Baron Global Opportunity Fund | Q3 2025

It was an okay quarter.

Baron Global Opportunity Fund® (the Fund) gained 7.7% (Institutional Shares), in line with the 7.6% gain for the MSCI ACWI Index (the Index), and the 9.0% gain for the MSCI ACWI Growth Index, the Fund’s benchmarks.

Year to date, the Fund is up 19.7% compared to gains of 18.4% and 19.1% for the benchmarks, respectively.

Three quarters of the year are in the books, and all the major global market indexes are sporting nice gains. Liberation Day tantrum, the market has continued to grind higher, driven by the anticipated interest rate cuts by the Federal Reserve (the Fed) and of course, the daily barrage of AI announcements and deals. On September 9, Oracle announced a massive backlog of AI bookings which soared to $455 billion (measured by Remaining Performance Obligations), an increase of 359% year-over-year. The news overshadowed the company’s miss of its quarterly revenues and earnings per share estimates, and its stock surged 36%, adding approximately $245 billion to its market capitalization. Less than two weeks later, on September 22, NVIDIA announced a $100 billion investment into OpenAI, which committed to purchase and deploy enough NVIDIA chips to power 10 gigawatts (GWs) of compute, worth approximately half a trillion dollars⁴. While we have been anticipating this accelerating pace of AI investments, we along with the rest of the market participants were caught a bit off guard by this massive scale. Add to this that company earnings have generally been better than expected, with more positive earnings revisions, and you have pretty good ingredients for the markets continuing to move higher.

There was not much insight to be gleaned from a quarterly performance attribution perspective, which is to be expected in an in-line quarter with 9bps of outperformance versus the Index. Our best sectors were Health Care and Information Technology (IT), which gained 25.4% and 12.4%, respectively. Within Health Care, results were driven by the strong performance in our biotechnology holding, argenx, which reported excellent financial results that beat investor expectations. Within IT, we saw broad strength, with Shopify continuing to post great financial results, while our semiconductor holdings, NVIDIA, Taiwan Semiconductor (TSMC), and ASML continued to benefit from everything AI. The strength in Health Care and IT was offset by weakness in Consumer Discretionary, mainly because of the pullback in MercadoLibre, due to weaker macro in Argentina, as well as the short-term impact to profitability from expanding free shipping in Brazil. From a geographical perspective, the Fund outperformed in developed markets, driven by Canada and the U.S., which was mostly offset by our investments in “other” markets (MercadoLibre and Globant in Argentina).

From an absolute return and stock specific perspective, we had 22 contributors against 14 detractors. Eternal (formerly known as Zomato), TSMC, ASML, Tesla, and GDS contributed over 50bps each to absolute returns, while Shopify, NVIDIA, SpaceX, and argenx generated over 100bps each. On the other side of the ledger, we had only two significant detractors: MercadoLibre, which cost us 108bps, and InPost, which detracted 54bps. We are confident that neither drawdown is likely to result in permanent losses of capital.

“History doesn’t repeat itself, but it often rhymes.” 
– Mark Twain

Is this the internet bubble all over again?

We don’t think so. One does not have to be a student of history to recognize some obvious similarities. On October 13, 1994, Netscape released its first browser – the Mosaic, which quickly captured the imagination of the tech community. It was the very first stock I had recommended to my first employer’s head of equity growth investing when the company filed to go public in August of 1995. “But they don’t make any money!” he said in the closest possible imitation to Mr. Wonderful. “Would you bet your bonus on it?” – “Yes, sir.” “Would you bet your career on it?” – “Yes, sir!” As a junior tech analyst, I wasn’t expecting much of either, and I was the only one on the desk who saw no connection between being a public company and having actual profits. The five million share offering was expected to price at $14 per share, but the underwriters had never seen this much demand, and so they priced it 100% above that, at $28. The stock opened at $71, peaked at $74.75 before closing the first day of trading at $58.25 and marking the beginning of the dot-com era. Many people started writing about how the internet will change the world and our lives, but few, if any, were able to articulate exactly how. In early 1996, Bill Clinton signed into the law the Telecom deregulation act, catalyzing a massive investment cycle into the infrastructure layer (fiber). But of course, there was no e-commerce yet, no social media, no streaming, no smart phones, no cloud computing, etc. – the application layer had not been built out yet. In December of 1996, Alan Greenspan first talked about “irrational exuberance” asking “how one could know when the stock market had become too excited and was potentially overvalued,” leading to speculation about the dot-com bubble that appeared to be growing at the time. Of course, grow it did, and it did not stop growing until the NASDAQ Composite Index reached a high of $5,049 in March of 2000 – 557% higher than it was on October 13, 1994 ($768) for a compounded annual return of 41.7%.

