Baron Global Durable Advantage ETF | Q2 2026

Baron Global Durable Advantage ETF™ (BCGD or the Fund) gained 12.7% (NAV) during the second quarter, compared to the 14.9% gain for the MSCI ACWI Index (the Index), the Fund’s benchmark.

Year to date, the Fund gained 4.4%, compared to the 11.3% gain for the Index.

There is a two-letter word that can best describe what happened in the second quarter – AI!

While the war in Iran weighed on stocks earlier in the year, with the Index declining 3.2% during the first quarter, the impact wasn’t significant. We have long argued that politicians, wars, and geopolitics in general have little lasting effects on businesses (barring long-tail scenarios of course, in which case writing or worrying about BCGD’s performance would likely be far from our first concern). 

What really drove stocks in the second quarter was the AI buildout, which produced a highly concentrated result with Information Technology (IT) as the only sector that outperformed the broader Index, and by a wide margin, up 39.1%, accounting for two-thirds of the Index’s return (and 70% in the first half). Even within IT, it was all about AI buildout beneficiaries. The semiconductor, semiconductor materials & equipment, and electronic components sub-industries, which accounted for only 16.1% of the Index’s average weight in the quarter and 14.5% in the first half, generated 50% of the Index’s quarterly return and 68% of its return in the first half. While Industrials and Financials were also up double digits in the quarter, up 12.6% and 11.2%, respectively, all other sectors lagged far behind. The stocks that performed best within Industrials were similarly driven by the AI buildout.

Against that backdrop, the Fund’s 12.7% return in Q2, which trailed the Index by 228bps, was not that surprising given our focus on the highest quality, durable compounders. As a result, we did not own many of the biggest winners – the AI bottleneck names that doubled or tripled in the quarter. Many of these businesses have historically been lower quality businesses with limited pricing power, low differentiation, low returns on invested capital, low margins and high cyclicality – and have now risen to prominence, thanks to the under-supply situation in the market. While in the short term their earnings can grow faster than their higher quality counterparts, the durability of their growth will be determined by the longer-term supply-demand dynamics, the barriers to entry in the industry and the ease with which incremental supply can be brought online (by incumbents as well as new entrants), as well as the durability of demand in the particular segment, and the availability of substitute products. Because we did not own the AI bottleneck stocks, IT was the Fund’s second worst sector in the quarter, despite the Fund's IT investments being up 28.1%, and was responsible for 301bps of relative underperformance.

In addition to IT, underperformance was also driven by Financials, which was a 312bps headwind on relative results. Within Financials, our entire underperformance can be attributed to being in the wrong sub-industries. Financial exchanges & data – the worst sub-industry within Financials, down 8.5% in the Index – was our biggest overweight, responsible by itself for 204bps of underperformance. We also held no domestic commercial or investment banks, which did well in the quarter on the back of strong U.S. capital markets and rate expectations that have shifted upwards. On the positive side of the ledger, we held the right Health Care, Consumer Staples, Real Estate, Communication Services, and Industrials stocks – which contributed 246bps to our relative results. Not holding any Energy, Materials or Utilities also helped, contributing 230bps.

From an absolute return and stock-specific perspective, Taiwan Semiconductor (TSMC) was our largest contributor, adding 272bps. ASML, Alphabet, Amazon, NVIDIA, and Eli Lilly contributed at least 75bps each, while Visa, Ajinomoto, Monolithic Power Systems, Keyence, and Welltower contributed at least 40bps each. Overall, 22 of our holdings were up double digits in the quarter. On the negative side, 11 names detracted from performance, though only CME Group detracted more than 50bps. Not only are we confident we suffered no permanent loss of capital, but we also believe the sell-off created an attractive opportunity to add to this quasi-monopoly business at an all-time low relative valuation multiple, even cheaper than during the great financial crisis. Historically trading at an average 47% premium to the S&P 500 Index, the stock dropped to a roughly 14% discount during the second quarter on emerging competition from crypto-native “perpetual futures.” We view those products as more of a retail-focused solution and hence unlikely to weaken CME Group’s competitive moat, which is rooted in the depth of liquidity on its platform. Moreover, as the stock sold off, fundamentals continued to improve, with 2026 revenue expectations increasing by 0.5% in the quarter and 3.2% in the first half and EBIT expectations increasing by 0.5% in the quarter and 4.5% in the first half. CME Group was our largest addition during the quarter.

