still early on in its phase, with <4% of active users having subscribed to the program. That said, due to ShopeeVIP’s 80% monthly retention rate, I believe it is very plausible that we could see Shopee introduce price hikes much earlier than 3 years in.

Subscribe now

Voucher Optimisation

Another way that Shopee could reduce its overall cost-to-serve, would be by optimising its voucher strategy, increasing minimum spending thresholds and thereby encouraging larger basket sizes.

Increasing the minimum spending requirement would enable Shopee to lower the effective discount rate per order and, larger baskets would reduce logistics cost per dollar of GMV, improving the unit economics of ShopeeVIP.

ShopeeVIP Unit Economics Improvement

Let’s assume that ShopeeVIP has a similar path to Coupang WOW, and hikes membership prices by 172% over the next 3-5 years to $3.91. That would rank it among the middle in the above chart which I think is plausible. Then, let’s also assume that the cost-to-serve falls from 3% to 2.5%.

Under the same assumptions of GMV and membership numbers as Q2 2026, the ShopeeVIP program would be able to reduce the total cost-to-serve from $211M per quarter to $54M per quarter, a massive 74% saving.

The bigger point is that the ShopeeVIP membership fees are a formality, and should not be of material value to Shopee’s bottom-line in the long run. Instead, it would be great if Shopee could run their program at a minimal loss, enabling customer LTV to expand and creating more customer lock-in.

GMV Growth & Logistics

One more thing to note is that ShopeeVIP and SPX Express investments are moving toward the same goal. SPX Express now moves more than 30 million parcels a day across Southeast Asia, Taiwan and Brazil (as of Q4 2025), and that scale is what enables free shipping perks for ShopeeVIP.

Density is ultimately what decides last-mile economics. Logistics being one of the key strengths of Shopee in its battle against TikTok Shop, has to double down and increase their advantage. For instance, a courier who drops six parcels on one street covers his route cost faster than one who drops two. ShopeeVIP is able to manufacture density by increasing order counts, enabling cost per order to fall. Some examples that prove this:

• Buyer purchase frequency rose about 12% year on year in Q1 2026

• Shopee also shared in Q2 2026 that fulfilment order volumes were up >20% QoQ, while >60% of fulfilled parcels arrive the next day in some markets

• Instant service in Indonesia can deliver in as fast as 1 hour; instant delivery

  order volumes grew ~80% YoY, with cost-per-order falling ~20% YoY

Conclusion

One thing to note, is that Shopee has likely already gotten their most engaged users signed up for the VIP program. Therefore, the 8x stat will likely continue going downwards. That said, having members spend 3x, even 2x more than non-members would be a win regardless.

There are a few key metrics to look out for to observe how the membership is going. The % of GMV that ShopeeVIP members spend is an obvious start. Monthly retention is also important. Currently it lands at about 80%. If we see it fall to 70% or lower while membership continues climbing, it could indicate that Shopee is buying users who leave and the cost-to-serve becomes a subscription to churn.

If the first price rise from $1.44 produces a visible drop in members, we would have to re-think whether ShopeeVIP’s perks are sufficient to make this membership as successful as Coupang’s, for example. Currently, the perks are largely built around vouchers and other perks such as free ChatGPT trials, Duolingo memberships etc. I’m not sure if that is enough to convert the longer tail of users.

Regardless, I think ShopeeVIP is a key part of Sea Limited’s story moving forward, and an important step ahead. What gives me confidence is that Shopee has proven over the years to have incredible execution. I think this will be no different.

As always, I will be monitoring this closely, with updates in every future earnings review.

Thank you for reading!

-Gab

Fiscal AI

The charts are courtesy of Fiscal AI.

For me, it is the best stock research platform available for retail investors. If you are interested in the service, they are offering a free 7-day trial for new users and 15% off any paid plan if you use the link below!

15% OFF FISCAL AI

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 (20 Deep Dives and counting)

You can get 30% off your first month with the link below.

30% OFF FOREVER

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.

Portfolio: +4.26%
S&P 500: +2.62%
NASDAQ: +3.99%

August was a strong month for the indexes, reversing a poor June to July stretch where the market had back to back red months. Yet, it was not a great month for all sectors and stocks. There was yet more bifurcation within the MAG7 with NVDA, MSFT and TSLA leading the pack while GOOGLE and AMZN struggled.

We saw a strong rebound in the AI trade in the first half of the month with stocks like MU, SKHY, NBIS, CRWV going up 30-50% in that period. Of course, this happened after the Leopold forced-selling that we discussed last month. Another sector that was incredibly strong was healthcare. Personally, I like that we are seeing healthy rotations on a consistent basis in the bull run that has now gone on for over 3 years.

It must be noted that September has historically been the weakest month for the markets over the past 20 years, with -0.6% monthly returns on average. It was only higher 45% of the time.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +222.14%

S&P 500: +85.78% (Outperformance: +136.36%)

NASDAQ: +113.66% (Outperformance: +108.48%)

Portfolio Performance (YTD)

Portfolio: +8.38%

S&P 500: +12.28% (Underperformance: -3.90%)

NASDAQ: +13.90% (Underperformance: -5.52%)

Personal Thoughts on the Market & Performance

One worry that has been constant in this bull market is inflation. That’s been led by 2 things: ongoing conflicts and AI bottlenecks.

1. The ongoing conflict in the Middle East and disruptions involving the Strait of Hormuz, have kept crude oil and fuel prices elevated, increasing transport and manufacturing costs globally.

2. The second has been the bottleneck in the AI supply chain, with the relentless demand for compute pushing up prices for core products.

As I discussed in the previous portfolio review:

Treasury yields have followed oil higher again, as we saw it make a new high for the year again, hitting 4.804% as I write this. This has very direct implications on stocks, as higher yields make risk-free government bonds more attractive than equities, drive up corporate borrowing costs and reduce the present value of future corporate profits.

This is exacerbated for high-growth stocks for a simple reason, a high discount rate shrinks large future numbers much faster than small current earnings. Since high-growth companies promise big profits years from now, not today, we often see a much larger decline in prices investors are willing to pay.

My Positioning

Read more

The objective behind this is to concentrate my capital on the highest-conviction ideas rather than watering down performance with a bunch of mediocre holdings. This is best summed up by 2 quotes from Warren Buffett:

More than the total number of holdings, I’m more concerned that many are at small single-digit percentages, which make it difficult to move the needle for the portfolio.

In this article, I’ll be sharing my portfolio plans for the next few months, which stocks I’m looking to trim, cut entirely, and how I intend to allocate the excess cash.

In case you missed the most recent article (linked above), I discussed all 15 holdings in my portfolio, ranking them by conviction, quality, risk and valuation.

Read more

Do consider checking out the recent free posts:

If you enjoy the content and would like to see more, consider joining over 100 paid subscribers where you will be able to get access to:

• Full Investment Cases (20 Deep Dives and counting)

• Portfolio Performance & Holdings

• Emerging Markets Focus (SE, GRAB, DLO, MELI)

You can get 30% off the subscription forever with the link below.

30% OFF FOREVER

Read more

Revenue: $399.7M v $369M est. (+56% YoY, +69% FX-Neutral) 🟢

Diluted EPS: $0.18 v $0.19 est. (+87% YoY, +7% QoQ) 🔴

Selected Key Metrics

dLocal saw continued YoY and QoQ growth across all 6 key metrics. Revenue hit $400M for the first time, driven by TPV growth that came in at 92%. The turnaround in TPV growth within the past 8 quarters have been nothing short of extraordinary.

The glaring issue, of course is that while TPV grew 92%, net income growth came in at just 28%. There are a few layers between that, but there is no doubt that take rates is the main culprit behind that gap.

• TPV: $17.7B (+92% YoY) ✅

• Revenue: $400M (+56% YoY) ✅

• Gross Profit: $127M (+29% YoY) ✅

• Operating Profit: $64M (+15% YoY) ✅

• Net Income: $55M (+28% YoY) ✅

• Adj. FCF: $69M (+41% YoY) ✅

Management also announced that they re-purchased 6.9M shares for $86.1M under their $300M share buyback program and have secured a new $150M senior unsecured facility at Term SOFR + 2%.

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

As I always state in my discussion for dLocal, there are really 2 key guiding metrics for the business.

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

Secondly, net take rates. This determines the amount of gross profit generated from TPV, which ultimately affects profitability of the business.

When we multiply these 2 numbers, we get gross profit, which I personally perceive as the true top-line metric for the business.