Almost exactly 28 years later, in October of 2022, OpenAI released to the public ChatGPT. It reached over one million registered users in the first five days and over 100 million registered users in the first two months.

Less than three years later… OpenAI is estimated to have $13 billion in annual recurring revenues and a valuation of $500 billion with an investor base highlighted by Microsoft, Softbank, and NVIDIA, making it the most valuable private company in the world. As soon as Oracle gave the AI trade a boost with its unexpected $455 billion order backlog, Alphabet said its own cloud-computing business has a $106 billion backlog, while Microsoft announced a $19.4 billion deal with Nebius to get more capacity. NVIDIA followed with its own order of up to half a trillion dollars, and Broadcom announced a $10 billion order from a new customer (most likely OpenAI). Add in an IPO frenzy (remember the talking dog sock puppet of Pets.com?) with companies that have yet to make a profit enjoying strong public debuts and the NASDAQ Composite Index hitting new record highs every week and the parallels to 2000 may start to feel eerily familiar.

But… that’s where the similarities end.

Bubbles are marked by “bubble thinking,” where investors stop questioning valuations and believe that an asset must be owned regardless of price. Today, we observe the opposite. Whether it is Liberation Day a few months ago, or a Fed-induced sell-off in 2022, still fresh in investors’ minds, market participants are clearly nervous and complaints about AI-driven irrational exuberance and excessive valuations are front and center across every financial publication we see. Ironically, valuations are far more rational and reasonable in 2025. In March of 2000, the seven largest “internet beneficiaries” traded at approximately 104 times earnings estimates5. The Roundhill Magnificent Seven ETF trades at about 32 times. That’s not cheap, but hardly bubbleliciously expensive either. Cisco, the mega-cap company that was the poster child for internet enthusiasm saw its multiple expand from around 30 times in 1996-1997 to 126 times in March 2000, while Oracle, 2025’s version of Cisco, went from a P/E of approximately 30 times to 106 times. Today’s valuations are not in the same zip code. NVIDIA is at 32 times (lower than its 5-year average of 39 times), Alphabet at 23 times, Meta at 24 times, both around market multiples, Microsoft at 32 times, and even Oracle is at 40 times. Obviously, valuations today should be higher compared to the late 1990s as interest rates are expected to be structurally lower, and the leaders of the digital economy today are driven by platform companies that have proven to be massively more profitable with stronger competitive advantages, and anticipated durable growth.

The dot-com bubble burst after a series of interest rate hikes from the Fed, which peaked at 6.5% in May of 2000, increasing the cost of capital and driving up borrowing costs for tech companies that had taken on heavy debt loads. In contrast, the Fed is likely to resume a rate-cutting cycle with the consensus expectations of 50bps of cuts in the December quarter. Stephen Miran, the odds-on favorite to ascend to the Fed chair post early next year, is on the record that interest rates in the mid-2% range are more appropriate today (compared to 4% to 4.25% currently), and so the path of interest rates in the foreseeable future is clearly pointing downward. With more than $7 trillion sitting in money market funds facing lower yields due to lower interest rates, plenty of dollars could be heading into the stock market.

