.png)
Zorea Capital Equity 3Q 2026 letter
Fellow investors,
Our performance for the third quarter of 2026 was 1.4% (gross) and 1.0% (net). This brings 2026 YTD performance to -4.4% (gross) and -5.7% (net). Net performance is calculated using our standard fee schedule ‘M’ which charges the highest management fee from our different alternatives but no performance fee. As long-term investors, we emphasize that evaluating our results over a short period of time is not informative.
Markets continued to be driven by macro themes during the third quarter, with share price movements heavily influenced by the same narratives present in the first half of the year. The portfolio’s look-through metrics remained strong. Over time, our overall performance should converge with the performance of our companies. We expect our companies’ underlying value to continue growing at attractive rates and look forward to sharing our look-through portfolio metrics in next quarter’s letter.
During the quarter, we reduced our investments in HCA Healthcare (“HCA”) and T-Mobile (“TMUS”). We will expand on the decision to begin reducing HCA later in this letter. The further reduction in TMUS reflects our continued loss of confidence in management. We reallocated most of the proceeds to existing holdings whose shares had declined amid concerns around AI. In these cases, lower prices coincided with greater clarity around how AI is likely to affect the companies, as well as continued strong execution, which strengthened our confidence in their long-term prospects. While we have underwritten each business individually, and each of their relationships with AI varies, their shares tend to move with changes in the AI narrative du jour. So, we will refer to them here as the “AI scare basket.”
The “AI scare basket” currently represents 23% of our portfolio. We believe the risk-reward of this basket is quite attractive and that the odds are in our favor. From a risk perspective, we think our downside is limited at current prices given the nature of these businesses, the value they add to customers, and their share prices. On the flip side, if we are right, the shares could be worth multiples of today’s prices over a multi-year period. The companies in this basket are Tencent Holdings, Topicus.com, Constellation Software, and MercadoLibre. Together, these four holdings detracted 4.5 percentage points from our year-to-date gross return, with Tencent and Topicus accounting for all of the negative contribution. Yet they are all experiencing strong operational and financial momentum.
Our willingness to invest amid AI uncertainty does not extend to every business. As discussed in our 4Q 2025 letter, we believe AI creates direct substitution risk for important use cases served by well established companies. In that letter, we identified Adobe and FactSet as two examples. We have stayed away because we are generally uninterested in companies whose products or services are losing utility in the eyes of the customer. The distinction is whether we can identify durable sources of customer utility at a price that adequately compensates us for the uncertainty.
We do not think we are being particularly clever in recognizing the potential upside in the companies in our “AI scare basket.” Their shares have largely declined because of uncertainty about how AI will affect them. The market does not like uncertainty. From our perspective, even though we think these companies are likely to experience a neutral-to-positive impact from AI, we also recognize that such an outcome is not certain. We own these companies despite the uncertainty because we believe our downside is limited if we are wrong.
Tencent Holdings is our largest investment in this group, so we will use it as an example to illustrate our thinking[1]. We believe consumer agents are AI’s most relevant application for Tencent. WeChat is the superapp in China with consumers increasingly using it as their operating system. Tencent would greatly benefit from the successful implementation of a WeChat native Muse-like agent. The company is very focused on it and is in a prime position to make it happen. However, our investment case does not depend on Tencent developing the preferred agent. Even if another company succeeds in developing the consumer agent of choice, we don’t think core social engagement in WeChat will decrease (do people use WhatsApp, Instagram, or Facebook less because of AI?). Furthermore, even if core social engagement in WeChat is impacted, there is more to Tencent than WeChat. Tencent owns Weixin Pay, several gaming platforms, a large cloud business, and an enviable VC portfolio. These businesses are doing well, and AI is either less relevant or positive for them. Tencent currently trades at an 11x P/E[2]. At the current share price, we believe the non-WeChat businesses are worth enough to substantially limit our downside.
Although we assess the companies in our “AI scare basket” individually, they share exposure to how AI may change customer behavior and business economics. We therefore consider their combined weight when sizing the portfolio. At 23%, we believe the exposure balances the attractiveness of the individual investments with the possibility that our assessment of a shared risk is wrong. While we may add at the edges, we are unlikely to increase that exposure materially without stronger evidence supporting our views. We would, however, allow these investments to grow through appreciation.
Avoiding biases
As the great late Charlie Munger said “if you make a public disclosure of your conclusion, you’re pounding it into your own head.” Writing these letters exposes us to anchoring bias, escalation of commitment bias, and confirmation bias. Please note that no investment operation has had a perfect success rate. We will not be the first. Accordingly, we reserve the right to change our minds and hope to do so quickly and swiftly when appropriate. We take on no obligation to communicate changes of opinion.