Total Payment Volume

TPV was up 92% YoY and 26% QoQ, marking the 8th successive quarter of acceleration in TPV. TPV is now at $17.7B and judging by the trajectory, we are likely to see TPV reach ~$25B by Q4 this year.

Net Take Rates (Gross Profit/TPV)

Net take rates are the true share of money that dLocal keeps after paying direct costs like local partner networks, payment rails, processing fees, etc.

This quarter, net take rates fell to 0.72%, less than a quarter of the net take rates 6 years ago this quarter. This is par for the course for any payments business, and especially one like dLocal that serves large enterprise merchants, and hence should not come as a surprise.

Notably, CEO Pedro Arnt said during the call:

Gross Profit

dLocal’s true top-line in my view, is gross profit. That was up 28.4% YoY, the weakest since Q4 2024. As Pedro shared during the call, this was due to the ramp up of a singular ride hailing merchant.

I would personally like to see this number grow at 30%+, and it was disappointing to see it dip below this quarter.

3. Guidance

Management updated guidance for FY2026, guiding for 60-70% TPV growth for 2026. With the first 2 quarters coming in at 74% and 92% respectively, this probably is still a sandbagged target. Gross profit guidance was also slightly raised, although only 2.5% at the low and high end. Operating profit was unchanged, and this is perhaps surprising to some.

Management elaborated, stating that this is due to the prior-year tax adjustment, that would otherwise have operating profit come in at the higher end of the range. They have also been very consistent in their guidance. For instance, in Q4 2025, they stated that the new OPEX baseline introduced post-2025 investment cycle would temporarily pressure 1H26 margins but drive operating leverage improvements in 2H26. (I believe this is where the opportunity for investors lie and I will elaborate later)

Operating profit growth was 15% YoY in both Q1 2026 and Q2 2026. As such, we should see numbers closer to 45% YoY growth in H2 2026 in order to bring numbers in line with management’s guidance.

4. Product Development

dLocal has continued to prioritise expanding coverage. In Q2 2026, they added 7 new APMs, reaching 217 APMs across 39 markets. Every additional market and payment method makes dLocal more valuable to existing and new merchants alike. For example, a business like Spotify having to decide between a provider that services 12 markets and 50 payment methods versus dLocal that services 39 markets and 217 payment methods. Even if Spotify only needs access to 10 countries today, they will still favour dLocal as it gives them optionality when expanding into other markets. Hence, this is a crucial customer acquisition and retention metric.

dLocal also shared that their optimisation suite can deliver up to 10 percentage points of conversion uplift for localised payments. This is huge for dLocal as it changes them from just a generic payment processor into a revenue optimisation product. Take for instance, a generic PSP approves 80% of payment attempts, while dLocal approves 85%. If Uber sends $1B of payment attempts, that extra 5% is $50M of additional completed transactions for Uber.

dMore is also a very important addition in the context of take rate pressure that, as we discussed, is the biggest worry for investors. Currently, dLocal’s primary revenue source is earning a fraction of TPV. Through dMore, it could capture economics across a much larger service stack. For example, if dLocal processes $100 worth of transactions, it keeps about $0.72 today in gross profit.

With dMore, that same $100 transaction enables dLocal to handle FX, tax administration, invoicing, compliance and eventually provide additional merchant services on top. As dLocal matures, this will be a key part of its business.

dLocal also highlighted the use of AI with ~60% of code autonomously written and ~10% in headcount equivalent savings through automation. This is all great news but I would expect it.

5. Management Commentary

1. Existing merchant wallet-share gains are materially exceeding expectations

dLocal’s growth is coming from existing global merchants giving dLocal more volume every quarter. This is corroborated by the 188% TPV retention, and 92% TPV growth. It means just 4% of TPV growth is from new merchants.

2. The wallet-share opportunity remains large

This quote was with regard to the ride-hailing business that largely led to doubling of volumes running through the dLocal platform QoQ. The one obvious concern after this doubling of volume of course is, doesn’t this mean dLocal has already earned most of its business?

Management is clearly saying no and stating that the extraordinary Q2 ramp was heavily concentrated in LATAM, and more gains can be had by expanding that relationship to Africa, the Middle East and Asia.

This means that even a merchant which has already generated enormous incremental TPV can theoretically remain a significant future growth source.

3. Pedro reiterated what is really most important to dLocal

The eventual aim of dLocal is to maximise shareholder value, and that comes by generating more dollars every quarter. Management is essentially telling investors that gross profit dollars matter more than take rate percentages, and I have stated this too in the past.

4. dLocal is prioritising scale for a reason

TPV growth can create a moat, and is not simply a means of stemming the take-rate decline. Scale gives dLocal better bargaining power with processors/acquirers, better FX liquidity and more payment data.

5. Management expects automation to structurally reduce middle and back-office hiring intensity

In the latest Q, dLocal added just 4 employees, compared to 27, 60 and 55 in the previous 3 quarters. As dLocal’s investment cycle comes to a close, I believe we should see similar numbers in the next few quarters as the one in Q2. This will bode very well for dLocal’s FCF per share generation as I will elaborate on later as part of my thesis moving forward.

6. Mexico was a disappointment due to dLocal’s failure to control costs

This is certainly one of Pedro’s strong points, acknowledging the mistakes they’ve made and seeking to correct them. He is one of the most excellent communicators and is always transparent.

6. Conclusion

Overall, this was a rather mixed quarter. TPV growth coming in at 92% is incredible, and I continue to be amazed at how Pedro has turned this around since becoming co-CEO and now sole CEO.

Margins continue to be an overhang, but I remain very optimistic.

If we look at the next few quarters, I really like the setup of dLocal. Gross profit growth is in the 30% range, they are extremely capital light, and taking market share. The investment cycle has just ended, headcount is now flat moving into H2 and AI efficiencies are expected to cut costs. We are seeing operating leverage in real-time. Combined with buybacks, I think we are about to see FCF/share growth go vertical soon, and valuation will look ridiculously cheap if the stock remains range-bound.

Below, I will write about what I plan to do with the stock:

Read more

EPS: $0.70 v $0.79 🔴

Sea Limited reported earnings on 11th August before the market open.

The stock closed the day up 15%.

Subscribe now

Selected Key Metrics

Total Gross Profit: $3.55B (+47.3% YoY)
Operating Income: $650.3M (+33.3% YoY)
Total Net Income: $458.1M (+10.6% YoY)
Total Adjusted EBITDA: $917.2M (+10.6% YoY)
Provision for Credit Losses: $555.2M (+71.5% YoY)
GAAP Diluted EPS: $0.70 (vs $0.65 in Q2 2025)
Cash & Cash Equivalents: $10.3B (down from $10.6B in Q2 2025)

E-Commerce (Shopee)

Gross Orders: 4.3B (+27.5% YoY)
GMV: $38.3B (+28.4% YoY)
Revenue: $5.6B (+48.2% YoY)
Adjusted EBITDA: $255.4M (+12.2% YoY)

Digital Financial Services (Monee)

Consumer & SME Loans Outstanding: $11.1B (+62.5% YoY)
Revenue: $1.4B (+58.9% YoY)
Adjusted EBITDA: $288.0M (+12.8% YoY)
90-Day NPL Ratio: 1.0% (down from 1.1% in prior Q)

Digital Entertainment (Garena)

Bookings: $763.5M (+15.5% YoY)
Revenue: $746.6M (+33.5% YoY)
Adjusted EBITDA: $429.8M (+16.7% YoY), 56.3% of bookings
Quarterly Active Users: 666.3M (roughly flat YoY)
Quarterly Paying Users: 68.1M (+10.2% YoY)
Paying User Ratio: 10.2% (vs 9.3% in Q2 2025)
Average Bookings Per User: $1.15 (vs $0.99)

Share Repurchase Program

Repurchased 4.7M shares for an aggregate of $417M = $88.72 per share

Sea Limited continued its top-line acceleration, this time to 48%. This is an incredible rate of growth for a business now run-rating at $30B annual revenue. The bottom-line missed estimates which is to be expected with the business very much still in re-investment mode (although management did suggest it could be coming to an end).

All 3 segments are firing on all cylinders and I continue to believe Sea is one of the most underrated businesses out there. This is potentially the most important business in all of Southeast Asia and $70B for it, appears cheap.

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 hit another all-time high in GMV and gross orders, and saw its 8th consecutive quarter of sequential growth. Management reiterated their target for 25% full-year GMV growth, which I believe is very achievable.

They also re-affirmed their target of at least $1B in Shopee Adj. EBITDA for FY2026.