Watching for tangible signs of AI demand and refreshing the lessons learned from the dot-com bubble is wise. But in our experience, bubbles do not get formed when everyone is anxious and valuation conscious. They certainly do not burst while the Fed is aggressively cutting interest rates, and the leading players are generating tens of billions of dollars in cash and are able to raise capital at will. In our experience the signs are more subtle, like your barber starting to give you stock tips, and your gardener handing in his notice because he is making more money day-trading the market.

The portfolio is constructed on a bottom-up basis with the quality of ideas and conviction level having the most significant roles in determining the size of each individual investment. Sector and country weights are an outcome of the stock selection process and are not meant to indicate a positive or a negative “view.”

As of September 30, 2025, the top 10 positions represented 58.4% of the Fund, and the top 20 represented 84.3%. We ended the third quarter with 42 investments.

Our investments in the IT, Consumer Discretionary, Industrials, Financials, and Health Care sectors, as classified by GICS, represented 98.8% of the Fund’s net assets. Our investments in non-U.S. companies represented 52.4% of net assets, and our investments in emerging markets and other non-developed countries (Argentina) totaled 32.9% of net assets.

We initiated three new investments during the quarter: the collaborative design platform, Figma, the AI-powered med-tech heart disease diagnostics company, Heartflow, and the network security solutions provider, Netskope.

We also added to seven existing holdings: core long-term position in Taiwan Semiconductor as well as our newer position in the leading Latin-American neobank, Nu Holdings, and the aerospace and defense aftermarket parts supplier, Loar Holdings. We also took advantage of stock price volatility to add to our investment in the leading website building platform, Wix and in the leading European out of home e-commerce delivery platform, InPost, as well as to the Identity, Governance, and Administration-focused cybersecurity provider, SailPoint, and the Chinese e-commerce platform, PDD.

We have reduced three existing holdings and exited our investment in Globant in order to reallocate capital to ideas in which we believe the risk-reward equation was more attractive.

Our largest new position in the third quarter was Netskope, Inc., a leading cybersecurity company specializing in Secure Access Service Edge (SASE). SASE is a large, fast-growing IT category that converges several historically separate networking and security products, including data loss prevention, secure web gateways, cloud application access, and private network access. Netskope provides a unified SASE platform that inspects and secures all of a company’s data traffic – whether to the web, the cloud, or private applications – allowing it to enforce security policies, prevent data loss, and stop threats in real-time, all while enhancing network speed and performance. The company serves over 4,300 customers, including more than 30% of the Fortune 100.

The rising cybersecurity threat environment, proliferation of cloud and AI applications, and new regulatory requirements have made SASE a critical priority for large enterprises. Gartner expects the $15 billion addressable market today to nearly double by 2028. Netskope has been gaining market share in this space due to its next-generation technology. Unlike many peers who rely on public cloud infrastructure, Netskope spent years and hundreds of millions of dollars to build and operate its own global network of over 120 data centers. This gives the company full control over performance and results in faster connectivity for its users. Critically, Netskope’s founders designed the platform to deeply inspect and understand the “language of the cloud,” (inspecting communication protocols such as JSON and APIs) allowing for far more granular control over how employees interact with modern applications, including new generative AI tools. This unique architecture has resulted in strong win rates, accelerating revenue growth, and best-in-class customer retention, with a 96% gross retention rate.

We believe Netskope is a big idea with significant market expansion opportunities. New customer growth has accelerated, while the company maintains a healthy 117% net revenue retention rate as existing customers expand their usage. With the average customer still using only a fraction of Netskope’s 20-plus products, there is a long runway for continued wallet share capture. Furthermore, with the heavy capital investment in its global network largely complete, the company has begun to generate strong operating leverage. We expect this trend to continue as Netskope scales, driving substantial free cash flow generation and creating long-term shareholder value.

