Dear Baron Discovery FundShareholder,

Performance

This was a challenging quarter for Baron Discovery Fund ^® (the Fund), both on an absolute and relative basis. In the first quarter of 2026, the Fund declined 10.65% (Institutional Shares), trailing the Russell 2000 Growth Index (the Index) by 7.84%. We don’t take this lightly, and we have doubled our efforts to understand what is going on in the market both in the short term, and (far more importantly) as it affects the overall long-term embedded valuations of our holdings in the Fund.

Of the underperformance, five buckets accounted for 7.88% (essentially all of it):

• 2.63% came from Information Technology (IT) (software exposure was entirely responsible for the relative shortfall in the sector, but was partly offset by solid relative performance in areas benefiting from the AI secular growth narrative, such as semiconductor, semiconductor materials & equipment, and electronic equipment & instruments related companies)

• 1.76% came from Consumer Discretionary (higher energy prices, inflation and AI induced unemployment fears, plus noise around “prediction markets” competitors to DraftKings Inc. (DKNG) )

• 1.22% came from Health Care (there were no real standout mistakes here, but the market was negative on life sciences tools and health care technology)

• 1.17% came from our lack of exposure to Energy (higher oil prices related to the Iran action moved the sector up 26%) and Materials (aluminum and chemicals prices are up, also related to Iran);

• 1.09% came from Industrials (some of which related to concerns about commercial aerospace suppliers like Loar Holdings Inc. (LOAR) due to the military action in Iran)

Annualized performance (%) for periods ended March 31, 2026 ^†

Of the underperformance, in IT, 3.71% of the relative deficit was attributable to software. If we include two health care companies that are software-related ( Waystar Holding Corp. (WAY) and Heartflow, Inc. (HTFL) ), the total adverse impact from software in the quarter was (4.36%) or nearly 60% of our negative relative performance. These software companies almost uniformly beat earnings, yet shares dropped considerably.

Software has been decimated by the so-called “SaaS-pocalypse” which is shorthand for how the revolution of AI is changing the industry. SaaS stands for software as a service. The market has decided that all software companies are AI losers and, as a result, every one of our software holdings saw significant declines in the quarter. Despite generally strong fourth quarter earnings, the sharp declines have pushed software valuations to levels not seen in more than 15 years. Although the short-term results have been difficult, we see this environment as a chance to invest in truly attractive opportunities across software companies that in our view have strong and sustainable competitive advantages. There are multiple potential catalysts that could quickly change the market’s thinking on these software companies, and we want to be there to reap the benefits when that happens.

Companies like Anthropic (ANTHRO) and OpenAI (OPENAI) have created models known as “frontier, ” “foundation, ” or “large language” AI models (LLMs) that have revolutionized the way we search for and categorize information that is generally publicly available. They have extended their LLMs into software coding, in a way that has become much more accessible to the general population, thereby democratizing software development. It is true that this revolution has made it much less expensive to develop basic software (for professionals and consumers alike). Companies that have value propositions based mostly on their actual code are truly at risk of disintermediation in the world of AI. However, we have largely avoided these types of companies. Our companies should have built-in competitive advantages, which extend far beyond the actual code. Our portfolio companies have their own internally developed AI which is custom tailored to their own domains. Here are a few examples of the differentiation which exists in our investments.

1. Deterministic Data/Infrastructure Protection – LLMs take the data that is available to them and search based upon it. If there is an actual answer to the question being asked, it will be returned. Where no actual answer can be found, a probabilistic “guess” is made in order to fill in the blanks. The answer may be correct, or it may not be (in which case you have what is called a “hallucination”). Software companies that deal with private customer data, not available to LLMs, have a prized possession because software using deterministic data will have an actual answer to a question being asked that in many cases cannot be addressed by an outside LLM. In fact, it may be unsafe, illegal, or out of policy for a company to use an external model, or to allow that external model to have access to its proprietary information.

Good examples of this are regulated companies in industries such as health care and finance. The more complex the environment, the more embedded the legacy software will be in the enterprise. Now these legacy software companies can use AI from an outside LLM through a link called an MCP Server (Model Context Protocol) to help fine-tune their own deterministic data. But there is a cost for using outside AI based on the amount of information “tokens” consumed. And breaches of MCP Server software have also been reported (see below). Cybersecurity companies in particular have the advantage of seeing all of a company’s data and parsing it for particular threats to the internal network or application structure of that company.