Core marketplace revenue was up 65.6% YoY, more than double the pace of GMV growth and is a clear indicator of monetisation inflecting in real-time.

Take-rates also increased, climbing to 14.6%, up from 13.7% in Q1. It is also notable that management addressed this specific point in the call:

Profitability

Another key point that was one of the reasons SE fell in the past 8 months is the question of Shopee’s profitability.

Management had guided to “at least flat” EBITDA in 2026. For context, Shopee EBITDA in 2025 was $880M. The market took it to mean that EBITDA margins would fall from 0.7% to 0.55%, with the assumption of 25% GMV growth.

This quarter put those fears to bed. Adj. EBITDA margin inflected upwards to 0.67%, and management shared that Adj. EBITDA would come in above $1B, which would mean EBITDA margins will come closer to 0.7% than 0.55%, satisfying the market.

ShopeeVIP

We also saw continued ShopeeVIP member growth, one of the primary reasons management is sacrificing near-term profits.

• 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)

• Q2 2026: 15M members (+45% QoQ)

Management shared that these 15M members, which represent about 3.75% of the Shopee user base of 400M people, make up 24% of GMV spend. It is simply a no brainer to sign up as many people as possible. At current growth rates, we are likely to see ShopeeVIP member numbers end closer to 30M at the end of the year, which is simply incredible, and would likely make it one of the most used subscription services in Southeast Asia.

Management also shared that there was ~80% average monthly retention, a number that would have been impossible to imagine half a decade ago. Monee has been crucial to this. The majority of users in Southeast Asia do not have access to credit cards, which makes a subscription system simply unfeasible.

Chris Feng alluded to this in the Q4 2025 earnings call:

Also, a reminder that ShopeeVIP was only introduced to Brazil in April and as of end June had over 1M members. MELI’s own subscription program, MELI+ is estimated to have about 30M members in the region, majority of which originate from Brazil. I think it is not a crazy shout to suggest Shopee can convert over 10M members in Brazil alone.

Shopee Logistics (SPX Express)

Shopee has continued to invest in its logistics capabilities, which is now a key cornerstone of their moat within the business. Instant service in Indonesia can now deliver as quickly as 1 hour, with instant delivery order volumes growing ~80% YoY and cost-per-order falling ~20% YoY in Q2 2026.

Fulfilment order volumes are also up > 20% QoQ, with >60% of SPX-fulfilled parcels arriving the next day in some markets.

Other Interesting Stats

• Shopee orders through Facebook increased by >85% QoQ, largely driven by Reels

• Extended IG collab to all 8 core markets, promising early results from ID (1st launch market)

• Orders from livestreaming and short-form video up >50% YoY, now accounts for >25% of physical good orders in SEA

• Avg. monthly new active buyers up 35% YoY, acceleration from previous Q's

2. Monee (Digital Financial Services)

Monee continued its hyper-growth, delivering another record quarter for the loan book, up 62.5% YoY to $11.1B, while the NPL ratio fell from 1.1% to 1.0%. Active credit users grew around 34% YoY to more than 40 million while ~5.3 million first-time borrowers were added in the quarter.

Management also shared that average loans outstanding per user grew around 20%, and the Philippines became the 5th market to cross $1B in loan book volume.

A large part of this is increasingly the introduction of AI tooling. Management shared that their latest underwriting models are pre-trained on behavioural and transactional data across the ecosystem, using transformer architectures similar to modern large language models.

This has led to approval rates growing by 10% while maintaining similar levels of risk. Sea has also deployed AI to verify user-submitted income documents across markets and languages, cutting review time by about 95%. These are all low-hanging fruit, and I believe AI has a long way to go to benefitting businesses that are open to it.

GAAP revenue for the quarter came in at $1.4B while Adj. EBITDA climbed to $288M. As I’ve highlighted in the past few quarters, one of the key items to watch within Monee is the expansion of credit use cases beyond Shopee.

Management disclosed that off-Shopee loans accounted for >20% of the total SPayLater portfolio as of end-June, and were as high as 35% in some markets. The standalone ShopeePay app which is currently live in Indonesia, Thailand, Malaysia and Vietnam, saw monthly transacting users more than double YoY in 2Q26. The app is set to be launched in Brazil soon.

3. Garena (Digital Entertainment)

Garena had yet another steady, good quarter like we have to come expect. One of the points bears keep raising is with regard to the sustainability of Free Fire, and how Sea would be in trouble if Garena stopped printing cash.

The truth is that is simply outdated. Yes, Garena continues to print cash, over $400M in Adj. EBITDA this quarter which is not a small amount. However, Shopee is now generating $250M a quarter in Adj. EBITDA while Monee is generating nearly $300M a quarter. More importantly, these 2 businesses are far from optimal profitability, with management continually re-investing excess profits back for more growth.

At 3% Adj. EBITDA margin, which management has acknowledged is a very clear and realistic target for Shopee, it alone would be printing north of $1B in Adj. EBITDA per quarter, dwarfing Garena.

Garena also announced 2 new mobile games this quarter, built on globally-recognised IP. First, Palword Online, an open-world multiplayer survival adventure game, developed and published by Garena under license from Pocketpair. Secondly, Monster Hunter Outlanders: a survival hunting action game developed by Tencent based on Capcom’s iconic franchise. These are 2 exciting titles that could potentially diversify the Garena revenue base away from just Free Fire.

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.

1. Shopee's investments are already producing better unit economics.

In short, management is saying that the investments currently suppressing Shopee's reported margins are not necessarily permanent expenses. They are being deliberate and systematic in their spend. This backs up my point that they are doubling down on re-investments because they see strong ROI.

2. Management thinks the move towards 2-3% Shopee EBITDA margins is relatively straightforward.

Management is giving the margin bridge pretty clearly here and I think this is the most colour they've given to this. My belief is that this gave the market a lot of confidence which contributed to the 15% move post-earnings.

3. Shopee still sees meaningful room to increase overall take rates.

One concern coming into the call was, especially with the incredible take rate trajectory over the past few years, would Shopee be able to maintain that moving forward? They answered this question pretty directly.

4. Shopee VIP is crucial and has scaled very quickly.

Shopee VIP was always inevitable and should fuel the next growth leg. The strategic value is not just about subscription revenue, though it will surely help. However, the key is retention, higher frequency and more spending which creates higher customer LTV.

5. Shopee Brazil is growing faster than the market, and is Shopee's fastest-growing market.

6. AI is materially improving Monee's underwriting models.

AI beneficiaries are everywhere, but nobody talks about e-commerce businesses. I continue to believe that they will be one of the biggest beneficiaries, firstly from simple tasks like customer service, next in logistics and finally in fulfilment.

5. Concluding Thoughts

The past few months have been much better for Sea stock, after several months of getting hammered.

Throughout this period, what has stayed constant is Sea’s consistently strong earnings results. I remain a huge believer in what management is doing, and the proof of their execution has been clear to see. Yes, the stock will remain volatile, but I believe this is a long-term compounder and the story is still in the early stages. If Sea continues to be ignored by the market, management will simply ramp up buybacks again, returning value to shareholders.

On any reasonable sum-of-the-parts or conservative discounted cash flow, the current price still does not reflect what these three segments are worth, and hence I remain a buyer on any dips.

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 (20 Deep Dives and counting)

• Southeast Asian coverage of industries and companies

Fiscal AI

The charts are courtesy of Fiscal AI.

For me, it is the best stock research platform available for retail investors. If you are interested in the service, they are offering a free 7-day trial for new users and 15% off any paid plan if you use the link below!

15% OFF FISCAL AI

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.

So instead, I will look to do a quick review of each company’s earnings here in one post. I will also detail what I intend to do with each position: add/trim/hold (actually just added to one of them yesterday).

Read more

Revenue: $997M v $991M est. (+22% YoY, +21% Constant Currency)

EBITDA: $168M (+54% YoY), 16.9% Margin v 13.3% A Year Ago

GAAP Profit: $235M v $20M in Q2 2025

EPS: $0.06 v $0.02 est.

Selected Key Metrics

• On-Demand GMV: $6,463M (+21% YoY, +22% CC)

• Group MTUs: 54M (On-Demand MTU +17% YoY)

• On-Demand GMV per MTU: +3% YoY

• Gross Loan Portfolio: $2,297M (+197% YoY, roughly doubled ex-Superbank)

• Loans Disbursed: $1.2B (+72% YoY)

• Customer Deposits: $2.5B

• Operating Profit: $19M (+$12M YoY)

• Adjusted Free Cash Flow: $73M (-39% YoY)

Subscribe now

What stands out most from these numbers is probably the GAAP profit of $235M v $20M last year. However, it must be said that this is largely an accounting abnormality and not a 10x in earnings growth.