Our second largest new purchase in the quarter was Heartflow, Inc., a medical device company providing an AI-powered engine to diagnose heart disease. Heart disease is responsible for 1 in 5 deaths in the U.S., and every 40 seconds someone has a heart attack creating an urgent need to quickly and accurately catch the disease before reaching this stage. Heartflow’s solution provides a minimally invasive way to catch blockages in the heart vessels, reducing both false negatives and false positives relative to standard of care today, thanks to their ability to measure the actual disease itself rather than surrogates and systems, which is the standard of care. Heartflow’s CEO, John Farquhar, at the Morgan Stanley conference in September: “30% of the time someone’s told they’re okay, they’re sent home, where in reality they’re not okay. They ought to be getting taken care of. And 50% of the time they’re sent downstream into the cath lab when they don’t need to be there.”

The company has a strong competitive moat, with a repository of 110 million images supplemented by human-aided training that has taken over 10 years to build. We also like the asset-light nature of the business, along with the margin expansion opportunity as the AI algorithm gets smarter with scale and data, enabling Heartflow to reduce the number of employee hours involved in real-time workflows.

Heartflow is also a big idea as it is still in the very early innings of its growth ramp. The company has over 600 peer-reviewed publications, its solution has been approved by the FDA, and its core product has 99% insurance coverage. The company is only low single-digit penetrated into its addressable market, and additionally is launching a new add-on product for plaque analysis that will utilize the same analytics backbone and thus comes in at extremely high incremental margins. We believe Heartflow is a model for the next generation of health care companies, which will be powered by big data and AI in order to make health care more efficient, less costly, and save lives on a broad scale. CEO at the Morgan Stanley conference in September: “And then longer term, we’re going to get into this asymptomatic opportunity... And I do think one day the market’s going to move towards screening for heart disease like you screen for other big killers like cancer.” After following the company for several years and holding multiple meetings with management while it was still privately owned, we decided to participate in its IPO in August 2025.

Lastly, we also initiated a new position in Figma, Inc., a software business that offers both designers and non-designers (designers work alongside developers, product managers, researchers, marketers, writers, and other non-designers) a collaborative product to ideate, visualize, build, and ship software – to help companies deliver the best possible user interface in order to optimize user experience and overall customer satisfaction. With the expansion of its product portfolio over the last few years, the company has broadened its focus from traditional designers to non-designers creating a platform that helps teams across companies of all sizes share and explore ideas, align on a vision, visualize concepts, and translate them into coded products – all on a single, connected, AI-powered platform that collaborators around the world can access with a web browser (think “Google Docs” for everything that comprises software design).

Figma’s business model is 100% subscription based, with a land and expand approach to the over $30 billion TAM comprised of the global workforce engaged in software design. Figma has been taking share over the last decade and has become the dominant vendor in the design space, enabling it to grow well above market – thanks to the quality of its offering, replacing a variety of point solutions across the design and collaboration landscape. The company generated $749 million in subscription revenue, over 48% year-over-year with 30% adjusted free cash flow margins in 2024. The company’s unique product helped it reach an 80% market share within its core design end market (around 85% of revenue) with the key longer-term opportunity to extend into attractive non-designer adjacencies. We believe the company also has a significant opportunity in AI due to the accelerated pace of software development, the growing number of surfaces for creation, and the fact that humans would always be visual creatures, increasing the importance of design as the differentiator for customers. Dylan Field, Figma’s Co-Founder and CEO, further emphasized this during the company’s second quarter earnings call: “Design is now the differentiator. It’s how companies win or lose. And our goal at Figma is to build and expand our platform so we can do even more to unleash the taste and craft of our customers who together are shaping and defining this next era of digital products.”

Our largest add during the quarter was Taiwan Semiconductor Manufacturing Company Limited (TSMC) as we continued to build our position. We believe that while near-term uncertainty is heightened due to the developing China-U.S. geopolitical relations, TSMC’s competitive positioning in leading-edge semiconductor manufacturing remains unmatched with a 90% market share (and 65% overall). TSMC’s unique positioning in the market is underlined with the company’s ability to raise prices as demand for its next generation nodes continues to be robust – this should also enable TSMC to offset any margin pressures from cost inflation or the headwinds from opening fabs in the U.S. at an accelerated pace. We also believe that TSMC will benefit from a long duration of growth as the adoption of AI continues to accelerate. Note that TSMC will benefit regardless of the ultimate market share split between NVIDIA, Advanced Micro Devices, OpenAI or Anthropic and whether application-specific integrated circuits would garner any significant market share. It’s the ultimate picks and shovels supplier to AI.

