DeepSeek undercuts belief that chip-hungry US players will win AI race

Anything vaguely related to artificial intelligence was smashed on Monday after investors spent the weekend frantically googling DeepSeek-R1, the low-cost Chinese AI model released last week.

The new generative-AI model, which claims performance on a comparison with OpenAI’s latest, stuck a knife into two beliefs that have come to dominate investment in the past two years. First, that AI needs massive amounts of new infrastructure, energy and microchips, mostly from Nvidia. And second, that the winners of AI will be the dominant American technology companies.

Monday’s AI crash was led by the first part, the “chips and shovels" that supported the development of AI, not the developers of these fancy programs themselves. Nvidia stock was down 17%, losing more than half a trillion dollars of value, nuclear-power stocks Constellation Energy, Vistra, Oklo and NuScale Power were down 21%-28%, and data-center supplier Vertiv Holdings was off 30%.

“This is symptomatic of more AI supply coming at a point where AI’s still looking for a problem to solve," said Robert Almeida, global investment strategist at MFS Investment Management.

Investors should have some questions.

How much was the selloff about fundamentals, and how much about sentiment? The moves in prices appear to show investors focused on fundamental issues of how DeepSeek’s approach will lead to lower power use and less demand for chips and data centers.

Yet, it is hard to believe that prices had run up so much purely on the back of smart investors plugging growth estimates into their spreadsheets and valuing the resulting cash flows. A lot of what’s been going on is similar to when investors discovered the internet. They have grasped that AI is A Big Deal, but can’t yet see exactly how or when it will make money.

In a sentiment-driven market, it is even harder to work out what happens next. I thought the market was overly frothy in mid-December, because prices seemed too far detached from reality. The trouble is that sentiment is hard to predict: Investors can always become even more excited about something, but sentiment becomes more vulnerable to a setback the more enthusiastic investors are. DeepSeek may be just that setback.

What about the AI companies? What are the prospects for Microsoft-backed OpenAI, Alphabet, Amazon-backed Anthropic, Elon Musk’s xAI and all the others to make money when faced with a low-cost competitor? DeepSeek is, after all, available, like Meta’s Llama, to download and tweak free.

On Monday this wasn’t the main concern, with Microsoft off 2.1%, Alphabet 4.2% and Amazon slightly up. They have big, profitable businesses they are using to finance AI development, and will also be able to use the techniques DeepSeek shared to lower their own costs. But they just lost one of the biggest barriers to entry. If a new AI model can be produced for just a few million dollars by the tech arm of a Chinese hedge fund, maybe others can do the same.

More competition will make it hard for Big Tech to make the oligopoly-like profit margins that investors hope for. If the companies can’t make fat profits, it will be even harder to justify their high valuations. These valuations, remember, rely on the assumption that AI tools will be both widely used and highly profitable, but even the experts have little explanation of how the business model will work. It will also be harder to explain why they are sinking so much money into AI data centers.

Is this just another bump in the road? As the dot-com bubble was building in 1999-2000 there were large price corrections, but they merely attracted people who had missed out to come in and buy what they thought were bargains. On Jan. 4, 2000, the Nasdaq-100 index tumbled 6.5%, double Monday’s fall, yet the tech bubble itself didn’t burst until March. Even then, the S&P 500 almost completely recovered by September, when a brutal selloff began.

What will the economic impact be? Even quite significant market corrections typically have little to no effect on growth. This time, though, a change of market signal might be under way that would affect on-the-ground activity, not just Wall Street.

Investors have been encouraging companies to pour cash into building new data centers, power stations and anything related to the power grid for the past year. Monday’s price drops suggest less appetite for real-world investments and could persuade companies to invest less into such projects. That might be part of the reason that 10-year Treasury yields fell so much on Monday, though it was also a flight to safety by investors.

Will investors now remember the benefits of diversifying their investments? At the start of the year the 10 largest stocks in the S&P 500 made up 37% of the index, and only one, Berkshire Hathaway, was outside Big Tech. The point of buying the S&P or other large market indexes is to gain exposure to a range of stocks with low fees, but the concentration created by the sheer size of the market behemoths mean the market offers less diversification of risk than any time since index funds were invented.

Monday showed the problem: The Magnificent Seven stocks of Apple, Alphabet, Amazon, Meta, Microsoft, Nvidia and Tesla were down over 3%, weighted by market value. Yet the rest of the S&P 500 was down only 0.4% on the same basis—while the average stock was flat.

Diversification means underperformance when the big get bigger, but offers protection when investors get overenthused by an idea and push prices too far—as seems to have happened with AI.

Write to James Mackintosh at james.mackintosh@wsj.com