Another example of divergence between narrative and fundamentals was in rating agencies such as S&P Global, where investors were concerned that it would be disrupted by AI, even though fundamentals have been robust with no negative impact from AI. We believe that AI could actually prove to be a tailwind driving increased demand for debt issuance and ratings, opening incremental distribution channels for data offerings, and increasing operational efficiency. S&P Global’s Market Intelligence segment, the business investors view as most exposed to AI disruption, reported an acceleration in organic growth to 6.3% in the first quarter. Overall fundamentals also continue to move in the right direction with 2026 FactSet consensus expectations for revenue and EBIT up by 0.1% and 0.3% in the quarter and by 0.6% and 0.6% in the first half, respectively. Yet, S&P Global’s P/E multiple traded at its steepest discount to the S&P 500 Index (around 5%) since the Great Financial Crisis levels seen in 2008 to 2010, with the stock trading at an average relative valuation premium of 27% in the last 20 years.

We see a similar widening gap between narrative and fundamentals in Amazon, which was trading in the second quarter at an all-time low P/E multiple of about 25 times. The multiple declined 1% in the quarter and 13.6% in the first half as investors worried that the large investments in AI will not generate attractive returns. Fundamentals however, show a robust business that is becoming more valuable and with early positive impact from AI - in Q1, retail units sold were up 15% year-on-year, the fastest growth rate since COVID, margins improved as Amazon reduced costs to serve and grew its advertising revenues (up 22% year-on-year in Q1), and Amazon Web Services revenues increased by 28%, the fastest growth rate in 15 quarters, as AI is beginning to materially impact numbers – with AI annualized recurring revenues (ARR) of $15 billion (more than doubled year-on-year), Bedrock revenues (Amazon’s AI platform) up 170% sequentially, and its internal chip business up 40% sequentially and 3 times year-on-year to $20 billion ARR.

Even among the direct beneficiaries of the AI buildout, stocks are not immune to narratives – no need to look further than the company standing at the epicenter of AI – NVIDIA. While NVIDIA’s stock was up 14.7% in the second quarter, ending the first half up 7.2%, it lagged other AI buildout stocks by a wide margin, with the semiconductor sub-industry in the Index up 54.6% in the quarter and 51.9% in the first half, as a result of investor concerns over potential market share losses to ASICs (such as Google’s TPU and Amazon’s Trainium). NVIDIA’s size and liquidity make it an easy short for hedge funds. Also, technical limitations on long-only investors to buy stocks that represent large weights in benchmarks, and renewed investor concerns of an AI bubble (‘TokenMaxxing³’ and open weight models) weighed on the performance of the stock.

Would NVIDIA be better off as the sole provider of compute for AI? Probably, though competition drives innovation! Should NVIDA trade at a 15 to 16 times P/E on calendar year 2027 estimates, implying a terminal growth rate below nominal GDP starting this year? Assuming a 10% weighted average cost of capital, that multiple implies terminal growth of 3%, which seems highly unlikely to us! 

Fundamentals tell a different story. NVIDIA’s revenue growth accelerated for the third straight quarter to 85% year-over-year, adding a record $13.5 billion of revenues in the last quarter alone (equivalent to ServiceNow’s total revenues in 2025!), reporting that the latest generation, Blackwell is having the fastest ramp in NVIDIA’s history, gaining significant market share in networking (with revenues up 3 times year-on-year) and seeing unprecedented levels of orders (having booked approximately $0.5 trillion in 4.5 months), just as AI is starting to show real return on investment (ROI) with ARR of $175 billion⁴ thanks to the explosion in agentic AI.

