Ed Zitron has been calling the AI bubble since March 2024, and he's still at it. His newsletter, with 80,000 subscribers, has predicted the bubble's burst repeatedly. But Kelsey Piper at The Argument argues his case has actually gotten weaker as the evidence against it has grown.
The shift is telling. In 2024, Zitron made a straightforward economic argument: companies weren't really using AI, and models like GPT-4 showed promise but lacked clear utility. That was a reasonable position. By 2026, though, he's comparing OpenAI and Anthropic to Enron and FTX, alleging outright fraud. Piper sees this escalation as an admission that the economic case no longer holds.
She has a point. The numbers have moved fast. GPT-4-level intelligence now costs roughly 1/1000th of what it did at launch. API costs dropped from $0.03 per 1K tokens to around $0.0025. Hardware got better. Algorithmic tricks like Mixture of Experts architectures and quantization slashed compute requirements. Open-source models like Llama-3 now offer competitive performance for a fraction of the price. This isn't speculative. It's the economics on the ground.
And people are using it. About 30% of Fortune 500 companies have enterprise deals with leading AI startups. More than half of Americans use AI chatbots weekly or more often. Piper notes that Zitron's 2024 argument relied heavily on earnings calls where companies admitted AI wasn't impacting revenue yet. Those admissions are harder to find now.
The challenge for critics is that you can't just keep saying "bubble" when adoption is growing and costs are falling. You need a new argument. Zitron tried to build one around fraud, claiming OpenAI is inflating revenue numbers by counting free tier usage. But as Piper points out, that would be straightforwardly illegal if true, and he offers no evidence.
Commenters on the story express fatigue with a certain style of AI criticism, grouping Zitron with Gary Marcus and Emily Bender as commentators who seem to repeat the same arguments regardless of what actually changes in the technology. That's not entirely fair. Skepticism is valuable, especially in a space awash with hype. But skeptics have to update when the facts change. When something gets 1000x cheaper in two years, the use cases that were marginal become obvious. Zitron didn't update. He just yelled louder.