Reducing HCA
HCA began 2026 as our largest investment, representing 12.4% of the portfolio. It has been a core holding since our strategy’s inception and became our largest through appreciation. During the third quarter, we reduced the position after concluding that a known policy headwind would have a larger impact than we initially expected. Our assessment of the appropriate position size changed, even though we continue to find the investment attractive.
In the United States, the ACA exchanges had been an increasingly popular way for people to get access to health insurance. In 2025, total marketplace enrollments reached 24.3mm, roughly doubling since 2021. Part of that popularity stemmed from the enhanced ACA subsidies available. These provided substantial assistance to enrollees within certain income brackets to help cover health insurance premiums. These enhanced subsidies were set to expire at the end of 2025, and the Republican controlled Congress decided against extending them. This meant that some people enrolled in the ACA exchanges were likely to churn out of the exchanges and become uninsured[3]. Coming into 2026, we analyzed this headwind and tried to quantify it. Management, which is generally conservative in providing guidance, also gave its estimate. We determined that the outcome was likely to be a manageable one-time headwind for the company. As a result, we decided to hold our investment.
Subsequent datapoints up to June seemed to support our conclusion. However, in July data began emerging that made it clear that the number of people leaving the exchanges and becoming uninsured would be much higher than previously expected. The headwind would therefore have a larger impact on HCA than we had expected. Looking ahead, proposed premium increases and indications of a less healthy exchange risk pool suggest that coverage pressures could persist into 2027. Insurers have proposed a median premium increase of approximately 15%, which increases the likelihood of continued exchange churn in 2027.
These developments changed our expectations and those of the market about the magnitude and duration of the headwind, prompting the company’s shares to sell off. This was clearly not the outcome we expected, but negative surprises are inevitable in investing. What matters is the impact that they have on our businesses and how we react to them.
The deterioration changed our earnings expectations, but the company’s competitive position – its dense network of must-have assets in growth markets – remains intact. It also seemed to us that the current backdrop was as negative as it would get and that there was positive optionality on how conditions evolved. In addition, with HCA being one of the strongest provider systems in the US, financial pressure on weaker competitors could create opportunities for HCA to gain market share and make attractively priced acquisitions, though that ride would be bumpy. We have always seen negative policy developments as a risk for this company and have thought through the potential consequences appropriately.
We therefore concluded that HCA’s near-term fundamentals had deteriorated but that it remained an attractive investment. However, it became obvious to us that the investment was too large given the circumstances. As a result, we decided to begin reducing it opportunistically.
Since mid-July, it has become increasingly likely that Democrats will win control of both chambers of Congress in the midterm elections (though that outcome is far from certain). The current government’s action in letting the enhanced ACA subsidies expire has been one of several decisions making affordability an important issue for voters. Democrats have jumped on the opening and have made enhanced ACA subsidies a key topic in the current election cycle. A Democrat sweep would not guarantee changes to healthcare policy, as Trump retains veto power, but it improves the chances that something will happen. This is the optionality we referred to above.
We believe these developments have made investors more constructive about the prospects for hospitals, with HCA shares rallying off their lows to trade close to flat for the year (as of this writing). Following our plan, during the quarter we reduced our investment by about 20% at around $430 per share and may continue reducing the position opportunistically. HCA is still a core investment, but smaller. Despite the negative fundamental developments of the last year, our investment in HCA continues to be a material positive contributor to our performance. Our HCA investment has generated a gross IRR above 20% since inception. But we are still shareholders in the company, so it is too early to declare victory.
We believe this example provides a good window into our decision-making and portfolio management. We initially invested in HCA because of its dense network of provider assets located in attractive geographies, anchored by hospitals that insurers need in their networks. We made the investment despite its regulatory risks. Not because we like the government in our businesses, but because we thought the impact of these risks would be manageable if they materialized. Once we invest, we monitor our investments closely. When we see disconfirming evidence, we recalibrate our expectations. If fundamentals deteriorate, we either reduce or sell our investment.
Updated thoughts on the AI spending supercycle
In previous letters, we have written about some of the bubble-like behavior we are seeing in the markets. We have updated some of our views on AI and believe we should update our readers on what has changed. While there are clear signs of froth in the markets, we believe we must be more nuanced in how we assess these developments.
Since our last quarterly update, new data has emerged showing that paid adoption of AI tools (as represented by OpenAI’s and Anthropic’s ARR[4]) is occurring significantly faster and at a much larger scale than anything the world has seen in the past. Customers are opening their wallets and paying OpenAI and Anthropic at a magnificent scale and at prices that seem to be attractive for the labs. By July/August, OpenAI and Anthropic reportedly had a combined annualized revenue run rate of approximately $105bn, compared with roughly $29bn at the end of 2025. That is approximately 3.6x the year-end figure in just a few months and represents an increase of about $76bn in their annualized revenue pace. Anthropic was the standout, growing from approximately $9bn to $65bn[5]. Truly remarkable.
