Still from Steven Spielberg’s “AI.”
I have written before on the AI bubble. The stock valuations we are seeing now look a lot like the valuations we saw before the Internet bubble began to crash in 2000. Even a quarter century later, people miss much of the story of that bubble.
It’s true we had companies that never made a profit, like Pets.com, with valuations in the billions. But the story went far beyond some flaky companies having ridiculous market capitalizations. The irrational exuberance, to use Alan Greenspan’s great term, infected everything.
Ford’s stock price more than tripled from 1996 to its peak in 2000. It then fell by more than 70 percent at its trough in 2003. McDonald’s, which is about as far from the Internet as you can get, had its stock price rise by almost 80 percent from 1996 to its bubble peak. It also lost roughly 70 percent of its value over the next three years.
But the best model for some of the AI giants may be a company like Cisco Systems. Unlike the flaky startups, Cisco Systems was a real company with a useful product. It makes much of the hardware that provides the backbone for the Internet. But its stock price nonetheless went through a crazy boom and bust.
From the start of 1996 to its peak in 2000, Cisco’s stock price went up more than 2000 percent. It briefly had the largest market capitalization of any company in the world, at more than $300 billion, which would be the equivalent of $1 trillion in today’s economy. It then lost more than 85 percent of its value over the next three years.
Cisco is still around and a profitable company. It has a market capitalization of $350 billion, which is considerable, but nowhere near the multi-trillion valuations of the giants like Nvidia or Alphabet. The surge and collapse of Cisco’s stock price could be a model for current AI leaders.
The AI Race
I have always been skeptical about how much money the AI folks would be able to pocket for themselves. Remember, the issue here is not how useful AI is or will end up being. The question is how much of the benefits (or harms) from AI that Nvidia, Alphabet, Meta, and the rest can capture for themselves.
Here, the competition from China is a very big deal. This is not just a question of which country at the end of the day ends up having better or more efficient AI; the issue is that the Chinese AI companies provide serious price competition for US models. This will limit the extent to which US companies can make huge bucks on their products.
At the moment, it looks like the cutting-edge Chinese AI company, Deep Seek, is coming out far better on price than the US leaders, OpenAI and Anthropic.
I realize that there are other measures of AI performance and efficiency than price per token, but this is clearly an important one, and it looks like China’s leading company is running away from the field. (Deep Seek has a V-4-Flash version with far cheaper prices, $0.14 per million input tokens and $0.28 per million output tokens. The leading U.S. producers also have much cheaper scaled-down versions.)
Deep Seek also has the advantage that it is an open-source system, which means that companies can alter the models and run them on their own computers rather than loading data onto the cloud. This means they don’t have to worry about losing control of proprietary information. By some accounts, usage of Chinese AI already vastly surpasses usage of AI from US companies.
Can Trillions of Dollars on Investor Money be Wrong?
Economists, myself included, like to think the markets are generally right in their assessment of the world. But in my lifetime, I have seen two cases, the tech bubble in the 1990s and the housing bubble in the 00s, where they were badly wrong.
I know that people who control big bucks can sometimes be big buffoons. (We all know the old saying, “if you’re so rich, how come you’re not smart?”) In the 90s, they were throwing huge bucks at hare-brained Internet entrepreneurs. In the 00s, they were insisting that house prices never fall.
The current crew doesn’t inspire confidence. Mark Zuckerberg threw $80 billion in the garbage pushing Meta. Elon Musk runs around pushing crazy conspiracies about 20 million dead people getting Social Security and millions of undocumented immigrants voting in elections in between the racist rants he posts on X.
We should always be checking and double-checking our assumptions and views of the world. I definitely need to do more homework on AI economics, but for now, it’s hard for me to see how the big money can be right.
This first appeared on Dean Baker’s Beat the Press blog.
