Photo by Kido Dong

It seems to have gotten relatively little attention here, but China’s AI stocks have taken a sharp dive over the last four months. Reuters put the plunge at 40% in the STAR50 index, which includes most of the country’s AI and tech-related companies. The article said this drop corresponded to a loss of $480 billion in market wealth.

There are several important points about this plunge that should provide lessons for investors in the United States. The first and most obvious lesson is that sharp plunges in stock prices like this can happen. That shouldn’t really be news. From its peak in March of 2000 to its trough two and a half years later, the S&P 500 lost almost half of its value. The NASDAQ lost almost 80% of its value. Stocks can move in both directions, and when they are seriously overvalued, they can fall a long way.

Many people investing in AI-related companies have no memories of the tech bubble and its collapse. So, China’s plunge should remind everyone that bursting stock bubbles are not just ancient history. And yes, it can happen here.

It’s also worth noting that there does not appear to have been any obvious triggering event. The same applies to the beginning of the collapse of the tech bubble in March of 2000. I didn’t see anything at the time to send stock prices spiraling downward, and even with a quarter century to look back at the situation, I still can’t see anything that might have triggered the plunge. Perhaps Alphabet’s loss of $500 billion in market value on Thursday, following reports of deteriorating cash flow, will be a triggering event, but we’ll have to wait on that one.

Another striking point about China’s crash is how little money is involved. None of us will ever see $480 billion, but this sum is trivial compared to the sums in play with the U.S. AI and tech companies. If this loss is equal to 40% of the June value, it implies China’s tech sector had a market capitalization of $1.2 trillion before the crash. By comparison, Nvidia alone has a market cap of more than $5 trillion. At its peak, a few weeks ago, it almost hit $6 trillion.

The point here is that the amount of capital going into U.S. tech and AI companies is more than an order of magnitude larger than what the Chinese companies are spending. Yet the models produced by their leading AI companies are neck and neck with the frontier models produced by Anthropic or OpenAI. This enormous gap in spending is reflected in the enormous gap in price of their AI.

That makes it hard to see how the U.S. AI companies win in the market. If the U.S. companies were selling cars for $50k a piece, and Chinese models were pretty much as good and selling for $5k, who is going to buy the U.S. cars? And it doesn’t really matter if the U.S. AI models have a small edge in performance.

Suppose the U.S. car has a maximum speed of 120 MPH, while the Chinese car maxes out at 110. Since almost no one is ever going to want to push their car that far, the higher speed really doesn’t matter. It’s the same story with AI. The overwhelming majority of uses will not require frontier capabilities. Businesses and people will go with the versions that save them money. This will either be the Chinese models or U.S. models selling at sharp discounts to be competitive.

A final takeaway is that China’s AI companies are continuing to move forward and produce new high-quality models despite the crash. This is a key point. The crazy stock prices that we see for AI-related companies are not needed for companies to make progress developing AI. A crash might slow the frenzy to build data centers everywhere, but it will not be the end of AI. With luck, it will get the companies and country to think more carefully about what they are doing. That would be a very good development.

This first ran on Dean Baker’s Beat the Press blog.