The $235M profit includes a one-time $307M non-cash gain from re-measuring its stake in Superbank that it has just consolidated into their portfolio, a $66M favourable tax movement from recognising deferred tax assets, and partially offset by a $183M fair value loss on financial assets and liabilities (this is largely the convertible-note derivative).

If we strip these out, the underlying result is a $19M operating profit, which is a +$12M gain YoY. Therefore, the key is to watch the operating profit of the business at least in the short to medium term as management expects “a degree of variability tied to fair value measurements” to keep distorting the profit line in H2.

Adj. EBITDA is where we can see true operating leverage in the business, with it growing +54% YoY, over double the revenue rate. This also represents 18 straight Q’s of Adj. EBITDA growth with expanding margins to boot.

Grab also raised FY 2026 guidance to $4.10B-$4.15B in revenue and $720M-$740M in Adj. EBITDA, reflecting the strength of their underlying business alongside the consolidation of Superbank and the acquisition of Stash.

The board also authorised an additional $750M in share repurchases, bringing cumulative authorisation to $1.75B since 2024.

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

Deliveries remains the largest segment by GMV and Revenue.

Revenue grew 21% YoY (19% CC) to $531M, driven by GMV expansion and continued momentum in advertising. GMV grew 22% YoY (24% CC). Adj. EBITDA margin also grew from 1.8% to 2.3%, tracking closer towards management’s long-term target of 4+% margins.

I think the increase in margins is particularly impressive as they continue to pour growth into the lower-margin grocery business.

Advertising continues to be a strong driver in deliveries. Quarterly active advertisers on the self-serve platform grew 21% YoY and average spend per advertiser grew 24% YoY.

Mobility

Mobility was certainly hit by the fuel crisis in Q2. Revenue grew just 12% YoY, a clear deceleration from 19% in Q1. GMV grew 18% YoY to $2,214M, but transactions grew 28% YoY.

This also means that part of the reduction in revenue is due to average ticket falling, which has been part of Grab’s affordability plan since last year. Through the fuel spike, it is likely that Grab has leaned more into saver products and driver incentives to keep both sides of the marketplace intact.

As I’ve said, and continue to say, supply is the limiting factor for Grab, not demand. These factors led to net take rate compression with mobility take rate coming in at 15.0% vs 15.8% a year ago.

Fortunately, Grab’s efforts to keep drivers on the road have paid off, with monthly active driver-partners hitting an all-time high, up 19% YoY, and driver earnings rising 4% YoY, funded partly by $7M of direct fuel-support spend.

2. GFin, Financial Services

Q2 was a particularly strong quarter for GFin, when we look at the headline numbers. However, some of it was inflated by the consolidation of Superbank.

On 20th May 2026, Grab announced that their combined direct and indirect shareholding in Superbank will increase to over 50% from 37% prior, due to the transfer of Singtel Alpha Investments Pte. Ltd.'s shareholding in Superbank to GXS Bank Pte. Ltd.

As a result, Superbank’s numbers was consolidated into the GFin segment for the first time this quarter. Grab’s Gross Loan Portfolio increased by 197% to $2.318B this quarter, but management shared that organic lending growth ex-Superbank was closer to 100%. Those are still excellent numbers.

It is good to see that numbers are growing more aggressively and faster than its previous pace, which I have spent the last few quarters arguing for. I believe at Grab’s scale, it should be growing its loan portfolio at ~100%, especially as it has the cash to support it.

Management also shared that the Fin Services segment will be profitable in the 2nd half of 2026, as promised almost 4 years ago.

Superbank

Management revealed more numbers on Superbank this quarter. Superbank now serves 7.4M customers with daily transactions above 1 million. It was profitable for the full year in 2025, posting a pre-tax ROE of 5.7% and a cost-to-income ratio of 55% this quarter. Management expects this to fall below 50% by year-end. Over 60% of Superbank users also use Grab and OVO, which shows the ecosystem flywheel appears to be working as expected.

Stash

In Q1, Grab announced that it would be acquiring Stash for US$425 million for a 50.1% initial equity stake, with the remaining 49.9% bought at fair market value over three years.

This initial acquisition is expected to be completed in Q3 2026, and would add an AI-powered wealth platform with $5.5B in AUM. Grab shared that its AUM is growing at 22% YoY, with over 1M active subscribers.

This moves Grab’s GFin beyond lending into wealth, which I believe is the logical next step for a deposit base of $2.5B.

Subscribe now

3. Key Call Commentary

Grab again hosted a live earnings video call with their IR head, Ken Lek, taking charge of summarising questions and reading them out. On the call were the usual trio of CEO Anthony Tan, CFO Peter Oey and COO Alex Hungate.

Here are some of the key quotes that I thought were useful for investors.

On AI Efficiency for the Business

On Uber’s CEO Dara Exiting The Board & Potential Competition

As I highlighted when the news first broke, this is nothing more than a necessary step-down to ensure compliance to regulations. It is not a fracturing of the Grab-Uber relationship and I do not expect it to change. Uber still remains the largest single shareholder in Grab.

On GrabMart’s Growth

As I discussed in the $65B quick-commerce opportunity piece, I believe Mart is a massive opportunity for Grab. It is great to see that Grab expects deliveries margins to continue growing despite the focus on groceries and the expanding contribution to overall GMV.

On Indonesia’s Commission Caps and Potential for Spillover

This addresses the precedent-and-spillover worry that I flagged last quarter, at least for now. I believe this is something that has to be continually watched in the next few quarters.

On Outlook for Fuel Prices & Mobility Margins

On the AV Rollout in Southeast Asia

4. Positives & Negatives

Positives:

1. Deliveries expanded margin despite grocery growth

Deliveries segment EBITDA reached 2.3% of GMV, up 45bps from 1.8%, on 24% constant-currency GMV growth. The subtle point behind this is that GrabMart, the lowest-margin segment, grew at 1.7 times the rate of food deliveries and therefore diluted the segment mix. This means the underlying food delivery margin expanded by more than just 45 bps.

Advertising is the key lever of this margin expansion. Active advertisers grew 21% and spend per advertiser grew 24%, so advertising revenue grew roughly 50% YoY, and advertising is close to pure incremental margin. I wrote a piece on GrabAds a year ago, which I highlighted as Grab’s 4th pillar. I would highly recommend reading the piece to understand the potential of this segment:

2. Management is prioritising Financial Services

GFin' revenues grew 59%, loans disbursed grew 72% to an all-time high $1.2B, while the segment loss narrowed to negative $15M from negative $26M. Management reiterated H2 2026 segment profitability and projected the loan book to exceed $3B by year-end.

What is so key about Grab’s Fin Services business is their funding model. Grab acquires borrowers at near-zero marginal cost through the existing super-app that they own. They then underwrite on proprietary transaction and behavioural data, and fund the loans with $2.5B of low-cost DigiBank deposits. This is the exact playbook that Nubank and Sea’s Monee have employed to great success.

It is a structural edge over standalone FinTech lenders who fund through wholesale markets. Encouragingly, Superbank itself is already profitable and Stash, their newest acquisition is too.

3. Capital Allocation has stepped up tremendously in the past year

One of the key negatives I highlighted last year was Grab’s massive $6B net cash position, which I thought demonstrated poor capital allocation.

In the past year, they have made several large acquisitions and investments, as well as share buybacks. I think these are very positive signs.

For instance, Grab has acquired FoodPanda Taiwan for $600M, Stash for $425M (initial 50%, which means total of ~$850M), invested in Infermove, Momenta, WeRide, Vay Technology, May Mobility etc. They have also authorised another $750M buyback following the completion of $400M of their $500M share buyback plan. Most of the deals will close in H2 2026 which should bring net cash down to ~$3B.

It is also good to see that Grab’s management views the current stock price to be depressed and attractive levels to repurchase shares. I continue to believe that there is a severe dislocation between the stock price and fundamentals of the business.

Negatives:

1. Overall incentives are on an upward trajectory

On-Demand incentives reached 10.9% of On-Demand GMV, up 72bps YoY. Since incentives are contra-revenue, this increase is mechanically dragging the take-rate down even though unit contribution might be healthy. Hence, part of the take-rate decline is an accounting side-effect.

That said, overall incentives increasing is still not good for the business. In Deliveries, Grab is still out-earning the rising incentives. However, in the Mobility segment this quarter, segment margin is contracting. This is partially due to multiple factors such as Indonesia’s commission caps, the oil crisis etc. However, it is still key to watch whether this continues in future.

