What does it say about the prospects for an artificial intelligence “golden age” that some of those who most enthusiastically predicted it are now tamping down expectations? Earlier this week, in what looked like damage control over the release of a new version of ChatGPT, OpenAI Chief Executive Officer Sam Altman said investors have been inflating a speculative bubble in AI. He predicted “someone’s going to get burned.” Altman and other AI insiders seem more or less fine with that, arguing that asset bubbles often coincide with technology breakthroughs. The thinking goes like this: the dot-com bust was bad, but at the end of it we had a new information infrastructure that led to lasting economic growth. Maybe that happens this time around, but there’s reason to think an AI bust would be economically devastating—and not just for businesses that bet heavily on the software and data centers needed to run it. On this week’s episode of Everybody’s Business, we explore the potential economic fallout of an AI implosion with Ed Zitron, a skeptic who makes a compelling case for panic in a recent essay and on his podcast, Better Offline. In other words, rather than continuing to embrace the new technology, maybe it’s time to hate it. The argument boils down to a problem of misalignment: For years, big tech companies have dumped hundreds of billions of dollars into developing ever-more advanced large language models (LLM) (like OpenAI’s GPT, Google’s Gemini and Anthropic’s Claude). Much of that money has gone to chipmakers, especially Nvidia, which sells the graphical processing units needed to train new models. All of this spending has sent asset prices soaring, creating a dynamic in which index funds are heavily weighted to a single industry—which you guessed it—threatens to crash the entire stock market if it falters. A similar dynamic may be playing out in the world of private credit, which tech companies are increasingly tapping to build data centers, creating another economic risk.At the same time, there are signs those wildly expensive LLMs are failing to generate commensurate financial returns. These include the blowback to the release GPT-5, which OpenAI had promoted as potentially god-like but which many users say is actually worse than the last version. There’s also a recently published study from the Massachusetts Institute of Technology that showed the vast majority of pilot programs involving so-called generative AI failed to lead to revenue growth. Also on this week’s episode: We discuss the struggles of fast food companies with Bloomberg Businessweek’s Deena Shanker. Normally, they thrive in times of economic uncertainty, but McDonald’s has been struggling, partly she says because the company’s food is no longer seen as cheap. One exception to this trend? Taco Bell, which has thrived because it’s managed to keep prices low while being just unique enough to go viral on TikTok. Finally, for our underrated story, we preview Fed Chair Jerome Powell’s big speech in Jackson Hole, Wyoming, on Friday. Get ready for some extremely low-key fireworks. See omnystudio.com/listener for privacy information.