by Kenneth J. Entenmann CFA®
Posted November 19, 2025
“If you build it, they will come.” That famous, whispered line from the movie Field of Dreams seems appropriate for today’s equity market. In April, the market sold off in an overreaction to China’s Deepseek AI model and “Liberation Day” tariff announcements. From Feb. 28th through April 8th, the S&P 500 was down nearly 20%. Since then, it has been on a tear, reaching record high levels multiple times in the Fall. The primary driver of this strong move has been Artificial Intelligence (AI). Indeed, the top 10 stocks in the S&P 500, led by the “hyperscalers” (Microsoft (MSFT), Alphabet (GOOG), META (META, formerly known as Facebook) and Amazon (AMZN)) account for nearly 40% of the S&P 500 market cap! But they also generate 33% of the S&P 500 earnings! The incredible earnings have lifted the Price/Earnings (P/E) ratio of the top 10 to a lofty 32x earnings. The rest of the S&P 500, the other 490 companies, trade at 19.5 and the overall S&P 500’s P/E stands at 23 times earnings. Historically, that is an “expensive” P/E relative to the 30-year average of 17.1 times earnings. Nonetheless, the power and potential of AI have been used to justify the high valuation of the AI universe and the overall market. Don’t worry about high P/Es, this time will be different! The most dangerous sentence in the world!
To be clear, AI does appear to be the next great innovation that has the potential to transform economic productivity. Over the last 150 years, there have been several periods associated with great innovation and high productivity gains. In the late 1800’s, mechanization transformed the economy from an agrarian economy to an industrial economy. The automobile in the 1920s. Post WWII brought electrification and aviation. More recently, the invention of the personal computer, mobile phones and the internet had a dramatic impact on productivity and economic growth.
The history of economic innovation provides lessons for investors. In every case, the initial projections for these transformations proved to be ridiculously low and the ultimate impact was magnitudes better. It is possible that today’s AI projections, as grandiose as they are, will underestimate AI’s overall economic impact. For investors, there is danger in this unbridled enthusiasm for AI. Let’s give AI the benefit of the doubt and say it will be transformative beyond our imagination. That does not mean every AI stock will be a great investment. And that includes the biggest and best “first movers.” There are graveyards filled with companies that were early in the great transformations of the last 150 years. Yet, when was the last time someone drove a Packard? Flew on TWA or Pan-Am? Took a picture on Kodak film? Typed a letter on a Smith Corona typewriter? Fired up their Compaq computer, with its Intel microprocessor, and hit the screeching AOL modem to get on the internet. Have you Asked Jeeves recently? Sent a text on your Blackberry? Went to your My Space page? A core maxim of this blog is that great innovation invites an onslaught of competition. And not every great innovation proves to be great. Think the Metaverse!
The S&P 500 peaked on October 29 at 6920. Today, it trades 6590, nearly a 5% correction in less than 3 weeks. Does this mean the AI trade is over? Hardly. Given the 40%+ rise since the April 8 lows, a modest correction should not be too surprising. But why now?
Overall, the economy is ok. Yes, the employment market seems to have cooled but is far from flashing recession warnings. Inflation remains sticky. This puts the Federal Reserve Bank between a rock and a hard place. The Fed has used the cooling employment market to justify its .25% rate cuts in September and October. But inflation has remained sticky and uncertain given the ever-changing impact of tariffs on prices. In addition, the Federal government shutdown has delayed the release of hard economic data. One reason for the sell-off is that expectations for an additional .25% rate cut in December have been put on hold. When the market trades at 23x earnings, lower interest rates are very beneficial. Market probabilities for the December 10th rate cut have fallen from nearly 100% to below 50%. That helps explain, at least partially, the timing of the correction.
A bigger catalyst for the sell-off is a bit of a reality check on the AI trade. The amount of money that is being spent in the name of AI is eye-popping. The big four hyperscalers are projected to spend over $300 billion/year for the next 3 years! Yet, when CEOs of these companies are questioned on how and when investors can expect a Return on Investment (ROI), the answer seems to be a vague “it’s going to be huge!” Ok, like history’s great transformations, AI will be huge. But how huge? And when? After all, the capital expense is real, already huge, and growing. CapEx can’t grow forever without a return! Certain expense with uncertain ROI is not a great investment thesis.
In addition, the concept of “circularity” in AI is becoming popular. Today, Microsoft announced a new strategic partnership with Anthropic and Nvidia (NVDA.) Anthropic has committed to purchase $30 billion of Microsoft’s Azure computing power, Microsoft will invest $5 billion in Anthropic, and Nvidia will invest $10 billion in Anthropic. So, Anthropic will provide AI services to Microsoft, while Microsoft will provide the computing power run on Nvidia chips to Anthropic. Can you square the circle?
Similarly, Open AI has agreed to pay Oracle (ORCL) $60 billion over 5 years for computing power. Sounds straight forward. However, Open AI has never made a profit and is not projected to make a profit until 2028. Where is the $300 billion coming from? ORCL has yet to build the data centers and is already a debt ladened company. Where is the construction money going to come from? Lastly, according to JP Morgan, it will take 2.5 Hoover Dams or 4 full-scale nuclear power plants to provide the electricity for these data facilities. Currently, there are no dams or nuclear power plants under construction in the U.S. So, an unprofitable company has promised to purchase $300 billion in computing power with no power source from an indebted company. What could go wrong!
AI better be “huge” if these agreements are going to be profitable. As discussed above, history demonstrates that many first movers wind up being bad investments. What happens to these circular deals when one of the players fails. According to ORCL’s CEO, he is not concerned about Open AI’s ability to pay the $300 billion. If they can’t, there will be plenty of other buyers! Let’s hope! But if one of the leading AI model producers has failed, it is not clear there will be others to step in!
Are we in an AI bubble? I don’t think so. Unlike the dot.com bubble, most of the hyperscalers are highly profitable companies that will provide some ballast to the AI trade. That said, investors should be careful when investing in big transformations. While it is probable that AI is the next “huge” thing, not every investment will win! The AI trade is currently a Field of Dreams “if you build it, they will come” scenario. AI is undergoing a reality check. The Cap Ex is real and already huge. The ROI, while projected to be huge, is highly uncertain and the competition is and will continue to be fierce. History is clear that great innovations take years and decades to reach full potential. For long-term investors, that doesn’t mean the AI trade is over. As always, patience, diversification and strong risk management will be required. The market is finally asking for a little more than “if you build it, they will come.” And unlike the AI trade, the Field of Dreams had electricity to turn on the lights on!