The brands of these companies are valuable as they have built up years’ worth of trust with their customers. This is why we have invested in SentinelOne, Inc. (S), which provides endpoint protection using its own AI algorithms for cyber-breach discovery and remediation, . The same is true for observability software (which “instruments” everything that moves through a network or attaches to it, as well as the applications and data related to that movement). We own Dynatrace, Inc. (DT) which is architected on its own internal AI to predict failures in network software and hardware (whether in the cloud or on-premise) and works to automatically remediate the issues. It’s used by the largest companies in the world that operate in the most complex environments (airlines, financial giants, and defense companies for example).

Deterministic/infrastructure oriented companies gain nearly all of their value by integrating with and servicing their clients’ needs, rather than just by selling an off-the-shelf software package. Such software provides high return on investment (ROI), auditable security compliance, and peace of mind at a reasonable cost. Even if cheaper software solutions that were coded using LLM platforms came out, they would still have to be integrated and maintained into the enterprise’s architecture, and they would have to link to outside LLM’s for AI capability (which could cost a LOT more to run in the future versus what existing vendors charge for their “tuned” and more specific AI models). We believe that these companies will become even more valuable in an “agentic AI” world, where software autonomously executes tasks based on user goals, operates with its own enterprise privileges, and must be monitored and controlled.

2. Network Effects Vertical Vendors – Some companies serve a very specific customer base and provide increased value by giving each customer the benefit of understanding (using hard to compile domain specific data) what is going on in their industry. Examples of this include ServiceTitan, Inc. (TTAN), which provides software for service trades such as plumbing and HVAC. It is an all-in-one platform for lead generation, job bookings, dispatching, estimating jobs, customer communications, and payments/financing. Each trade has its own specific characteristics and regional data on pricing, competition, service times, and contract terms that ServiceTitan understands deeply. It is not easy to switch the software out, particularly because it helps businesses automate their processes and minimize the overall personnel needed. Procore Technologies, Inc. (PCOR) provides integrated construction software, which is required by many of the major general contractors in order for subcontractors to be able to participate in a construction project. The software combines computer-aided design software blueprints with job scheduling, cost estimations, materials costs, and change order management. In this manner, the job site can be coordinated among all the different parties involved in the construction project. It is truly a community-oriented platform that is not easily replaced.

3. Atoms Plus Electrons – These are hybrids of software and hardware. They are in some ways the most protected because AI in and of itself can’t create hardware. Companies like Netskope, Inc. (NTSK) fit into this category. Netskope is a misunderstood company which provides secure access service edge (SASE) functionality for zero trust network access (ZTNA), data loss protection, and threat protection to its enterprise customers. It uses a proprietary network of worldwide data access centers as gateways for access to enterprise network resources, web resources, and applications. These physical data centers allow much faster data movement as well as for in-line scanning of network data for security purposes. The company is not earning full margins yet because it has invested in building its physical network (which is part of the reason it is down in the quarter). However, NetSkope is now starting to reap scaled revenue benefits, and its physical network gives the company an advantage over purely software-based ZTNA solutions in that it is safer and provides much faster overall network access (lower latency or delay). It cannot be replicated by software alone.

4. Regulated Industries – Some industries like health care in particular are heavily regulated, with extreme penalties for misuse or loss of patient information. And in some cases, such as with Heartflow (which uses AI software to map coronary arteries to assess blood flow and plaque buildup without an invasive procedure), clinical trials and Food and Drug Administration (FDA) approval are required before the software can be used.

While this discussion is important, the more practical question is when the market will begin to recognize the wide dispersion in intrinsic value across the software universe. We believe several catalysts are emerging that should separate the winners from the losers.

First, it is likely that we will see increased merger activity. Private equity funds specializing in software have recently raised tens of billions of dollars and would be very sophisticated buyers of high-quality companies at historically depressed evaluations (we have had eight companies acquired in this space in the last six years). Additionally, we are seeing strategic buyers from within the technology space purchase software companies. Last year we had two software companies purchased by such buyers, including CyberArk Software Ltd. (CYBR), a high-end cybersecurity company which was bought by Palo Alto Networks (PANW) (announced in July 2025 and closed in February 2026).

