Sam Altman said AI budgeting has recently become a “huge issue” for some companies, driving AI bubble watchers and doomsayers into a frenzy.
At a corporate event on Tuesday, Altman referenced memes such as “My company used its entire 2026 budget in the first quarter.”
“It went from an issue that wasn’t talked about at all at the beginning of the year, where people were completely satisfied with the amount they were spending, to suddenly become a big issue,” Altman said.
The reaction from readers was great. Some said it was a warning of the dark ages or a failure of the AI business model. The word “bubble” came up a lot. Others said this is to be expected and is a normal stage where people learn what to actually spend their tokens on after a period of experimentation.
Commentators from Gary Marcus to Michael Berry also participated. Here are some of the most interesting reactions.
Some say it’s a dark warning.
Ed Zitron, one of the Internet’s foremost AI bubble alarmists, wrote in X that OpenAI is “completely cooked.”
“These are the words of a loser,” Zitron wrote. “Four years into the bubble, we can’t say, ‘Sure, our customers have a big problem with the height of our business.’ We just raised $122 billion!”
Programmer Eric S. Raymond (also known as ESR) agreed: “The bubble is bursting.”
“Make no mistake about it, this is a very useful technology, and its adoption will continue and accelerate,” Raymond wrote. “However, the overinvestment we are seeing in data centers is not sustainable. The business models of large providers are unworkable and dependent on VC funding.”
Academic and author Vivek Wadhwa wrote that AI researcher Gary Marcus appears to have been right: “AI’s revenue model is collapsing.”
Marcus himself commented that the death of tokenmaxxing was “a potentially very serious problem for all three big IPOs.”
Big Short investor Michael Varley, who is skeptical of AI, also mentioned X’s article.
Some say it’s token abuse.
As the TokenMax craze dies down, some engineers are wondering, “Are we spending our money the right way?” Altman’s comments may be more a rational cost-benefit analysis than a grim warning.
“Extracting value from agents is still too difficult for most engineers, so they end up just burning tokens,” said Google software engineer Patrick Toulme.
“80% of LLM’s economic value comes from 20% of the tokens,” wrote Peter Berezin, chief strategist at BCA Research. “There is a long tail of questionable token usage that can be significantly reduced without significantly negatively impacting productivity.”
Kun Chen has worked at Meta, Microsoft, and Atlassian. He wrote that some reductions are “inevitable” because AI spending is “driven by FOMO.”
Still, he remained optimistic.
“We are bullish that real demand will slowly rise again,” Chen wrote.
Corey Quinn, chief cloud economist at Duckbill, took a more cynical approach. Altman is starting to realize that OpenAI’s tokens (the tokens he sells) can become expensive, he wrote.
“He was truly the Copernicus of his generation,” Quinn wrote.