We also added to our position in the website builder, Wix.com Ltd. Shares responded negatively to the company’s second quarter results despite reporting accelerating key performance indicators - its Q1 2025 cohort was the best post COVID-19 cohort growing 12% year-on-year, while the Q2 2025 cohort was even better at over 14%, and the July cohort grew 20%. Wix also announced the acquisition of Base44 – which is a microcosm of the acceleration in entrepreneurship generative AI enables. Base44 is a vibe-coding startup that was founded by Maor Shlomo, a 31-year old Israeli, in January 2025 by himself, raised no capital and yet was able to get first revenues in March, become profitable in May, and reached $3.5 million annual recurring revenue (ARR) in June. Wix acquired it for $80 million (with additional undisclosed earnouts to keep Maor onboard for the next four years at least), and the company guided to $40 million to $50 million ARR for Base44 by year-end 2025 – with Similar web data released later in the quarter showing a potential acceleration to these numbers. Vibe-coding is essentially the new generation of no-code app builders. It enables users to build a SaaS app without coding at all – by just telling an AI what you want to build. Shlomo described how instead of using external SaaS apps for managing Base44, he would use Base44 to build custom apps, enabling him to save money. While there is a range of outcomes of how successful it could become in a crowded space, Base44 opens a new opportunity for Wix, and if it succeeds, it could become fairly significant and quickly – we are seeing ARR growth rates in the AI space, like we have not seen elsewhere. Apart from Base44, the stock remains attractively valued for the opportunity with free cash flow margins approximating 30%, and a mid-teens growth profile, trading at a 13 times EV/FCF multiple.

While we continue to believe that Shopify Inc. (up 28.6% during the quarter) and argenx SE (up 34.8% in the quarter) are high-quality businesses with significant opportunities ahead, we took advantage of market volatility to manage position sizing and to reallocate to other ideas. We also significantly reduced our position in Endava plc and sold Globant S.A. as we have lower conviction in their future prospects.

Linear thinking vs. exponential growth

In September, we spent a week on the West Coast meeting with management teams of public companies, early and later stage private companies, entrepreneurs, venture capital (VC) investors, and AI researchers. Our goal during these trips is to absorb as much information as we can, test our hypotheses, search for disconfirming evidence (prompting us to update our thinking) and build relationships with companies. It felt different this time than it did even six months ago and even more so than it did last September. The excitement in the air was palpable!

We came away from the trip with several observations:

• We are still very early in the days of AI - enterprise adoption remains relatively low as AI talent is very expensive and hard to find/procure. Production-grade solutions that are easy to use are not generally available yet. We expect this to change over the next few years as the application layer of AI starts to mature and as the cost of using AI continues to drop. According to International Data Corporation, spending on agentic AI is forecast to rise to around $1.3 trillion over the next four years. Agentic AI adoption is similarly early - we expect this to change as the accuracy of models continues to improve (agentic workflows require a much higher accuracy due to the compounding error dynamic in more complex and lengthy workflows).
• Nevertheless, AI is already having a growing tangible impact across the ecosystem - it is a huge driver of VC investment with more than 50% of total global VC investment year-to-date going towards startups in AI⁶.
• Proprietary data continues to be a key ingredient in advancing AI and should serve as a competitive moat driven by a data feedback loop as customer data helps improve the product with the use of AI, driving greater product differentiation, which in turn drives market share gains, more customer data, and so on. We continue to expect winner-take-most dynamics for companies with proprietary data to accelerate.
• On the other hand, in segments in which there is no proprietary data and in which differentiation has mostly been driven by unique product features, we expect AI to reduce barriers to entry and competition to intensify - as imitating the latest features of competitors becomes easier and faster to do.