Second, it is almost inevitable that there will be cyber-attacks based upon usage of LLM based AI within enterprises if the technology is not properly secured and controlled. We have already seen such an attack. In March 2026 LiteLLM, an LLM gateway tool (which allows developers to link their applications to over 100 different LLMs) was used as an attack vector. Poorly secured coding in this widely used tool led to widespread malware infiltration. The attack was so sophisticated that it allowed the attackers to rapidly spread the malware across on-premise and cloud resources and exfiltrate sensitive data to an external server. SentinelOne recently released a technical paper which showed how its own AI-driven software automatically and rapidly protected its users by finding and shutting down this attack and provided an audited trail of the attack vector itself.

Third, we are likely to see partnerships between legacy software companies and LLM providers, which will highlight the “last mile” deterministic data value of legacy software companies. Recent examples include partnerships with OpenAI and transaction processors such as Instacart (CART), as well as a partnership with SentinelOne and Google (GOOGL) (to provide autonomous, AI-based cloud security for Google Cloud customers).

Finally, we expect continued solid financial performance from companies with the protected characteristics described above. During the past quarter, our holdings generally delivered results ahead of expectations and raised guidance. We believe this trend will persist, and that growing free cash flow will ultimately capture investors’ attention. Yet valuations are lower than they have been in over a decade. As we have noted in past letters, software companies have incredible financial characteristics, including outsized margins, strong balance sheets, and the ability to actually generate more free cash flow as they grow (due to the upfront payment of subscription fees). For all these reasons, we have maintained our overweight in the software space, and we believe that we will see significant outperformance for years ahead of us.

Top contributors to performance for the quarter

Advanced Energy Industries, Inc. is a designer and manufacturer of products used to transform, refine, and modify electrical power for use in semiconductor, industrial, medical, data center, and telecommunications end markets. Advanced Energy’s stock rose during the quarter as earnings and guidance were better than expected and as the market began to appreciate the strength that the company would see in both its data center and semiconductor end markets. The company is enjoying the fruits of having repositioned its data center segment to focus on sole-source, differentiated, higher margin business. AI’s increasing power requirements play to Advanced Energy’s strengths in power density and efficiency. The company also recently launched new products into the semiconductor market which are expected to drive strong growth through this year. Combined with the early stages of a recovery in its industrial and medical end markets, Advanced Energy is poised for several years of continued strong growth and margin expansion. The company also remains focused on acquisitions to bolster its product offerings, particularly in the large fragmented industrial and medical spaces.

Masimo Corporation is a medical device company that manufactures and sells a variety of non-invasive patient monitoring technologies, including its well-known pulse oximeters used to measure blood oxygen levels. Shares outperformed for the quarter after Danaher Corporation (DHR) announced that it would acquire Masimo at a 38% premium. This was a special situation driven by an activist investor that worked out very well for the Fund.

Arcellx, Inc. is a biotechnology company which uses CAR-T technology (modifies a patient’s own immune cells to recognize and destroy cancer cells) to treat multiple myeloma. It is due to be acquired by Gilead Sciences Inc. (GILD) in June (around which time we expect that Arcellx will receive FDA approval for its drug called Antio-cel).

Top detractors from performance for the quarter

Intapp, Inc., a vertical software platform serving private equity, legal, and consulting firms, detracted from performance this quarter. The drawdown was driven by a sector-wide AI disruption narrative that hit legal-adjacent software stocks particularly hard, with Intapp declining sharply through mid-February after Anthropic announced new legal tools. We sold our investment in the quarter as we believe that our other software holdings have better overall competitive advantages.