Now, a narrative or a rumor is just that, until the outcome proves it right or wrong. The problem is that it is unclear whether the narrative is right or wrong until the future is revealed and even then, it is often difficult to know whether a bad outcome was due to a bad decision or just bad luck. It is on the investor to decide whether the probability of the narrative being correct is higher or lower than what is priced in. In other words, are you getting paid to play that hand or not?

However, with the proliferation of AI and information becoming much more readily available to investors today than in the past, why do narratives still cause wild swings in stocks, rather than investors correctly incorporating changes in the intrinsic value and stocks trading accordingly even in the short term? Why does the market remain inefficient?

In our view, the reasons are both structural and psychological. On the structural side – passive investing in market cap weighted indexes, quant strategies, and low-vol strategies exacerbate the momentum of stock prices on both sides. Passive strategies push stocks that have worked higher (and as they outperform, their weights in the benchmark rise, driving even more purchasing) and those that haven’t, lower. Financial leverage, quants, low-vol strategies and increased retail investor participation in the stock market add fuel to the fire, which creates further pull on both extremes. Window dressing causes investors to buy stocks that have worked and sell those that haven’t.

Psychology further exacerbates the situation – one can think of every investor as holding a bucket for every stock they own. As long as the bucket is not full, they will hold on to the stock, but once water overflows, they will no longer be able to bear the pain and will sell. Different investors have different sizes of buckets depending on their mindset, patience, incentives and experience. Data points that reduce conviction fill the bucket with water and data points that increase conviction remove water. The highest conviction ideas have the least water in their buckets and the lowest have the most.

As long as the prevailing narrative doesn’t conflict with the thesis, everything is OK. Once however, the narrative starts conflicting with the thesis, there are various psychological biases that spring into action as the bucket is filling up with water. The heavier it is, the more Cognitive Dissonance the investor feels such that at some point, the weight becomes too much to bear prompting the investor to sell to cure the dissonance – this even if they consciously believe the probabilistic or expected value is in their favor. Groupthink makes the water flow stronger if the stock is going against the investor, as they increasingly feel that the market must know something. Of course, a stock that works makes the investor think they must be right because everyone else ‘agrees.’ Confirmation Bias causes investors to actively search for evidence that supports their thesis, even though objectively, conclusions should follow the data. Sunk Cost and Loss Aversion play in the same direction as selling would crystalize a loss. This is like a lid that puts pressure and why the bucket overflows too late and when it does, it does so violently (no valuation multiple becomes too low to sell and cure the cognitive dissonance).

Periods of accelerated disruptive change, such as today with AI, increase the flow of water into investor buckets, creating big moves in stocks whose fundamentals have been (and in our view, continue to be) quite durable such as in the examples we discussed above. This tests investor conviction, creating an opportunity for patient investors that hold large enough buckets to counter cognitive dissonance.

Taiwan Semiconductor Manufacturing Company Limited is the world's largest contract chipmaker and the leading manufacturer of advanced logic semiconductors used in modern AI accelerators. Shares rose 41.6% during the quarter as the company continued reporting stellar financial results underpinned by AI demand with revenue growth of 35% year-on-year and EPS growth of 58%, with 66% gross margins and 58% operating margins. High-performance computing now represents the majority of TSMC's business. AI demand is consuming so much leading-edge capacity that smartphone and PC production is increasingly shifting to older technology nodes, reversing a dynamic that defined the foundry industry for much of the past decade. Management also raised its full-year outlook and increased capital spending to support demand that remains well above available supply. We retain long-term conviction in TSMC and view its leading-edge manufacturing monopoly, pricing power, and technology roadmap as durable advantages that support a long duration of growth.

ASML Holding N.V. is a Dutch company that builds lithography machines used to print circuit patterns onto semiconductor chips. Shares rose 50.9% during the quarter as accelerating AI infrastructure spending reinforced ASML's critical role in the semiconductor supply chain. Sitting at the center of advanced chip production, lithography is in high demand: leading-edge logic capacity remains sold out at ASML's largest customers, while the high-bandwidth memory buildout is driving a parallel wave of lithography equipment investment at memory manufacturers. Management raised its full-year outlook guiding revenue to €36 billion to €40 billion in response to these favorable demand trends with EUV capacity of at least 60 units in 2026 and 80 units in 2027, reinforcing our conviction in ASML as a core long-term holding. We maintain conviction and continue to view ASML as an indispensable enabler of the leading-edge semiconductor roadmap and one of the primary beneficiaries of the ongoing investment in AI.