As far as we can tell, this combination of growth rate and revenue scale is unprecedented. Today’s AI companies benefit from established digital distribution and cloud infrastructure, allowing adoption to spread much faster than in earlier technology waves. That advantage helps explain some of the growth, but it does not change the fact that customers are eager to part with their dollars and give them to the labs.
Based on the companies’ alleged gross margins and general commentary, current prices for intelligence justify paying artificially elevated prices for access to data center capacity. The AI labs are paying high prices to access compute capacity because of supply constraints. This should normalize at some point, but what seems evident is that OpenAI and Anthropic make money at current infrastructure prices, implying that investing in data centers at premium prices has been a reasonable use of capital. The lease agreements between the hyperscalers and the AI labs range from five to ten years, allowing the hyperscalers to (at a minimum) recover their capital as long as the AI labs don’t get into financial trouble. At some point OpenAI’s and Anthropic’s revenue growth must normalize, but in the meantime, we see no reason for AI infrastructure spending to meaningfully decelerate. This, of course, creates a high degree of dependence on OpenAI and Anthropic to keep the party going. So far, they are delivering.
Given the facts, we think it is incorrect to assert that current spending on data centers is a bubble. We nevertheless believe that plenty of individual companies are overearning because of supply/demand imbalances. The current buildout is bigger and faster than any other we have seen before (see below). It is extraordinary to come to the view that there is no bubble when all other previous buildouts of this type conducted by private capital led to a bubble. Interesting times.
Joining the AI party?
While we now recognize that the current spending on core AI infrastructure is reasonable, we continue to believe that investing in AI carries significant risk.
For usage-based AI services, revenue depends on the volume customers purchase and the price charged for that usage. We expect demand for useful AI capabilities to keep growing. What is harder to assess is how that demand translates into billable usage, computing requirements, and profits. Efficiency improvements can reduce the resources needed for a given task while lower prices make many more tasks economical. The balance between those forces will matter enormously for businesses supplying the technology, and having a high confidence opinion on that balance is difficult.
We doubt that growth in token consumption will track growth in intelligence demand. Significant (human) brainpower is being applied to the token efficiency problem, with interesting early results. One example is TypeSafe AI’s recently launched Jev, which makes structured decisions rather than generating text. Specialized models of this kind can take over tasks that would otherwise be handled by more expensive general-purpose models. Jev is noteworthy because for certain code generation use cases, it can reduce token consumption by 70% to 95% with no degradation in the quality of the output. While Jev’s scope is still relatively limited, it seems reasonable to assume that many more of these tools will be developed. Depending on their level of success, token consumption could grow much more slowly than end-user demand for intelligence.
Turning to token prices, open-source models are about six months behind frontier models, causing pricing pressure. We have no doubt that the cost per token at a given quality of inference will continue to fall rapidly. This is the consensus view. The chart below from Epoch AI shows a rapid decline in the price of achieving a given level of performance on selected benchmarks. Decreased costs lead to more usage, but if AGI-level intelligence starts to become commoditized, a race to the bottom on price across most use cases can pose a material risk to revenue growth and margins for the AI labs. We already find ourselves using the top models much less and the next-best models more. That trend seems likely to continue.
Customer concentration adds another source of uncertainty. According to Ramp, the top 1% of business customers accounted for approximately 80% of enterprise revenue at OpenAI and Anthropic. Some large customers serve many underlying users, but concentrated purchasing still creates exposure to customer bargaining power, changes in model selection, and abrupt shifts in spending.
For most AI related investments in the public markets, we need a defensible view of future demand, pricing, and the durability of the current unit economics at OpenAI and Anthropic. We find that difficult today. The paid adoption is impressive, but the range of outcomes remains particularly wide. For now, these investments remain too hard for us.
Alignment
A cornerstone of our portfolio management philosophy is to manage client portfolios in the same way we invest our own capital. At Zorea, we cannot guarantee results, but we can guarantee that we are in the same boat as our investors.
Thank you for your trust.
Yours truly,
Simon Bennaim
Footnotes
[1] Our initial Tencent investment dates back to 2024.
[2] Calculated using the share price as of 9/30/26 and trailing earnings. We deduct the value of its private investments at their stated balance sheet value from market capitalization. This is our estimate of the valuation of Tencent’s operating businesses.
[3] Uninsured patients have significantly worse payment behavior.
[4] Annualized Recurring Revenue.
[5] These figures are estimates for private companies.
Disclaimer and disclosures
The information in this letter was prepared by Zorea Capital LP (“Zorea”). It has been obtained from public sources believed to be reliable. Zorea makes no representation as to the accuracy or completeness of such information. Opinions, estimates, and projections in this letter constitute the current judgment of Zorea and are subject to change without notice.