2. Revenues have not accelerated

Despite GFin becoming a larger portion of overall revenues, it is notable that Grab’s revenues have not accelerated in the past 8 quarters or so. Personally, I find that disappointing as Grab’s rivals have weakened considerably in the past few years.

Gojek is a declining force in Indonesia, their largest market, while FoodPanda and Deliveroo have gradually exited the region. Much of the growth appears to be coming inorganically now too, through acquisitions. I would like to see more organic growth from the core business.

5. Conclusion

Overall, I thought this was a strong quarter for the business amidst the oil price headwinds. We saw a narrow beat in the top-line, no doubt due to uncertainty in the region.

In March, Grab invested heavily in on-demand driver incentives to spur supply amidst the peak of the oil crisis. Since then, monthly active drivers have scaled to an all-time high. This is extremely bullish news for the business, as supply remains the main limiting factor for growth.

Since then, driver incentives have declined which is a good sign considering the still-elevated oil prices.

My position in the stock remains unchanged for now. I have not added to the stock recently as I want to see a genuine acceleration in revenues. The thesis remains on schedule as operating leverage is driven across the business. I remain hopeful that the Fin Services business can accelerate and garner larger market share in the region.

Grab is still one of my core holdings, and I believe the stock price is dislocated from the business. We are now moving into the 2nd year of full profitability, and I believe the stock will eventually follow the earnings trajectory.

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 typically 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)

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.

Portfolio: -0.94%
S&P 500: -0.82%
NASDAQ: -4.15%

July was the second red month in a row for the indexes. It was a very mixed month for the market, with a few Big Tech names performing strongly: MSFT +24.58%, AMZN +13.95%, AAPL +6.76%.

The pain was largely in the AI trade, especially memory names that saw a pretty sudden and large pullback, with the reason for the speed of the pullback explained in the last couple of days, Leopold Aschenbrenner’s margin calls and subsequent forced selling. This was not the only forced selling, as Koreans also saw massive liquidations, with sources revealing that 1.2 million retail accounts hit margin calls, erasing over $2T in market value from the KOSPI since its June peak.

In my view, there were 3 drivers for these mixed results:

1. The AI CAPEX trade was confusing

In the past year, the rising tide lifted all boats, with every company in the AI stack benefitting from increasing CAPEX, largely fuelled by the hyperscalers. This month was a good tell that things have changed.

Frankly, from my point of view, it was very confusing. If we rank the hyperscalers by their cloud growth numbers during earnings this Q, we get GCP at 82%, Azure at 43% and AWS at 36.7%. However, when we rank them by share price reaction, we get the exact opposite.

The reason was also not CAPEX discipline, as I intuitively thought. Amazon raised full-year CAPEX from roughly $200B to $220B and reported TTM free cash flow of negative $7.6B, yet the stock went up 13%. Google raised its guide by $15B at the midpoint with free cash flow at negative $5.9B and went down 7%.

2. Moonshot AI released Kimi K3 on 17 July

Moonshot AI unveiled Kimi K3 at the World Artificial Intelligence Conference in Shanghai on 17 July and within hours the market treated it as a second DeepSeek moment.

Kimi K3 is an open-weight Chinese model, similar to DeepSeek. The results of the release was especially surprising, with the model benchmarking within a few points of the frontier, trailing only Claude Fable 5 and GPT-5.6 Sol, and doing it under export controls, which implied the American compute advantage was worth less than the CAPEX assumed. This meant the terminal value of the Western infrastructure buildout had to be marked down structurally.

However, when we look deeper, it is much less worrying for the infrastructure players. Kimi K3 is 2.8 trillion parameters, the largest open model ever released, up from roughly one trillion for K2, and it is priced at $3 and $15 per million tokens, identical to Claude Sonnet 5 and 3-4x what the previous Kimi cost.

4 days after the launch, Moonshot paused new subscriptions because its own GPU capacity had run out. It turned out that Kimi K3 was much bigger, more expensive and compute-constrained than what the market initially thought.

3. Forced Selling

In my view, this was the most important reason for the magnitude of the sell-off this month. Leopold Aschenbrenner's Situational Awareness ran roughly $225M into as much as $45B in under two years and was up 439% net YTD going into July.

His fund was long AI infrastructure and short software at nearly 4x leverage. In July both legs went the wrong way, with many infrastructure names falling 40-50% while software names rallied. This led to a complete wipeout of capital for the portfolio. Hilariously, Citadel, the very market maker that shifted its outlook on 28th July to predict a surprise Federal Reserve interest rate hike, bought his entire book. Of course, the rate hike didn’t happen, but it was enough to force Leopold to sell out.

This, coupled with the leverage single-stock ETFs that have been discussed about a lot in the Korean market, led to cascading liquidations.

Portfolio Performance (Since Inception)

(Portfolio Inception Date: 27th October 2023)

Portfolio: +207.68%

S&P 500: +79.77% (Outperformance: +127.91%)

NASDAQ: +103.43% (Outperformance: +104.25%)

Portfolio Performance (YTD)

Portfolio: +3.52%

S&P 500: +8.65% (Underperformance: -5.13%)

NASDAQ: +11.43% (Underperformance: -7.91%)

Personal Thoughts on the Market & Performance

Read more

It is difficult to understand how a business with fewer than 1,000 employees can make $5.48B in revenue, grow 70% year-on-year, and convert 65% of revenue into net income. In many ways, it is a 1 of 1 business.

It has seen an incredible 91.5% drawdown in 2022, followed by an even more astounding 80x return from those lows within 4 years.

Diving deeper into the history of the business makes it stand out even more as an anomaly, and I think there is not a more misunderstood business in the large-cap space.

In this piece, I hope to bust the myth of AppLovin and determine if this is a truly generational business, or a deservedly overlooked one.

Table of Contents

1. Introduction

2. Company History

3. Business Model & Products

4. Moats & Differentiation

5. Competitive Landscape

6. Financials

7. The Short Thesis & SEC Investigation

8. Ownership & Management

9. Valuation

10. Bull and Bear Case

11. GabGrowth Quality Score

12. Concluding Thoughts

1. Introduction

To understand how AppLovin makes money, we must first understand the concept of performance advertising.

If we break it down to first principles, it is simply the business of getting a specific person to take a specific, measurable action. For instance, installing an app, buying a product, or subscribing to a service, and charging the advertiser based on how well that action is delivered.

It is the unglamorous, yet deeply lucrative core of digital advertising, distinct from brand advertising (where you pay to be seen and hope it works). In performance advertising, the advertiser can measure return on ad spend (ROAS) to the dollar and will keep feeding budget into any channel where a dollar in, reliably returns more than a dollar out.

For roughly 15 years, this has been a duopoly. Meta and Google capture the majority of global performance advertising dollars because they own both halves of the equation:

• The inventory (billions of logged-in users and the attention they spend)

• The data (who those users are, what they search, what they buy, what they linger on)

The closed loop between owning the audience and owning the outcome data is what makes their ad machines so hard to beat. Everyone else, also known as the "open internet" and the independent ad-tech sector, has fought over the remainder, usually at thin margins, frequently going bankrupt.

Meta and Google tapped into that durable profit pool for years, leaving it effectively closed to new entrants.

AppLovin is the first credible attempt in years to pry open a third position in that profit pool and it arrived from a strange direction, mobile games.

Games turned out to be the perfect training ground for a performance-advertising AI, because a game generates an enormous, continuous stream of measurable user actions (installs, sessions, in-app purchases), and the advertisers (game studios) are ruthless ROAS optimisers who will move budget instantly when the math improves.

AppLovin built an AI engine, AXON, that learned, inside that environment, to predict which user would convert and how much to bid for the impression, in real time, across billions of auctions.

Once that engine was demonstrably better than the alternatives at selling mobile games, the obvious question was, is this technology transferrable to other non-gaming sectors? For instance, would AXON be able to sell running shoes, supplements, software subscriptions, and streaming services?

Management estimates the non-gaming opportunity is five to ten times the size of the gaming market that AppLovin currently dominates. At the end of June 2026, AppLovin has opened its self-serve advertising platform and a dedicated e-commerce product to general availability.

This product launch will determine where the stock goes over the next decade, and will be a large part of what we will discuss below.

2. Company History

Before we get any further, it is important to understand the journey of AppLovin as knowing the full origin story gives a much better context of the business it is today.

