Baron Durable Advantage's institutional shares returned 2.23% in the third quarter of 2026, just behind the S&P 500's 2.30% gain. Through Sept. 30, the fund was up 5.70% for the year, compared with a 12.75% rise in the index. In its quarterly letter to shareholders, the management team said the index's steady advance masked sharp swings in AI infrastructure stocks, while earnings growth at technology companies remained a key market focus.
AI infrastructure stocks fell, then rebounded
Early in the quarter, concerns mounted over whether AI investment would generate sufficient returns and whether demand could be sustained. The letter cited speculation that Meta might rent out some of its AI computing capacity, prompting questions about the company's internal demand. Growing adoption of open-source models, falling token prices and discussion of slower growth at Anthropic and OpenAI also weighed on AI-related shares. Sentiment took a further hit after the collapse of a hedge fund with large leveraged positions in semiconductors and AI infrastructure.
By the end of July, semiconductors were down 8.4%, a sector representing about 16.6% of the S&P 500. Semiconductor materials and equipment fell 32.5%, while electronic components dropped 27.7%. Previous leaders including Micron, Sandisk, Corning and Lam Research declined by nearly 20% to about 45%. Industrial groups tied to AI buildout also weakened: construction machinery and heavy transportation equipment fell 16.5%, and heavy electrical equipment lost 15.7%. Gains in energy, financials and the so-called Magnificent Seven limited the S&P 500's decline for the month to 0.1%.
Results from hyperscalers, cloud providers and semiconductor companies eased some concerns in August. Hyperscalers' cloud businesses posted a 50% year-over-year increase in annualized revenue run rate, to $420 billion, the letter said. Cloud providers reported higher contract prices rather than declines. CoreWeave secured a contract through 2029 to lease Ampere GPUs it had launched six years earlier, while Nebius said prices had risen for both new and older GPUs. The fund owns Nebius but not CoreWeave.
In September, Nvidia reported revenue of $96 billion, up 106% year over year, and forecast 70% growth in the next fiscal year. The company said supply constraints had affected its guidance. Broadcom's revenue rose 86% year over year; it expects AI revenue to double in the next fiscal year and double again in the following year. The fund's managers viewed the results as evidence that AI-related demand is still expanding, while noting that the sector remains volatile and sensitive to changes in growth expectations.
Earnings growth drove the S&P 500's advance
The fund's letter said the S&P 500's price-to-earnings ratio fell to 19.3 from 20.2 during the third quarter, meaning earnings growth accounted for most of the index's 2.3% return. Year to date, the multiple had declined to 19 from 22. Meanwhile, year-over-year growth in S&P 500 earnings per share accelerated from 13.9% in the fourth quarter of 2025 to 28.9% in the first quarter of 2026 and 52.2% in the second.
Earnings in the first two quarters beat market expectations by 16.4% and 25.8%, respectively, well above the roughly 8% average surprise over the past decade. The fund expects third-quarter earnings-per-share growth of 29.6%; applying the average earnings surprise would put growth at about 37%.
AI-related companies were among the significant sources of earnings growth. The letter cited second-quarter year-over-year EPS growth of 111.8% at Nvidia, 29.9% at Microsoft and 242% at Amazon. For Amazon, it said operating profit growth of 43% better reflected business performance after excluding gains from its investment in Anthropic. Alphabet's EPS rose 294%; excluding gains from investments in Anthropic and SpaceX, operating profit increased 30%. Broadcom's EPS grew 96.4%.
Stock selection helped, while tech holdings weighed
Thirteen stocks in the fund contributed positively in the third quarter and 16 detracted. Microsoft, Meta and Nvidia each added more than 100 basis points to absolute returns, while Taiwan Semiconductor Manufacturing Co. and Lam Research each subtracted about 50 basis points. Stock selection contributed 62 basis points, but sector allocation detracted 70.
Positions in communication services, consumer discretionary, financials and real estate contributed positively. Having no utilities holdings added 32 basis points. Information technology holdings detracted 135 basis points, and the fund's lack of exposure to energy stocks, the best-performing sector in the quarter, cost a further 45 basis points.
The fund named Microsoft, Meta, Nvidia, Amazon and CME Group among its leading quarterly contributors. Microsoft added 1.74% to returns, Meta 1.30% and Nvidia 1.15%.
As of Sept. 30, the fund's retail shares had gained 2.15% in the quarter and its institutional shares 2.23%. One-year returns were 8.22% and 8.49%, respectively, compared with 15.74% for the S&P 500. Over five years, annualized returns were 13.45% for retail shares, 13.74% for institutional shares and 13.79% for the index.
Since its launch on Dec. 29, 2017, the fund's institutional shares had returned 15.64% annualized, versus 14.59% for the S&P 500. The letter also reported that the fund outperformed the index in 80% of rolling three-year periods and the Morningstar Large Growth category average in 84%. For rolling five-year periods, the figures were 93% and 100%, respectively. These figures describe past performance and do not indicate future results.
Fund groups AI holdings into three categories
The managers divide AI-related investments into three groups: companies building AI infrastructure, cloud providers operating computing facilities, and businesses applying AI to core operations and generating revenue from it. Infrastructure holdings listed in the letter include Nvidia, Broadcom, Taiwan Semiconductor Manufacturing Co., Monolithic Power Systems, Amphenol, Lam Research and Quanta Services. Cloud provider holdings include Microsoft, Amazon and Alphabet.
The letter cautioned that the AI industry remains at an early and changing stage. Small shifts in growth expectations or key inputs could have an amplified effect on long-term outcomes, while the direction of the industry and its eventual winners remain uncertain.