Any investment in any strategy, including the strategy described herein, involves a high degree of risk. The description of the approach of Zorea Capital LP (“Zorea”) and the targeted characteristics of our strategies and investments is based on current expectations and opinions and should not be considered definitive or a guarantee that the approaches, strategies, and your investment portfolio will, in fact, possess these characteristics. In addition, the description of our risk management strategies is based on current expectations and should not be considered definitive or a guarantee that such strategies will reduce all risk. These descriptions are based on information available as of the date of preparation of this letter, and the description may change over time. Past performance of any strategy we employ is not necessarily indicative of future results. There is the possibility of loss, including loss of principal.
Any projections, forecasts, or estimates contained in this letter are necessarily speculative in nature and are based upon certain assumptions. It can be expected that some or all of such assumptions will not materialize or will vary significantly from actual results. Accordingly, any projections are only estimates and actual results will differ and may vary substantially from the projections or estimates shown. This letter is not intended as a recommendation to purchase or sell any commodity or security.
Performance information in this document reflects the actual performance of the account established by Zorea’s Chief Investment Officer as of May 1, 2024. Reported net performance is net of all actual trading and other account expenses, reinvestment of all income, as well as Zorea’s fees. Zorea’s fees, as presented here, are composed of our standard fee schedule for non-Qualified Clients, consisting of a 1.8% management fee. Our Qualified Clients may elect from other fee schedules we offer. Qualified Clients who elect a different fee schedule may pay higher (or lower) fees and therefore realize lower (or higher) net returns depending on the portfolio’s performance. The specific fee charged to a client will be identified in the client’s advisory agreement.
Because this account was established prior to Zorea becoming a registered investment advisor, this means the performance results are ‘hypothetical’. Different types of investments involve varying degrees of risk and there can be no assurance that any specific investment will either be suitable or profitable for a client’s investment portfolio.
Index information is included for illustrative purposes only, as it is not possible to directly invest in an index. Indices are unmanaged, hypothetical vehicles that serve as market indicators. Index performance does not include the deduction of fees or transaction costs which otherwise reduce performance of an actual portfolio.
Broader market events will generally have some corresponding impact on our results and the client portfolios managed in accordance with our strategy. For example, if US equity markets rise overall, that will frequently help the performance of portfolios with exposure to US equities, while declines in the overall US equity markets will frequently hurt the performance of portfolios with exposure to US equities. Similarly, increases or decreases in interest rates will have an inverse relationship on bond market prices (higher interest rates generally result in lower bond prices, and vice versa) and also some corresponding impact on the returns of fixed income investments. No investment approach can guarantee a positive return or prevent loss.
Performance results shown are not a guarantee of future results and are not a guarantee or prediction of how any client portfolio will perform.
The information contained in this letter is provided for informational purposes only, is not complete, and does not contain certain material information about our strategy, including important disclosures relating to the risks, fees, and expenses. The information in this letter does not take into account the particular investment objective or financial or other circumstances of any individual investor.
This letter is strictly confidential and may not be reproduced or redistributed in whole or in part nor may its contents be disclosed to any other person without the express consent of Zorea and/or its managing partner.
Zorea Capital LP is a registered investment adviser domiciled in the state of New Jersey. We may not transact business in states where we are not appropriately registered, excluded, or exempt from registration. Individual responses to persons that involve either the effecting of transactions in securities or the rendering of personalized investment advice for compensation, will not be made without registration or exemption.
Portfolio Metrics – notes on methodology:
For all companies we use latest reportable information as of the date of the quarter. The numbers are our best estimates and what we use internally. However, coming up with the different numbers requires assumptions and some subjectivity. We try to be consistent with our methodology through time but there is no guarantee of accuracy. We use source documents for our calculations. The growth calculations under this table are a weighted average of the underlying growth of our investments. Some important callouts below.
Revenue per share growth (LTM): only includes our single names (indexes are excluded). For SYF and AER, we substitute Revenue growth for Book value growth as we see that metric as more relevant to underlying business growth.
# of Positions: excludes T-Bills.
ROE, PE, and EPS growth: all three of these metrics use ‘earnings’ either in the numerator or the denominator. For the companies that we see GAAP or IFRS Net Income as a good proxy of earnings power, we use that. For the companies where it isn’t, we use the metric we deem most appropriate, which is usually Free Cash Flow or company adjusted earnings, but can be something else. We believe we are conservative when publishing these numbers, but we cannot guarantee that such is the case. Both ROE and EPS growth calculations exclude our exposure to indexes and cash. HCA is excluded from our ROE calculation because the company has negative equity. TBBB and AKER BP are excluded from these three metrics as earnings are not meaningful.
PE (LTM, ex-cash & index): This metric represents the calculation of the PE of our individual company investments. It excludes cash as well as any index exposure we may have.