Founding (2012-2014)

AppLovin was founded in 2012 by Adam Foroughi, John Krystynak, and Andrew Karam.

Adam, born in 1980 to a family that emigrated from Iran around the Iran-Iraq war, studied economics at UC Berkeley. Prior to AppLovin, he had already run 2 successful advertising businesses focused on performance advertising.

AppLovin was built to solve a problem that began brewing from 2008, the year the App Store was launched by Apple. By 2012, there were hundreds of thousands of games in the App Store competing for attention from consumers.

The problem facing every mobile developer was the same: how do you get users to install your app when no one knows it exists?

The mobile advertising market was new, developers had difficulty attracting users and building a steady income, and the market was split into many parts without easy access to good marketing and monetisation tools. There were ad networks, but they were fragmented, opaque, and often ineffective. A gaming studio in San Francisco had to deal with ten different networks, each with separate dashboards, separate contracts, and no unified way to know which one was actually working.

AppLovin's earliest product was straightforward. It was a performance advertising network connecting two parties:

• Advertisers (mobile app developers, particularly game studios) who wanted new users for their apps.

• Publishers (other mobile app developers) who had users already inside their apps and could show ads to those users.

The company most commonly charged on a cost-per-install (CPI) basis for each successful conversion. Publishers connect to ad networks like AppLovin to find the best bids for their inventory. Whenever the network successfully places an ad with a publisher, it pays the publisher on a cost-per-thousand-impressions (CPM) basis and keeps a portion of the proceeds for facilitating the transaction.

This was the entire original business model, which was effectively a brokerage where AppLovin matched supply (publisher ad inventory) with demand (advertiser budgets) and took a spread.

Impressively, AppLovin took almost no outside money at the start (around $4M in seed funding from Streamlined Ventures, the Webb Investment Network and others) and was profitable early. Adam took founder's equity and no meaningful cash salary for the better part of a decade. This shaped a culture of capital discipline that still defines the company.

Early Growth & First Pivot (2014-2018)

The company scaled as an ad network during the surge in iOS and Android app install campaigns. AppLovin was ranked #10 on the 2016 Deloitte Fast 500 North America list.

In September 2016, AppLovin agreed to be acquired by Chinese private equity firm Orient Hontai Capital for $1.42B. The acquisition was subsequently abandoned after opposition from CFIUS, and was converted instead into a debt investment. In 2018, AppLovin sold a 9.98% minority stake to KKR, raised debt, and restructured to preserve US-aligned governance and founder control.

The CFIUS block turned out to be a crucial decision. It forced the company to stay independent, retain founder control, and find its own path to scale rather than becoming a Chinese-owned ad network, which would have severely constrained its data access and US market positioning.

In July 2018, AppLovin launched Lion Studios, which works with mobile developers to publish and promote their games. However, the real purpose was to feed AppLovin’s advertising engine. Owning games gave AppLovin full-funnel, first-party data on how users behave after they install, which was the exact signal its ad models needed to predict and price conversions.

In September 2018, AppLovin acquired MAX, an eight-person in-app header bidding startup founded less than a year earlier by Jim Payne, the same Jim Payne who had co-founded MoPub and sold it to Twitter in 2013. (More on this later)

This acquisition solved a structural problem that was plaguing the business. As a pure ad network, AppLovin could help buyers acquire users but couldn't actually help developers sell their inventory well.

The industry standard at the time was the "waterfall" model, where publishers offered each impression to ad networks sequentially, in a pre-set order ranked on historical eCPMs and fill rates. If network A passed, the impression went to B, then C. The process was queue-based, slow, and systematically underpriced inventory, because rankings reflected stale historical averages rather than what any bidder would actually pay for that specific impression.

MAX replaced the waterfall with in-app real-time bidding where every impression is auctioned to all demand sources simultaneously and clears at the highest bid. Developers immediately earned more per impression, while advertisers still came out ahead as long as they bid below expected return. MAX was the first move toward owning both sides of the transaction. AppLovin would now sit between publisher inventory and advertiser demand, capturing bid-stream data from both directions, and it transformed AppLovin from a platform that only served buyers into one serving both advertisers and publishers.

Acquisitions and Vertical Integration (2018-2022)

Also in 2018, private-equity firm KKR made a minority investment that valued AppLovin around $2B. That capital funded a sustained acquisition spree across both sides of the business. The logic was to build a closed-loop stack where AppLovin owned supply, demand, measurement, and first-party behavioural data simultaneously.

On the supply/demand side, AppLovin built out its ad infrastructure by acquiring the MAX in-app bidding/mediation platform as we discussed, and acquiring mobile-attribution firm Adjust, and then purchasing MoPub from Twitter for $1.05B.

Adjust, acquired for roughly $1B in 2021 alongside the IPO, solved the measurement problem. Attribution (the process of identifying which ads, clicks, and marketing channels lead to a sale or action) in mobile advertising was a black box where an ad passes through impression, click, install, usage, and payment, with data scattered across systems advertisers can't see.

Advertisers could measure outcomes cleanly on Meta but not on independent networks, capping how much budget AppLovin could win. Adjust let advertisers break down what each dollar produced, turning AppLovin into a platform that could take accountability for results, which was the foundation of its LTV-based optimisation pitch.

MoPub can be best understood as a massive ad marketplace that aggregated ad space from numerous apps (45,000 to be exact), selling it to advertisers, taking a cut in the process. This acquisition solved the scale problem that had limited AppLovin (as their data was focused only on gaming). MoPub aggregated inventory across tens of thousands of non-gaming apps, (news, utilities, lifestyle) reaching roughly 700M daily active users, expanding both AppLovin's inventory and its audience diversity beyond gamers, and injecting a continuous stream of bid data into the platform.

On the content side, it bought game studios such as Machine Zone, PeopleFun, and Belka Games to generate first-party demand and, critically, the proprietary data that would train the ad engine. This combination meant AXON could train on more diverse outcome data than any competitor.

On April 15, 2021, AppLovin went public on Nasdaq at a valuation of approximately $28B, pricing at $80 per share.

However, Apple's App Tracking Transparency (the post-IDFA privacy regime) was released about a week after AppLovin went public, which gutted the signal that mobile advertisers relied on. To add to that, the SPAC/growth bubble burst in 2022, resulting in AppLovin stock that was burning cash at the point to lose over 90% of its value.

In the midst of that in August 2022, AppLovin made an offer to buy Unity Technologies for $17.54B in stock. Unity's board rejected the offer and committed to completing its acquisition of ironSource. Had the Unity deal gone through, AppLovin would have owned the dominant game engine and the dominant mobile ad network simultaneously, an extraordinary chokehold on the mobile gaming economy. Its failure left both companies weaker heading into 2022.

Adam Foroughi’s Bet on Himself (2022-2024)

AppLovin’s stock price fell to as low as $9 in 2022. Around that period, Adam went to the board and proposed that his entire compensation for the next 4 years be paid out as a performance-stock-unit (PSU) grant, only if the stock recovered massively. The first hurdle was ~120% stock price appreciation, and the final hurdle requiring over 380% appreciation with a limited time window and no cash bonuses.

The recovery was arguably driven by a single product. In mid-2023, AppLovin shipped AXON 2.0, a rebuilt machine-learning engine for the ad business. The improvement in ROAS was large enough that advertisers shifted incremental budget almost immediately. Revenue and margins inflected, and the stock began a rebound that has now exceeded thousands of percentage points off the 2023 base.

Pivot To Pure Ad Company (2024-2026)

The management team soon realised that the underlying algorithm behind AXON 2.0 was the asset. Therefore, it doubled down on their crown jewel. Here is the timeline of events that unfolded:

• Feb 2024: KKR sold down part of its stake in a secondary offering as the stock recovered.

• Late 2024: AppLovin raised debt (senior notes) to fund buybacks rather than dilute.

• May 7, 2025: Announced the sale of its entire mobile gaming business. All ten studios, including franchises like Wordscapes, Project Makeover, Hexa Sort, Cooking Madness, West Game, and Clockmaker were sold to Tripledot Studios.

  • AppLovin received roughly $400M in cash plus equity equal to about 20% of Tripledot (reported total value around $800M). Overnight, AppLovin stopped being a games company and focused on software.

• Sept 22, 2025: Added to the S&P 500.

• April 2, 2026: Adam Foroughi stepped down as Chairperson (remaining CEO), separating the chair and CEO roles.

Selling the games was a difficult decision at the time, but ultimately the right call. The studios were profitable and produced useful data, but they were lower-margin and capital-intensive. More importantly, they put AppLovin in direct competition with its own advertising customers.