Alphabet Inc., the parent company of Google, contributed to performance with the stock up 23.2% as the market increasingly recognized its unique position vis-à-vis AI. Alphabet is the industry's most vertically integrated AI player, with ownership across every layer of the stack, including custom TPU silicon, global cloud infrastructure, the Gemini foundation models, and distribution across 13 products with more than 1 billion users each. This full-stack ownership is translating into strong demand. Google Cloud revenue grew 63% year-over-year and backlog surged nearly 300% to roughly $460 billion, prompting management to raise 2026 capital expenditure guidance and signal a significant further increase in 2027. Search revenue grew 19% year-over-year as AI features drove record query volumes, demonstrating that generative AI is expanding, rather than eroding, the core franchise. Capital access has itself become a competitive moat, allowing Alphabet to fund supply aggressively and outbid peers for scarce compute. We maintain strong conviction and believe that Alphabet's vertical integration, talent density, multimodal proprietary data, and highly scaled consumer offerings are durable competitive advantages that will underpin a long duration of growth for the company ahead.

CME Group, Inc. operates the world’s largest and most diversified derivatives marketplace. Shares fell 25.0% due to a slowdown in trading activity, reflecting tough comparisons against the prior year and easing market volatility following the onset of the U.S.-Iran war in March. This cyclical softening was exacerbated by concerns over emerging competition from crypto-native “perpetual futures” and uncertainty triggered by the announcement that long-time CEO Terry Duffy will step down next year. We continue to own the stock because we believe that CME enjoys significant competitive advantages and should benefit from increasing adoption of exchange-traded derivatives and episodic volatility spikes.

Shares of Wix.com Ltd., a cloud-based platform for building and managing websites and online businesses, declined 41.6% during the second quarter. The stock came under pressure as growth in the company's core business decelerated, with bookings and new user cohorts softening relative to prior quarters. While the newer AI-driven offering (Base 44) continued to grow, it was not enough to offset slowing momentum in the established core business, raising questions about the durability of the overall growth algorithm. Our thesis on Wix was driven by the combination of its durable subscription revenue, expanding free cash flow, and disciplined capital returns, but the deceleration in the core business undermined a key pillar of our thesis. We have therefore exited the position during the quarter, redeploying capital toward opportunities with clearer growth trajectories.

Tencent Holdings Limited is China's largest internet company and operator of the WeChat ecosystem. Shares declined 11.7% despite a largely in-line first-quarter report, which included 17% year-on-year growth in adjusted operating profit excluding new AI initiatives. Investors instead focused on headline revenue growth, which was skewed by the timing of Lunar New Year, as well as a limited near-term catalyst path and a rising AI cost base, with chip depreciation expected to weigh on second-half profitability and cause full-year operating profit growth to trail revenue growth. Our conviction remains intact. The quarter provided initial evidence that Tencent's AI investments are gaining traction, with its new Hunyuan 3 AI model, becoming one of the leading open weight models on OpenRouter⁵ and its agentic solution, WorkBuddy becoming one of China's most-used agentic AI⁶ suites. The WeChat AI agent opportunity is difficult to replicate given its billion-plus daily users and embedded merchant ecosystem. Meanwhile, the core business comfortably funds the AI build, and management is accelerating share buybacks. At a low teens P/E multiple, we believe the durability of Tencent’s core business and its AI opportunity are not properly valued by the market. We remain invested.

The Fund is constructed on a bottom-up basis with the quality of ideas and level of conviction playing the most significant role in determining the size of each investment. Sector and country weights tend to be an outcome of the portfolio construction process and are not meant to indicate a positive or a negative view.