DraftKings Inc. is the leading U.S. digital sports betting and iCasino operator. The stock declined as investors grappled with a guidance range that implied handle (amount bet) deceleration, elevated prediction markets investments to compete with firms like Kalshi (KALSHI) and Polymarkets, and lingering debate around structural hold (the percentage of overage profit per bet) sustainability. The headline concerns obscure what we believe are strong fundamentals in the core sports betting business customer cohorts. Management built 2026 guidance on flat actual hold, a figure that has expanded every year in the industry’s history. Parlay mix, the primary mechanical driver of hold, increased 500 basis points during NFL season and 200 basis points year to date. The $800 million EBITDA midpoint also embeds a $200 million headwind from prediction markets investment, which currently carries no associated revenue. Excluding that impact, implied core business EBITDA exceeds $1 billion. We believe the stock is trading at attractive multiples relative to the company’s long-term earnings potential and think the total addressable market for prediction markets, while nascent, has the potential to accelerate growth.

Shares of Netskope, Inc., a cloud security and networking platform for enterprises, were down due to a combination of sector-wide and technical factors rather than fundamental weakness. The entire application software sub-sector experienced a sharp drawdown as investors weighed AI disruption risks, and recent IPOs like Netskope bore the heaviest losses. Adding to the pressure, Netskope’s lock-up expiration in mid-March made roughly 390 million shares eligible for sale, creating a supply overhang that coincided with the worst of the sub-sector selloff. The business itself performed very well— fiscal fourth quarter (ended January 31, 2026) revenue grew 32%, annualized recurring revenue (ARR) reached $811 million, and grew 31%, the company posted record quarterly net new ARR, and achieved positive free cash flow for the first time. Management guided fiscal 2027 revenue above consensus expectations. We maintain conviction in Netskope’s long-term positioning in the SASE market, where demand for securing cloud and AI workloads continues to grow, and view the current valuation as disconnected from the company’s growth trajectory and competitive standing.

Portfolio Structure

Top 10 holdings

Recent Activity

Top net purchases for the quarter

Forgent Power Solutions, Inc. is a leading manufacturer of electrical distribution equipment used in data centers, the power grid, and energy-intensive industrial applications. Forgent is a low- and medium-voltage equipment specialist and focuses on custom, “engineered-to-order” products (90% or more of revenue) whereas larger competitors in the industry generally focus more on higher voltage and standard products. Forgent differentiates itself from competitors by engaging deeply with customers in the design phase and then offering custom products in shorter lead times than the standard products sold by competitors.

The company has nearly completed a manufacturing footprint investment which will support $5 billion in revenue, giving it one of the largest state-of-the-art manufacturing footprints in the industry. Plus, it has very good visibility with about $3 billion in annualized orders, with a $1.5 billion current backlog. Electrical equipment, especially power transformers, remains a key bottleneck in the broader data center infrastructure build, and Forgent’s capacity planning and manufacturing efficiency are uniquely positioned to take advantage of this supply/demand mismatch. Despite inefficiencies from excess capacity, Forgent already has near best-in-class adjusted cash flow margins, which we expect to continue to expand as the company drives more volume over its large manufacturing footprint. To date, most of its data center business has focused on colocators and neoclouds, with very large opportunities to engage with and support larger hyperscale customers going forward. We believe Forgent can grow its revenues to over $5 billion in the next five years (from $296 million in 2025 and an expected $1.3 billion in 2026) supported by continued robust grid and data center capital expenditure as well as share gains from competitors in the market.

Enpro Inc. is a diversified industrial technology company whose proprietary, value add products and solutions provide critical functionality and protection across a wide range of demanding environments. Today, more than half of revenue is generated from recurring, high margin aftermarket applications, and a similar proportion is exposed to structurally higher growth end markets. Enpro’s Sealing Technologies segment designs, engineers, and manufactures metallic seals, soft gaskets, wheel end products, and gas analyzers and sensors serving general industrial, commercial vehicle, power generation, food and pharmaceutical, aerospace, and petrochemical markets, supported by strong brands such as Garlock, which is widely regarded as the “Kleenex” of its category. The Advanced Surface Technologies (AST) segment is focused on the semiconductor market and provides precision manufacturing, cleaning, refurbishment, and coating services to leading wafer fabrication equipment original equipment manufacturers and foundries, with a particular emphasis on leading edge production.

