ChatGPT is more popular than ever, but is the AI bubble about to pop?

It’s been a big couple weeks for OpenAI. The most valuable startup in the world recently announced that ChatGPT will become more like an operating system, released its first social networking app, and even spread rumors about launching a device designed to make us happy.

There is some tricky accounting along the way, of course. These big product announcements happened after chipmaker Nvidia, the most valuable company in the world, invested $100 billion in OpenAI to build more data centers, which OpenAI will fill with Nvidia chips. OpenAI then made a deal with AMD, Nvidia’s rival, to build even more data centers and then fill them with AMD chips. Some analysts call these kinds of deals “circular,” since one company is investing money in another company that gives some of it right back. Others call it “bubble-like behavior.”

All things told, OpenAI has inked $1 trillion worth of computing deals this year alone. That staggering amount of money will help you do things like shop for houses on Zillow without leaving ChatGPT, star in your own AI-generated sitcom, and carry around an artificially intelligent surveillance device in your pocket. A trillion dollars is also a very silly sum when you acknowledge that OpenAI has never turned a profit and reportedly expects its losses to triple to $14 billion in 2026. And yet OpenAI’s valuation climbed to $500 billion last week. (Disclosure: Vox Media is one of several publishers that have signed partnership agreements with OpenAI. Our reporting remains editorially independent.)

Math like this is what’s got more and more people talking about the AI bubble and its imminent popping. On Wednesday, the Bank of England cautioned that the risk of a “sudden correction” to global markets is growing as the valuations of top AI companies increase. The same day, IMF managing director Kristalina Georgieva issued a similar warning and said tech company valuations “are heading toward levels we saw during the bullishness about the internet 25 years ago.” The Nasdaq index reached a peak on March 10, 2000, before imploding. The Nasdaq closed at an all-time high on October 6.

Bubble-like behavior, briefly explained

The idea that we’re looking at another tech bubble is not new. It’s been around for at least a decade, and people were wringing their hands about AI hype even before ChatGPT stunned the world with its popularity. But now the stakes are higher than they’ve ever been, as investment in the AI industry has expanded into every corner of the economy. The AI boom is no longer just a Silicon Valley venture, as building the data centers needed to power apps like ChatGPT relies on the real estate, construction, and even air conditioning industries. Then there’s the chip industry, which is ultimately reliant on a single company in Taiwan to manufacture the most advanced semiconductors for AI. Everyone seems to believe that the AI bet is too big to pass up.

The AI hype is so powerful, it’s lifting up the rest of the economy. It’s also obscuring a lot of bad economic news in the United States, including inflation, stagnated growth, and a dreadful job market for young people, which the rise of AI probably contributed to. If the AI boom has indeed become an AI bubble and that bubble bursts, the shock wave would hit everything.

That’s a scary thought. It’s also increasingly looking like a very real possibility. The circular dealmaking is just one red flag and it’s not just OpenAI. Elon Musk’s xAI recently raised $20 billion, some of which came from Nvidia, in order to buy Nvidia chips.

Another red flag is the simple fact that we don’t know if these big bets on AI will pay off. AI companies expect demand for their products to keep growing, which is why they’re investing so much in infrastructure so they can meet that demand if and when it comes. But everything is speculative. The hundreds of billions of dollars being thrown at data centers is reminiscent of the huge investment in internet infrastructure back in the 1990s. Eventually, however, the supply of fiber optic cables outpaced demand, and the telecom industry crashed.

The most salient signs of danger, however, are what you’ve probably experienced yourself: the bad vibes. Americans in general are pessimistic about AI and have only grown more concerned about the technology since ChatGPT’s launch. We don’t really know how AI will make our lives better. Sure, ChatGPT is popular — OpenAI says it has 700 million weekly active users — but it’s far from clear if it will become our new operating system or the new front door to the internet. AI’s ability to boost productivity is so far unproven, too. An MIT study released last month found that 95 percent of organizations surveyed found zero return from their AI initiatives.

It’s certainly possible the AI boom remains just that, and we all ride into the future with virtual assistants in our ears and data centers in our back yards. It’s also possible that the bubble-callers are correct, and that we’re about to relive not only the dotcom crash of the early 2000s but also the aftermath of Railroad Mania in the 1840s. In both of those periods, companies collapsed, and lives were ruined. The infrastructure survived, however. Victorian England ended up with a railway system, and Silicon Valley got tubes to run the internet on. Eventually we figured out how to make it all work.

A version of this story was also published in the User Friendly newsletter. Sign up here so you don’t miss the next one!