As of June 30, 2026, the top 10 holdings represented 50.2% of the Fund’s net assets, and the top 20 represented 74.4%. We exited the quarter with 40 investments, unchanged from the end of the first quarter.

IT and Financials represented 51.5% of the Fund, while Consumer Discretionary, Industrials, Communication Services, Health Care, and Consumer Staples represented another 45.7%, with the remaining 2.9% held in Real Estate (Welltower) and cash.

During the second quarter, we initiated three new positions: a leading software and semiconductor company, Broadcom, the Korean giant and one of the leading memory, semiconductor manufacturing, smartphone and display providers, Samsung Electronics, and an engine provider for power generation, INNIO.

We took advantage of inflows and market volatility to add to 36 existing positions. Our largest additions included the following names: CME Group, TSMC, Amazon, NVIDIA, Alphabet, Visa, ASML, S&P Global, Stevanato Group, and Brookfield.

We also exited three investments – Indutrade, HDFC Bank, and Wix – reallocating to names in which we saw a more attractive risk-reward profile.

We took advantage of the stock’s sell-off post earnings to initiate a new position in Broadcom Inc., which designs and supplies semiconductor and infrastructure software solutions that sit at the core of modern computing and networking. The company is a global leader in high-performance digital and mixed signal technologies spanning networking, connectivity, storage, and custom accelerators (ASICs). Through VMware, Broadcom also owns critical software layers used to virtualize and manage large-scale compute environments.

Broadcom has one of the most durable and formidable competitive moats in the AI infrastructure buildout. As AI labs and hyperscalers look to not only optimize their hardware for AI workloads to achieve peak performance but also optimize costs to extract the highest intelligence per dollar of capex spent, they partner with Broadcom – the best silicon design player, one that excels at both digital and analog circuit design. Customers also choose Broadcom due to its ability to innovate at an annual cadence, similar to NVIDIA. Speed of innovation matters because lagging would mean that a competitor using NVIDIA's product cadence will have access to more powerful and cheaper compute. Moreover, ASIC designs are growing increasingly complex and require a multitude of technologies to bring to life through full system optimization. A hyperscaler or an AI lab cannot afford to go with an inferior solution when its revenue and long-term success are tied to the quality and the quantity of the compute power it has available.

While there has been a lot of noise about customers considering vertically integrating for future silicon designs, this remains only a bear narrative at this stage. The increasing complexity of AI systems demands very tight integrations between the different components, which is done via extreme co-design of compute, memory, input/output on dies, and networking fabrics. Broadcom, being at the frontier in all of these technologies, has therefore a far more durable moat than the market appreciates, in our view. Hock Tan on recent earnings call (Q2’2026) addressed this issue with respect to key customer, Google: “Our relationship continues to be strategic and very substantial as we continue to deliver vastly superior technology and execution compared to other alternatives. This ability to provide differentiated value to Google ensures that our business will sustain and grow for the foreseeable future.”

We expect Google to continue accelerating its investments in TPUs, driven by growing internal and external customer demand. Broadcom has also recently extended its multi-year agreement with Google to 2031, validating its strong position within the Google silicon ecosystem and establishing it as a key player benefiting from Google’s investments in AI infrastructure. Hock Tan commented on the agreement in the last earnings call:

“... a very, very strong agreement and it basically reflects the strength of the partnership we have simply because of the products we do… and any intellectual property we deploy into this whole program… it's a commitment that is very substantial in dollars. Very, very substantial amount of dollars.”

Beyond Google, other customers such as Anthropic, OpenAI, and Meta each showed incrementally positive signs in their respective ASIC adoption journeys. Anthropic, following its recent extraordinary success, is planning for larger compute requirements, which should translate into a larger TPU compute base over the next few years. OpenAI, working with Broadcom, taped out its first inference silicon, Jalapeno, in a record nine months, targeting 10GW of ASIC-based AI infrastructure this decade. Meta continues to hold a bullish view on AI and the associated investments in AI infrastructure. Additionally, Apple signed a multi-year agreement with Broadcom spanning multiple products, not only conventional radio frequency components and next-generation wireless connectivity technologies, but also ASICs across multiple generations of Apple products. Hock Tan expects ASICs to match GPU units in volume by next year, and Broadcom, as the leader in this space, should continue to be the biggest beneficiary of this growth. We believe that Broadcom is uniquely positioned to capture the lion's share of the custom silicon market for years to come, with strong competitive moats, underpinning a long duration of growth.