We believe Enpro can deliver mid to high single-digit organic revenue growth over time, with EBITDA margins expanding into the high 20% range from the low to mid 20% range today, supported by contributions from both segments. Sealing Technologies should continue to achieve above GDP organic growth driven by strong pricing power and ongoing investment in innovation and attractive growth markets. AST is positioned to benefit from a multi year secular growth opportunity driven by increasing leading-edge semiconductor spending and a rising U.S. share of global manufacturing, particularly supported by AI driven demand in the near term. We also expect the company to continue deploying its strong free cash flow toward highly complementary acquisitions, leveraging its operational excellence capabilities to drive value creation. As Enpro continues to scale and margins improve, we believe the business will warrant a more premium valuation, supporting further upside over time.

We added to our position in Dynatrace, Inc., a provider of “observability” software. For the reasons we laid out above we believe that this is a great deterministic data-oriented company, benefiting from significant competitive advantages. However, it is trading at a rock-bottom multiple (13 times free cash flow, with that metric is likely to grow in the mid-teens for the next few years).

We also added to Heartflow, Inc., whose software analyzes CT scans of a patient’s coronary arteries done with contrast, and shows calcification, plaque buildup, and blood flow quality in a three-dimensional model. It is hard to understand how Heartflow would be easily disintermediated, given the customer trust it has built up, and its FDA approved software based on significant clinical trials and millions of real-world CT scan analyses.

Finally, we added to Waystar Holding Corp., which like Heartflow has been lumped into the “AI software losers” bucket. Waystar is a provider of revenue cycle management software to health care providers. The company has an AI driven, end-to-end suite of solutions that saves clients massive amounts of working capital costs by getting claims submitted quickly and correctly, and by automating insurance appeals when necessary. At under 11 times adjusted cash flow, but growing cash flow in the low teens, we believe the company is competitively advantaged and very cheap.

Top net sales for the quarter

We sold several positions in the first quarter, mostly relating to companies set to be acquired. These included Exact Sciences Corporation (a cancer diagnostics company acquired by Abbott Laboratories (ABT) in March), Masimo Corporation, Clearwater Analytics Holdings, Inc. (an investment accounting SaaS company due to be acquired by multiple private equity firms in June), and Arcellx, Inc. We also sold our remaining position in GitLab Inc. (a software company that enables enterprises to coordinate the development and production of software), as we came to the view that the company had the potential to be disintermediated by LLM developed solutions.

Conclusion

We hate to underperform. We “eat our own cooking, ” as we have personally invested meaningful amounts of our net worth in the Fund. Rest assured that we are devoted to our process of investing in competitively advantaged companies with great management teams for the long term. We spend hours every day performing due diligence on our companies, including speaking with management teams, competitors, industry experts, and customers. So, we have true conviction in our investments for the reasons laid out above. Sometimes we are too early. But we believe we are not far away from seeing outperformance related to our hard work. We are grateful that you have chosen to take this journey with us.

Randy Gwirtzman, Portfolio Manager

Laird Bieger, Portfolio Manager

References

1. † Historical performance was impacted by gains from IPOs. There is no guarantee that these results can be repeated or the level of IPO participation will be the same in the future.
2. 1 The Russell 2000® Growth Index measures the performance of small-sized U.S. companies that are classified as growth. The Russell 3000® Index measures the performance of the largest 3,000 U.S. companies representing approximately 98% of the investable U.S. equity market, as of the most recent reconstitution. All rights in the FTSE Russell Index (the “Index”) vest in the relevant LSE Group company which owns the Index. Russell® is a trademark of the relevant LSE Group company and is used by any other LSE Group company under license. Neither LSE Group nor its licensors accept any liability for any errors or omissions in the indexes or data and no party may rely on any indexes or data contained in this communication. The Fund includes reinvestment of dividends, net of withholding taxes, while the Russell 2000® Growth and Russell 3000® Indexes include reinvestment of dividends before taxes. Reinvestment of dividends positively impacts the performance results. The indexes are unmanaged. Index performance is not Fund performance. Investors cannot invest directly in an index.
3. 2 The performance data in the table does not reflect the deduction of taxes that a shareholder would pay on Fund distributions or redemption of Fund shares.
4. 3 Not annualized.

Original Post

Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.

Editor’s Note: This article covers one or more microcap stocks. Please be aware of the risks associated with these stocks.