In addition to this trip, the quarterly earnings reporting season offered a flurry of data points supporting the significant opportunity behind AI in terms of both productivity enhancements and incremental revenues:

• Meta reported quarterly revenue growth of 22% and highlighted AI as a tailwind to engagement and to advertiser’s return on ad spend, prompting the company to increase 2026 estimated capex expectations to $100 billion, $30 billion higher than Wall Street expectations. We thought Mark Zuckerberg’s comments regarding “Early glimpses of AI improving itself” were particularly insightful⁷. If AI can improve itself, a self-reinforcing flywheel can emerge resulting in super-intelligence.
• Andy Jassey, the CEO of Amazon, described AI as the “Biggest technology transformation in our lifetimes. Will change customer experience. Will change how we work - code, analyze, research, business processes, customer service. Every area will be impacted meaningfully by AI".⁸ This is the company that invented Cloud Computing!
• Alphabet reported a doubling in AI tokens in the last two months alone.⁹ “AI is positively impacting every part of the business.” Later, in a separate blog, Alphabet announced that AI will be deeply embedded into YouTube, enabling both video creation with their excellent Veo 3 model, and making videos shoppable.¹⁰ AI would help video creators and brands tag products, significantly expanding the monetization potential of YouTube. Meantime, the latest Gemini model climbed to #1 in the LLM comparison leaderboard across a variety of different tasks.¹¹
• Microsoft’s Azure revenue growth accelerated to 39% with demand continuing to outstrip supply despite the company bringing on 2 GW of new DC capacity online just in the last 12 months. Microsoft’s AI foundry used over 500 trillion tokens in the last quarter, up 700% year-over year.¹²
• NVIDIA now expects $3 to $4 trillion in infrastructure spend over the next five years as demand is proving to be higher than the recent expectations presented to investors during their GTC conference in the spring.¹³ NVIDIA also unveiled its next generation Rubin CPX, an inference-focused GPU that offers 7.5 times more performance than the current generation GB300 NVL72 GPU. This GPU is optimized for the prefill phase of inference and is great for AI video generation and long context windows. Since the prefill phase of inference is highly compute-heavy (and yet memory light), using a regular GPU that has a lot of expensive HBM memory is wasteful. Similarly, running the decode phase, which is memory-heavy on a regular GPU, wastes compute resources. Disaggregating the two processes enables a leap forward in performance¹⁴.
• CrowdStrike described that while the cloud increased their opportunity by 10 times, they expect agentic AI to increase it again… by 100 times as the sheer volume of assets to be protected and the complexity of protecting them is increasing exponentially.¹⁵

This is a non-exhaustive list of recent developments – just over the last three months. Human beings tend to perceive and process change linearly, i.e., incrementally. But as disruptive change investors we know that tectonic shifts in computing platforms often result in changes that are exponential. For example, the length of software engineering tasks AI could do has been doubling every seven months, for the past six years now.¹⁶ That’s around 10 doubles or 1,000 times better with a similar pace of progress identified across a wide range of other tasks.¹⁷ Can you imagine what another 1,000 times in capability would mean for AI over the next six years? To what extent would it be able to make information workers more productive? Accelerate R&D? Discover new materials and sources of energy? New medicines? Cure cancer? What is the value of that to humanity?

There is no question that the very nature of these questions and some of the behavior that we see in the markets is starting to resemble some aspects of the internet bubble. But with the benefit of hindsight, whoever was the most bullish about the internet in the year 2000 was still not bullish enough! There were no smart phones, no tablets, no Wi-Fi, no Google, no Facebook, no TikTok, no cloud, and no video streaming. The internet has become much more important than even the biggest optimists believed. There is little doubt in our minds that 25 years from now, AI will turn out to be even more important and even more value-creating than we currently imagine.