We also initiated a new position in the Korean giant and one of the leading memory, semiconductor manufacturing, smartphone and display providers, Samsung Electronics Co., Ltd. Our thesis has 4 key aspects:

1. The demand s-curve is large, and we are very early on it due to AI and specifically agentic AI.
2. Supply growth is limited.
3. The memory industry is becoming less cyclical.
4. We’re getting the non-memory parts of Samsung for free including its foundry business which offers optionality and a partial hedge on the tail risk of a Taiwan invasion.

First on demand, the throughput of AI inference (the amount of tokens generated per second by the AI factory), is dependent on how fast the system can pull data and model weights from memory and run the calculations on it to infer the next token. The greater the bandwidth, the more tokens can be generated. The more tokens, the more revenue the AI factory can produce and the greater the ROI. Agentic AI, which generates a lot of internal tokens (due to reasoning) to generate 1 external (consumer-facing) token, requires even greater bandwidth. It also creates a long context window (according to Micron, context window has grown by 30 times year-on-year!⁷) which requires more memory and storage, and it calls tools (search, file use, application use) which require even more memory and storage – as users want their AI agents to remember previous interactions rather than start from scratch every time. This has created the strongest demand in history for memory. And this is before autonomous driving has really started scaling (with autonomous vehicles expected to have 20 times the memory of non-autonomous⁸) and before robotics.

Second, regarding supply of memory, we believe that memory will remain structurally supply constrained through the end of the decade at least. High bandwidth memory (HBM), the stacked DRAM that sits beside every AI accelerator, is gaining market share of total DRAM, and it’s more wafer intensive than regular DRAM. Additionally, agentic AI drives demand for NAND, and so NAND capacity can no longer be converted into DRAM. Then, it takes two to three years to build a new greenfield production facility, and tool supply is limited.

Third, the memory industry may be becoming less cyclical. Memory suppliers are increasingly signing long-term agreements that provide better demand visibility and share volume and pricing risk with customers. In addition, memory – especially HBM – is becoming less commoditized as rising AI complexity makes co-design with accelerators more important. Co-optimizing the full system should increase differentiation and customer stickiness. Finally, as memory complexity rises, technology transitions to the next node are producing fewer incremental bits, making a larger portion of supply growth discretionary and giving the industry more flexibility to align bit supply with demand.

Lastly, having fallen behind SK hynix through the HBM3E generation, Samsung regained ground recently when in February 2026 it became the first in the industry to begin commercial shipments of the sixth generation HBM4⁹ for Nvidia’s next platform.

Beyond memory, we view Samsung’s display, mobile, automotive, and foundry businesses as meaningful optionality. While the foundry business is currently loss-making, success with external customers could create meaningful profits and drive a re-rating. Samsung’s 4x P/E multiple offers an attractive risk-reward to own a secularly growing memory business, with potential hedging value against global dependence on TSMC, HBM4 recovery as a near-term catalyst, and foundry as a long-duration call option.

This quarter we purchased shares of INNIO N.V., a leading reciprocating engine OEM and service provider that owns the Jenbacher and Waukesha brands. These engines, long used for gas compression and power generation, are now seeing record demand for behind-the-meter (BTM) power at data centers, which increasingly bring their own power solutions rather than rely on the grid.

After two decades of flat U.S. electricity load, the AI buildout has accelerated growth while the aging grid infrastructure has created interconnection shortages, fueling BTM demand. We believe that INNIO is well positioned to benefit from this dynamic given its engines' reliability, efficiency, optimal sizing, and fast-response load-following capability, which is critical for AI's volatile power demands, and provides a lower total cost of ownership than alternatives. With two to three year lead times versus four to five years for large combined-cycle turbines, INNIO logged a record 2 GW of orders in Q1 2026 and 8 GW of backlog exiting the quarter. Its advantages are difficult to replicate, requiring years to develop a competing engine platform and overcome INNIO's decades-long reliability moat.

