Some investors who remember the harsh, dramatic stock market downturns of 2000 and 2008 may fear that we are on the precipice of a third downturn, driven by excessive enthusiasm for artificial intelligence (AI) stocks. The AI investment cycle resembles the dot-com internet bubble of 2000 in that there is a massive buildout of infrastructure based on the hopes of a new technology. AI looks to be at least as revolutionary as the internet in changing our daily lives and our economy. The current market is less comparable to the Great Financial Crisis of 2008-2009, when a speculative bubble in house prices, spurred on by lax lending standards, brought down a much larger part of the economy.
Some investors can’t get enough “AI” in their portfolios. Not all investors have a clear sense of which businesses will prove to be economically viable, and they are throwing money at AI stocks indiscriminately. This will likely turn out to be a mistake: any number of stocks will turn out to be bad investments.
The largest corporate investors in AI are the hyperscale data centers, Alphabet, Microsoft, Amazon, and Meta Platforms. In addition to their capital expenditures, Microsoft invested $13 billion in OpenAI, and Amazon invested $8 billion in Anthropic. Both investments are now valued at multiple times those initial amounts. The four largest data center players had strong businesses before the advent of AI, and continue to generate strong cash flow from those businesses. These companies didn’t become some of the largest businesses in the world by having lax financial discipline. Their creditworthiness remains high. Their stocks would certainly fall if they did not earn decent returns on their AI investments, but they would remain very solvent falling back on their core businesses.
AI promises to do a fairly good job of automating the routine, mindless parts of our jobs for us. That’s not to say results from AI algorithms aren’t wrong with some regularity and should not be supervised,[1] but we are banking on the technology improving and safeguards being in place. There are multitudes of real-world applications driving interest in AI. Companies are using AI to optimize everything from advertising placements, shipping routes, and processing insurance claims to new ingredient combinations, drug discovery, and molecular modelling simulation. Self-driving cars and robotics will consume considerable amounts of computational power in the coming years. Like the internet, these technologies will take years to develop.
AI could make millions of employees redundant. That was true for the industrial revolution, and it has been true for technology ever since. Observers were worried that the automatic switchboard would put telephone operators out of work in the 1960s. It did. Observers were worried that the automated teller machine (ATM) would put bank tellers out of work in the 1970s and 1980s. It did. The more optimistic take is that if AI improves productivity, competing firms can lower their prices, providing more services, and providing services to more people. That is the stuff of economic expansion. While some areas of the economy will contract, others will expand. We are optimistic this will lead to growth overall, as has been the economic norm. It’s too early to know which companies implementing AI will be winners and losers.
AI is already generating billions worth of revenue for the hyperscale data centers, which are saying demand is exceeding what they can provide. Demand is only likely to grow. Spending on AI hit $1.75 trillion globally in 2025,2 with about $344 billion of that coming from the big four US hyperscalers. Global spending is expected to hit $2.5 trillion in 2026 and $3.3 trillion in 2027, according to Gartner. The global economy was estimated to be $104 trillion in 2025, making AI a noteworthy but still small component of the overall global economy. These forecasts could prove too optimistic and AI developers could spend too much.
Some see the AI spending boom as a race among the largest technology companies to develop artificial general intelligence (AGI), a level of intelligence which exceeds that of humans. We don’t believe our clients should worry about AGI, and believe the spending boom is justified based entirely on real world applications for the more rudimentary AI. AI proliferation could, however, be limited by our ability to build enough electrical generation capacity to fuel the AI data centers. Current data centers in the US use an estimated 62 gigawatts (GW) of capacity. Demand is expected to rise to 134 GW by 2030.3 While energy constraints could be a cause for the bubble to burst, it would suggest underinvestment in power generation capacity rather than overinvestment in AI compute capacity.
None of this is inconsistent with the idea that AI is in a bubble. The internet was wildly hyped in 2000. That hype turned out to accurately forecast how the internet would develop, if not underestimate its potential, but there were also bad investments. Many people saw their net worth decline, and the US suffered an 8-month recession in 2001.
Diversify!
Because the Magnificent Seven stocks comprise such a large percentage weighting in the S&P 500 Index, the bursting of an AI bubble would likely lead to a decline in the index, but that doesn’t even mean most stocks would decline. There could be plenty of stocks not directly related to AI, and hitherto obscured by the AI enthusiasm, which come into their own as the bubble bursts. The year 2000 was an excellent time to invest in value stocks. While both value and growth stocks fell from the market top on March 10, 2000, looking at 5-year returns starting from the market peak, the Russell 1000 Value index was up 46.6% for the period, or 7.9% per year, while the Russell 1000 Growth Index was down -39.8%, or -9.6% per year. Investing at any other time than the day the market peaked, an investor would likely have higher returns. For the current cycle, it’s also possible some investment factor or theme other than value outperforms. Value stocks did not outperform during the Great Financial Crisis— growth stocks did.
Portfolios at Woodstock for the most part have moderate exposure to AI. We believe most clients should have some exposure, since the industry is still developing and may still have some strong winners. We endeavor to invest in high-quality companies, with defensible moats and strong free cash flow. Many of these dominant companies will likely prove to be most advantaged by AI. The more important point is to maintain broadly diversified portfolios, which are likely to have some winners and some losers in almost any macroeconomic regime.
While the broader stock market can and does decline, the Equal-Weighted S&P 500 Index, which minimizes the importance of the Magnificent Seven, trades for an estimated 17.2x forward earnings, not far off its 10-year median price-to-earnings ratio of 16.6x. The Magnificent Seven as a group trades for 29.8x forward earnings. That’s still well below the nosebleed levels technology stocks traded at during the height of the dot-com bubble. For clients with a long-term time horizon, we recommend maintaining positions in high-quality AI stocks, and that would be all the more true if they were to sell off. It’s not clear what level of investment will prove to be excessive, and when bubbles have burst in the past, the market has always come back. Clients who are not comfortable with the intermediate-term risks should discuss their exposure with their portfolio manager. For all clients, regardless of their investment time horizon, we recommend keeping enough cash on hand to meet near-term obligations.
— Adrian G. Davies, CFA — President