In June 2026, AppLovin fully opened its AI-powered AppLovin Ads platform to the public, removing prior referral restrictions to scale access for e-commerce and non-gaming performance advertisers globally. This is a pivotal moment for the business, and it remains to be seen how this will impact the business from a financial standpoint.

3. Business Model and Products

AppLovin has gone through several iterations of its business model, as we have discussed. Today, the concept of their business model is pretty simple.

It runs an AI engine that matches advertisers to users, and takes a cut of the ad spend that it directs. This is essentially the oldest business model in advertising and what makes it possible, is AXON.

AXON is a real-time machine-learning system built around reinforcement learning. For every available ad impression, it predicts the probability that a given user will take the action the advertiser cares about (install, purchase, subscribe) and bids accordingly.

AXON’s scale is the key here:

• Processes on the order of 2 million ad auctions per second

• Draws behavioural signal from over 1 billion devices

• Trained on trillions of daily in-app events from AppLovin’s former portfolio of games

Crucially, AXON does not primarily rely on the demographic or social-graph targeting that defined the old ad world. Instead, it is a behavioural prediction engine that learns, impression by impression, what kind of user-context-creative combination produces a conversion, and it updates continuously.

Every install, purchase or even a “non-click/tap” is a training signal that sharpens the next prediction. This is the same closed-loop dynamic that made Meta and Google's ad machines unbeatable, but AppLovin built a version that operates across the open mobile ecosystem.

The commercial output of all this is ROAS.

Ultimately, what advertisers truly care about is that a marginal dollar spent through AppLovin returns more than the same dollar spent elsewhere. When AXON 2.0 launched, measured ROAS jumped enough that advertisers reallocated budgets, and because the model improves with more spend (more spend → more data → better predictions → higher ROAS → more spend), the advantage compounds on itself. That self-reinforcing loop is the crux to this economic engine.

AppLovin’s Full Stack System

Of course, AXON is not the only part of the puzzle. AppLovin’s acquisitions over the years have built a full stack pipe that few competitors can match:

• AppDiscovery: This is the demand side where advertisers buy user acquisition.

• MAX: The supply and mediation side, an in-app bidding platform that auctions publisher inventory.

• Adjust: The measurement layer that attributes outcomes back to ad spend.

• AXON: The decisioning brain that sits on top of the 3 above.

Owning demand, supply, measurement, and decisioning together means AppLovin sees more of the signal end-to-end and can optimise across the whole chain compared to other competitors who have to stitch together 3rd party tools.

The Advertiser Expansion

This is a huge part of the thesis for AppLovin. Historically, AppLovin's advertisers were almost entirely mobile-game developers buying installs. It is a large market, but relatively capped. The story moving forward is AXON catering its services to everyone else, outside of just the gaming market.

• E-commerce:

  • The target market here would be direct-to-consumer and online brands that intend to acquire customers. AppLovin ran this as a referral-only, white-glove pilot through 2024-2025.

  • Management has repeatedly said early e-commerce ROAS is comparable to Meta's Audience Network, which if true at scale, is massively bullish.

  • Self-serve:

    • At the end of June 2026, AppLovin launched its AXON Ads Manager self-serve platform plus a dedicated e-commerce platform to general availability.

    • Previously it was hand-held onboarding, which holds back the growth in usage. This is perhaps the most important near-term catalyst.

• Onboarding bottleneck:

  • Roughly 57% of qualified e-commerce leads that have been vetted and want to advertise on AppLovin today. This means that the other 43% is still hesitant and is lost revenue sitting on the table.

  • The friction is largely creative where e-commerce brands without big creative teams struggle to produce enough ad variations.

  • AppLovin is piloting GenAI creative tools with 100+ advertisers specifically to break this bottleneck.

• CTV and Streaming:

  • Through the 2022 acquisition of Wurl, AppLovin sells into connected-TV inventory, a growth vector management has flagged alongside the core. It is still small today but as it is not gaming-related inventory, it could be a great diversifying option.

• Lead generation & new verticals

  • A forthcoming lead-gen campaign type is designed to push AppLovin beyond mid-market e-commerce into high-value verticals like automotive, insurance and subscriptions, meaningfully broadening the addressable advertiser base beyond Shopify sellers.

• First-party inventory

  • AppLovin has signalled it is exploring its own social/owned-inventory surface. This is still early and speculative, but it would extend the model beyond renting other people’s inventory.

Monetisation Rate, Potential Hidden Optionality

AppLovin today monetises only about 1.3% of the impressions it serves. The other ~98.7% generate no revenue at all.

This is not a model quality issue, but rather a demand diversity problem. At the moment, the majority of ads AppLovin has to show are for other mobile games, which results in the engine quickly running out of relevant things to serve. Most users are perfectly happy with the game they are already playing and have no interest in installing another, so the vast majority of impressions get a repetitive, low-intent ad.

As the advertiser base broadens beyond gaming, AXON now has something relevant to show on inventory it already controls, a personalised product on almost every impression, while reserving scarce gaming ads for the high-intent moments where they already convert. Lifting the monetised share of impressions from 1.3% toward even 3% would roughly double the revenue capacity of the existing publisher base with no new supply added at all. Adam himself has mentioned previously that he believes 5% is a realistic target that they eventually hope to get to.

The e-commerce opportunity is therefore not a linear opportunity, but a massive, perhaps exponential opportunity for the business.

Huge Margins

AppLovin’s incredible margins are because it is essentially a software business, not a media business. AppLovin doesn't own most of the inventory and doesn't employ armies of salespeople.

AXON is the sales pitch and each incremental dollar of ad spend it processes costs almost nothing to serve. The easiest way to explain AppLovin, is that it is a tollbooth on performance advertising, where an AI sets the toll, and the road is being widened from just mobile games to all of digital commerce.

4. Moats & Differentiation

This is the most important section in the piece because everything else, from margins to growth and valuation, is downstream of whether the business moat is durable.

1. Data and Feedback flywheel

This is probably the most important part of AppLovin. As more ad spend flows through AXON, more conversion outcomes are observed, enabling the model to get more accurate. ROAS then improves and advertisers route even more spend, starting the next cycle.

It is self-reinforcing and, crucially, has increasing returns to scale. The marginal data point is most valuable to the player who already has the most data because they can place it in the richest context.

AXON serves over $10 billion in annual media spend and learns from trillions of in-app events across 1B+ devices at 2M auctions/second. This is structurally the same strength that makes Meta and Google's ad machines impregnable.

What makes a moat important or durable of course, is that a competitor must not be able to buy their way in.

In this case, I believe this holds true. Data scale economies have a cold start problem where a new entrant has no data, which means a poor model with weak ROAS. This means they can’t attract the spend that would generate the data. Essentially, a chicken and egg problem.

2. Control of Supply via MAX

In my view, this is the most defensible moat that AppLovin owns. AppLovin’s in-app bidding and mediation platform, MAX, controls an estimated 60-80% of the mobile ad-mediation market.

Firstly, what is mediation? It refers to the layer that decides which ad network gets to fill each impression in an app. Owning the dominant mediation platform means AppLovin sits at the chokepoint between the world's app publishers and every ad network bidding for their inventory, including its competitors.

This provides AppLovin with two advantages, (1) privileged, low-cost access to an enormous, exclusive stream of supply and bid-level data that trains AXON, (2) a structural toll position over rivals who must bid through AppLovin’s channel to reach that inventory.

Why is this sticky? Because MAX is integrated into the publisher's app at the SDK level (literally compiled into the app's code). Therefore, ripping it out and replacing it requires engineering work, re-integration, QA, and re-tuning of monetisation across every app in a portfolio. Publishers are unlikely to do so, especially when MAX is also delivering them top-tier fill rates and yield.

There is also a second, more subtle advantage that AppLovin possesses because of MAX. As MAX runs the auction, AppLovin observes every bid (who bid, how much, who won, who lost, on which inventory, at what price). A pure demand-side competitor, such as Meta's Audience Network, sees only its own bids and its own outcomes. That is a structural information asymmetry no bidder can replicate, and it means that when a rival like Meta bids more aggressively inside MAX, it is not purely zero-sum.

The auction pie grows, publishers earn more and recycle it into user acquisition on AppLovin. AppLovin also collects a ~5% fee on impressions won by outside demand bidding through MAX, and each competitive bid becomes training signal that makes AXON smarter.

3. Vertical integration that closes the loop

AppLovin owns every layer of the chain: AppDiscovery (demand), MAX (supply/mediation), Adjust (attribution/measurement), and AXON (the decisioning brain) sitting on top.