We also like INNIO's razor-razorblade model: solid margins on original equipment, followed by more than 2 times that EBITDA from service revenue over equipment lifespans exceeding 20 years. Industry-wide price increases of 10% to 20% annually should further support margin expansion. INNIO plans to expand capacity from 3 GW today to 10 GW by 2030, primarily through capital-light brownfield expansions at existing facilities – supporting strong growth with minimal execution risk. We believe rising equipment demand combined with a growing installed base's long service tail will support meaningfully higher stock prices over time.

“It’s tough to make predictions, especially about the future.”
Yogi Berra

Short term market performance is driven by a whole host of external variables that are impossible to consistently predict - Geopolitics, macroeconomics, wars, Federal Reserve decisions and Truth Social posts. Even on a micro stock-specific level, short term results are driven by rumors, the narrative du-jour, investor positioning and expectations compared to the narrative and, as we discussed in the first section of the letter, good luck also trying to predict the bucket sizes of various investors!

Rumors and noise are constant – copper will be disrupted by optics! Amphenol’s stock falls 20% in the first 3 weeks of May. Actually, CPO (co-packaged optics) is delayed. The stock bounces by 48% in the last 10 days of May and June! NVIDIA’s next generation architecture is delayed, data centers are delayed, tokenmaxxing¹⁰, Meta is building a cloud business. AI stocks sell off. Then IBM pre-releases earnings disclosing customer budgets have shifted to server purchases related to AI – and AI stocks rise! Then new open-source models show further progress on the cost-intelligence curve and AI stocks sell off again (even though lower cost of intelligence has proven to be a net positive to the AI buildout) and so on and so forth… Our job is to separate the signal from the noise. Make sure we don’t waste time on the latter nor react to it. Stay the course. Focus on understanding the drivers of intrinsic values. Understanding whether the company’s competitive positioning is strong, whether it solves real problems for customers, and whether growth is durable. This is our bread and butter and what has worked for us in the past – this is what we continue doing day in and day out.

While we have no idea how stocks will trade over the short term, the direction of longer-term returns can be glimpsed from deconstructing the Fund’s performance into its two components – valuation multiples and fundamentals. If fundamentals are revised higher, intrinsic values are expanding. If valuation multiples rise by less than the rise in intrinsic values, our margin of safety has increased, and with it our prospective returns.

This was the case in the second quarter and the first half of 2026. Starting with fundamentals, weighted average revenue expectations for 2026¹¹ increased by 3.3% during the second quarter and by 8.1% in the first half. Operating income expectations were even stronger, having been revised up by 3.9% in Q2 and 10.4% in the first half¹². This means that the intrinsic values of our businesses continue to rise. As to valuation multiples, after contracting by 10.5% in the first quarter, the weighted-average multiple¹³ for the Fund expanded by 6.3% in the second quarter, though year to date, the multiple is still down 6.4%.

Note that while the multiple is down year to date, business fundamentals have been revised higher – which means that the Fund has become more valuable, and yet it can be bought at an even more attractive valuation. This implies that we are getting a collection of higher intrinsic values at a more attractive valuation than we did six months ago – this despite the Fund’s positive 4.4% return, year to date.

From a historical perspective, as of the end of the second quarter, the weighted average stock in the portfolio traded at a 9.3% discount to its 5-year average¹⁴ - which is another positive indicator for prospective returns.

Our goal is to invest in companies with strong and durable competitive advantages, proven track records of successful capital allocation, high returns on invested capital, and high free-cash-flow generation. It is our belief that investing in great businesses at attractive valuations will enable us to earn excess risk-adjusted returns for our shareholders over the long term. We are optimistic about the prospects of the companies in which we are invested and continue to search for new ideas and investment opportunities.

We thank you for your trust and for being our partners on this journey.

Sincerely,