AI, the Elephant in the Market
The second quarter of 2026 was a strong one for US stocks, with nearly all major indices posting double-digit returns. The S&P 500 gained 15.20% in the quarter, and the Russell 2000 gained 21.49%. That performance more than offset a somewhat weak start to the year, leaving the S&P 500 up 10.21% year to date, and the Russell 2000 up 22.57%.
The strength of the US stock market was underpinned by a number of factors. The US economy continued to grow despite a multitude of headwinds. Interest rates remained stable despite rising inflation. Corporate profits continued to advance at impressive rates. And valuations of US stocks remain reasonable, near the averages of recent years.
The most apt adjective to describe the US economy is resilient. Despite numerous geopolitical tensions, the disruptive effects of tariffs, elevated oil prices, and troubling rates of inflation, the US economy continues to expand, with GDP (Gross Domestic Product) growing at about 2.1%. That rate may sound modest, but it is near the long-term average for the US economy and is adequate to enable most businesses to report healthy profits. Corporate profits for the S&P 500 rose 13% in 2025 and are estimated to increase by 26% in 2026.
While the economy in aggregate is expanding at an acceptable rate, the drivers of that growth have changed. Historically, consumer spending has been the biggest contributor to growth, as it represents over two thirds of GDP. But recently consumer spending has been restrained. High net-worth individuals, whose fortunes have been buoyed by record stock prices, have continued to spend, while rising prices for necessities like gasoline have forced lower-income consumers to cut back. The result is what has been described as a “K-shaped economy.”
The relatively modest growth in consumer spending has been offset by a dramatic increase in capital spending, particularly the massive investments being made in data centers. That spending is driven largely by the hyperscalers (AMZN, GOOGL, META, MSFT, ORCL), which are racing to build the computing capacity needed to support AI (artificial intelligence) applications.
The cumulative spend by the hyperscalers is enormous and unprecedented in US history. The five largest companies are expected to invest over $750 billion in 2026, up roughly two thirds from 2025, and that spending is increasingly being funded with debt, as it outruns even these companies’ prodigious cash flows. Capital spending on AI is estimated to contribute anywhere from 0.3 to 0.5 percentage points to GDP growth in 2026.
It is worth reflecting on what all this money is actually buying, because the capabilities of AI have advanced considerably in a short period. The first wave of applications centered on large language models (LLMs) such as ChatGPT and Claude, which could write software code as easily as a term paper. The current generation goes further. Reasoning models can work through multi-step problems, check their own output, and match skilled professionals on a growing range of analytical tasks. The newest systems, commonly called “agents,” do not simply answer questions. They conduct research, write and test code, and carry out multi-step workflows with limited human supervision. Surveys indicate that more than half of large organizations now deploy AI across multi-stage workflows, with software development, customer service, and document-intensive functions such as legal and compliance being reshaped first.
With AI capability improving with each new model generation, tasks that were impossible three years ago are becoming routine today. Moreover, unlike most prior technologies, AI is improving fastest at precisely the kind of cognitive work that was long considered immune to automation. That combination explains both the enthusiasm and the anxiety surrounding it.
What the future of AI will mean is the subject of much speculation, and we are skeptical of anyone claiming certainty. The optimistic case foresees a broad productivity boom, with some estimates placing the potential economic value in the trillions of dollars as the technology spreads through the economy. The labor-market evidence is more nuanced. So far aggregate employment numbers show little impact from AI, but hiring for entry-level positions in the most AI-exposed occupations has slowed noticeably, and employers increasingly expect new hires to work alongside AI tools. As with prior technological shifts, new roles and industries will emerge, but the transition is likely to be uneven.
For investors, the central questions are familiar ones of capital allocation. Will the trillions of dollars being spent on AI generate an adequate return on that investment? How durable is the competitive advantage of any given AI model when each generation of technology depreciates quickly, and rivals close the gap within months? The stock market is attempting to sort the winners from the losers, and that sorting has introduced a high level of volatility, particularly among the early entrants.
Much of the media coverage of the AI revolution focuses on the largest players: model builders like Anthropic and OpenAI, semiconductor companies like Nvidia and Micron, and the previously mentioned hyperscalers. But the AI supply chain is quite long and encompasses a large number of companies, including many less glamorous industries that supply the vital infrastructure needed to construct data centers. Power generation, cooling and HVAC systems, communication equipment, water management, backup power, and specialized construction are all in demand. Many of the companies providing that infrastructure are smaller industrial businesses, and a number of them are seeing improved order books and pricing power as demand runs ahead of supply. That is the segment of the AI economy where much of our research effort is focused, because it is where careful analysis can still uncover quality businesses at reasonable prices.
Stock market valuations are not unrealistically high, as earnings growth is exceeding stock appreciation. The forward P/E for the S&P 500 is now about 21 times, not far from its 10-year and 20-year averages. The future of AI is unknown in detail, but rest assured, it will be a big deal, and the spending will continue for at least the next few years and drive the US economy.
The US has experienced transformative and risky developments before. The railroads opened a continent, yet many railroad investors were ruined. Many early investors in building the electric grid lost substantial sums. The fiber-optic boom of the 1990s laid the cables that carry today’s internet, but much of the capital that financed them was lost. Transformative technology and rewarding investment are not the same thing. The difference usually comes down to the price paid. Some companies will see tremendous hype and extraordinarily high valuations. There will also be many great businesses that benefit from the adoption of AI at prices that make sense. That is where we will keep our focus: buying great businesses at attractive prices.
As always, we appreciate your continued confidence and support.
Sincerely,
