Altman’s budget warning splits AI into two camps

AI budgeting – Sam Altman said AI spending has suddenly become a “huge issue” for some companies, sparking a wave of bubble-pop and token-misuse debates—ranging from claims that AI business models are imploding to arguments that teams are simply learning how to spend effecti

Sam Altman dropped a line about AI costs that landed like a flare. During a Tuesday enterprise event, he described how quickly budgets can change once companies try to turn AI experimentation into something that looks like a business.

“My company spent my entire 2026 budget in Q1.” That meme—followed by the reaction Altman said he’s hearing—wasn’t just internet humor. Altman framed it as a shift in how customers feel about spending: “That went from. at the beginning of this year. an issue that never came up — people were totally happy with the amount they were spending — to all of a sudden. a huge issue. ” he said.

The immediate reaction was loud and emotional, with the word “bubble” showing up again and again. Some commentators treated Altman’s remark like a warning that dark times are coming. or that the AI business model is failing. Others treated it like the messy. predictable phase in which teams discover what they’re actually getting for the tokens they’re burning—and adjust from there.

Ed Zitron, one of the internet’s most prominent “AI bubble” voices, put it bluntly on X. “OpenAI was absolutely cooked.” He added: “You can’t be four years into the bubble saying ‘yeah our customers have a huge issue with how expensive our business is.’ You just raised $122 billion!”

Eric S. Raymond—often referred to as ESR—agreed with the diagnosis that “the bubble is popping. ” while trying to separate that view from outright pessimism about the technology. “Make no mistake, it’s a hugely useful technology and uptake will continue, even accelerate,” he wrote. “But the overinvestment in datacenters that we’ve been seeing is not sustainable; the business model of the big providers doesn’t work. and is floating on VC money.”.

Vivek Wadhwa, an academic and author, weighed in by pointing to the argument that some AI revenue models are breaking down. “It seemed like AI researcher Gary Marcus was right: ‘the AI revenue models are imploding,’” Wadhwa wrote.

Marcus, for his part, tied the budgeting tension to market timing. He said the death of tokenmaxxing was “potentially a very serious issue for all three big IPOs.” Michael Burry—the “Big Short” investor who has taken a more AI-skeptical turn—also referenced the story on X.

Not everyone who jumped in thought the story was about collapse. As tokenmaxxing fever faded, some engineers focused on a more practical question: are teams spending in the right ways?

Google software engineer Patrick Toulme argued that the problem often isn’t AI itself, but the difficulty of extracting value from it. “Getting value out of agents is still too difficult for most engineers, so they end up just burning tokens.”

Peter Berezin, BCA Research’s chief strategist, offered a different kind of critique—less about doom and more about waste. “80% of the economic value of LLMs come from 20% of the tokens,” he wrote. “There is a long tail of dubious token usage that can be greatly curtailed without much negative impact on productivity.”.

Kun Chen, who has worked at Meta, Microsoft, and Atlassian, connected overspending to emotion rather than economics. He said AI spending was “driven by FOMO,” so some cutback was “inevitable.” Even with that warning, he stayed optimistic: “I’m bullish that real demand will slowly build up again.”

Corey Quinn, Duckbill’s chief cloud economist, chose irony instead of alarm. In his view. Altman may be arriving at a cost lesson that many customers already learned: “Altman was starting to realize that OpenAI’s tokens — the tokens he sells — can be expensive.” Quinn then wrote: “Truly the Copernicus of his generation.”.

What ties all sides together is that Altman’s “huge issue” line didn’t just describe an abstract market trend. It mapped onto a lived experience: budgets that were stable early in the year. and then suddenly weren’t—big enough. in at least one retelling. to turn a full-year allocation into a single quarter.

In the end, the debate wasn’t only about whether AI spending is too high. It was about what that spending means. For the bubble-pop crowd, the sudden pain signals a model that can’t hold. For the token-efficiency camp. it signals learning—teams finding where value actually comes from. and cutting away what doesn’t pay its way.

Sam Altman AI costs tokenmaxxing OpenAI AI bubble enterprise event AI budgeting datacenters VC money token misuse LLMs