Each of these layers feed the others in a way competitors who own only one layer structurally cannot match.

Adjust's attribution data continuously sharpens MAX's bidding, MoPub's acquired traffic became training samples for AXON, MAX's supply gives AXON the inventory to act on.

Owning the whole pipe enables AppLovin to see end-to-end signals from impression → click → install → in-app purchase, with no data lost at the seams where most ad-tech stitches third-party tools together.

There is also a counter-positioning angle here. The Trade Desk built its brand on being a transparent, non-principal DSP that doesn't own inventory nor take hidden margin. That positioning makes it very hard for TTD to copy AppLovin's closed, principal, take-the-spread model without betraying its own value proposition and cannibalising its agency relationships.

Incumbents are often trapped by their existing model. AppLovin's full-stack, principal approach is something the open-internet players can't easily mirror. I believe the key lesson here that the industry is learning from AppLovin is that owning the infrastructure beats owning the inventory, and infrastructure of this completeness would take a rival many years and many billions to assemble.

5. Competitive Landscape

Read more

Circle stock $CRCL fell 17% on the day and now trades around $66, a market cap of roughly $15.7 billion. Circle is down 20% YTD and over 70% from all-time highs.

Two months ago, I published my Circle deep dive and said I would get interested below $80. We are now well below $80, but the fundamental picture has also changed. In this post, I will discuss what has changed and my plans moving forward.

What Open USD Actually Is

OUSD is run by an independent company led by Zach Abrams, co-founder of Bridge (the stablecoin firm Stripe acquired for $1.1 billion). The design has three defining features:

1. Partners can mint and redeem at zero cost with no volume caps.

2. Nearly all interest earned on reserves flows back to partners after a management fee.

3. Governance sits with an independent board of partner companies rather than any single firm.

It launches natively on Solana later in 2026. Coinbase has confirmed OUSD will also be launching on Base, and Stripe says it will make OUSD the default stablecoin for businesses on its platform.

How This Affects Circle

I think this has several wide-ranging impacts on Circle, if successful. We will debate the likelihood of success later, but let’s assume for argument sake that this is successful.

1. Coinbase re-negotiation

Coinbase and Circle have a revenue-sharing agreement which expires next month, and Coinbase now negotiates as a founding member of the rival consortium.

Circle currently keeps roughly 40 cents of every gross revenue dollar. I find it hard to see this going higher post-August. I believe the renewal will happen, and odds are that terms are unchanged. However, there is now a likelihood that Circle gets worse terms, and the probability of Circle getting a better deal than the one they have now is most likely zero.

2. Neutrality moat

In my view, and as I discussed in the deep dive, I believe Circle’s strongest moat is market neutrality. It is seen as the most credible stablecoin, and one that competitors could build on without conflict.

The consortium structure replicates neutrality through shared governance. For instance, we have Visa and Mastercard, clear rivals, both co-owning one coin. If OUSD gains traction, it would be a legitimate contender.

3. Enterprise and payment flows

One huge advantage for OUSD is that it has all the tools to build massive volume. Stripe making OUSD the default for platform businesses attacks precisely the segment Circle identified as its next growth leg: B2B payments, merchant settlement, CPN corridors.

If OUSD becomes the settlement asset inside Stripe, Shopify and the card networks, we could see an explosion in volume. At the same time, Circle's payments expansion story could get boxed into crypto-native and emerging market channels.

4. The re-pricing of distribution

OUSD’s model is a direct counter-position to USDC. Circle’s model treats distribution as a cost to be negotiated bilaterally while OUSD's model makes distribution the shareholder where nearly all reserve yield goes to partners by design.

Every future Circle partnership negotiation would therefore happen against that benchmark. For instance, Binance, wallets, neobanks would compare what a consortium would pay them and use it as a benchmark. This damages Circle even if OUSD never launches, and is therefore the most immediate and permanent impact.

OUSD is not a done deal, nor a guarantee

This is not the first time that a consortium has come together to build shared money rails.

In June 2019, Facebook unveiled a stablecoin (Libra) backed by a Geneva-based association of 28 founding members including Visa, Mastercard, Stripe, PayPal and eBay. The pitch was nearly identical to OUSD's, with shared governance, no single owner, a neutral digital dollar for the internet. Within five months, PayPal, Visa, Mastercard, Stripe and eBay all walked out under regulatory pressure.

Three of the companies that abandoned Libra at the first sign of regulatory heat (Visa, Mastercard and Stripe) are now founding partners of Open USD.

The second precedent is a working project. The Global Dollar Network (USDG), launched in late 2024 by Paxos with Kraken, Robinhood, Galaxy and Anchorage is a functional OUSD blueprint. It is a consortium coin that shares nearly all reserve yield with distribution partners. It has everything OUSD promises, including yield sharing, real exchange distribution and a fully licensed issuer, which is more than OUSD currently has.

18 months in, USDG has gathered roughly $3 billion against USDC's $74 billion. Yield sharing alone evidently does not create demand.

We can also go further back beyond stablecoins. In 2012, the merchant consortium behind CurrentC, assembled Walmart, Target, Best Buy and dozens of other retailers to build a shared payments network and kill card interchange. The incentive to cooperate was obvious and enormous.

However, the product never got out of pilot, because every member had its own priorities and they didn’t deploy their best engineers to work on it. Apple Pay emerged and took the market. I believe the issue is with consortiums. They are excellent at announcing partnerships, signing memoranda, but poor at actually getting stuff done.

Apart from this, I believe there are also a few other reasons Open USD is unlikely to gain traction:

1. The incentives are questionable

OUSD's incentive architecture is funded by reserve yield, and reserve yield requires float. The partners incentivised to push OUSD (Stripe, Shopify, the card networks) are payments companies whose flows generate velocity, which produces volumes but almost no balances. There are of course banks too, such as BNY, DBS, Standard Chartered, U.S. Bank, BBVA. However, they are structurally disincentivised from moving them, because every dollar of client deposits converted into OUSD leaves the bank's balance sheet.

It therefore trades a 3%+ net interest margin franchise for a diluted share of ~4% reserve yield, and cannibalises the deposit base that banks ultimately care most about. I believe banks are in the consortium because they want the optionality. However, it will not be a priority.

2. No named issuer, custodian or reserve detail

As it stands, there are no disclosures on which regulated entity issues OUSD, in which jurisdictions, or under whose supervision. There are tons of requirements that will be required to launch a stablecoin and these are questions that remain unanswered today.

The launch of OUSD is the easy part, what comes next is probably the toughest.

3. Membership is a free option, not a commitment

The most impressive part, where there are 140+ partners for OUSD, I believe also reveals the weakest point. There are no capital commitments, exclusivity or minimum volumes disclosed.

For Visa and Mastercard, membership is cheap insurance against disintermediation. Coinbase earns billions from USDC economics and joined anyway, proving membership is compatible with betting on the incumbent. I think this dis-incentivises partners from wanting this to succeed.

4. Committee products lose to owned products

As we discussed earlier, we’ve seen many cases of ideas like this that have surfaced and disappeared. CurrentC had Walmart, Target etc, but never gained traction. OUSD's competition is Circle, which shipped CCTP, CPN, Gateway and Arc in under three years, and Tether, which is one of the most impressive businesses in the world, garnering the most dollars per employee of any company in the world.

5. Float is the largest unsolved issue

USDC's ~$74 billion and Tether's ~$190 billion sit in exchange trading collateral, DeFi positions, emerging market dollar savings and crypto-native treasuries, and OUSD has no distribution edge in any of them. The marginal holder in Buenos Aires or Lagos holds USDT because everyone around them does. Offshore trading collateral is Tether's fortress, where no consortium member operates. It will be tough to disrupt the incumbents.

Where does this leave Circle?

Near term, I believe Circle should be fine. USDC adoption will continue growing steadily. I think OUSD is a 2027/2028 and beyond issue where the specific worry is with terminal value, and specifically a terminal margin problem.

Even in scenarios where OUSD underwhelms, the multiple the market assigns to a "winner-take-most" reserve income stream has to come down when the winners' biggest customers have publicly organised an alternative.

While CRCL now trades at $66 which was an attractive entry point based on my deep dive, the facts have indeed changed. All the assumptions used in the valuation model therefore also have to change.

Personally, I am watching 3 things:

1. The terms of the Coinbase renewal in August.

2. Whether/when OUSD names its issuing entity and licensing structure.

3. RLDC margin trajectory in the quarters ahead.

I hope this was helpful!

-Gab