Sam Altman Says AI Budgets Have Become a "Huge Issue" for Companies

The companies racing to adopt AI have a new problem, and Sam Altman is hearing about it directly: the bills.

Speaking at OpenAI's Intelligence at Work event, the CEO said cost has gone from something clients never raised to one of their most common complaints in a matter of months. "People are really saying, you know, it's kind of a meme now, but 'My company spent my entire 2026 budget in Q1. Can you make this more efficient?'" Altman said on stage, as reported by Tom's Hardware.

He described the shift as sudden. The concern went from "an issue that never came up" at the start of the year, when "people were totally happy with the amount they were spending," to "all of a sudden, a huge issue," Altman said. According to Axios, he ranked cost as the second most common issue he hears from customers, behind only simplifying AI workflows.

The token problem

The pressure point is the token, the basic unit of AI computing that every prompt consumes. As companies rushed to prove they were AI-ready, usage exploded and so did the invoices. According to a Wall Street Journal report cited by IBTimes, some companies have burned through annual AI budgets in just a few months, while others have seen costs double or triple.

Altman put the scale of consumption in context. The top token user at OpenAI now goes through roughly 100 billion tokens a month, he said on a livestream about enterprise adoption, per Axios. Six and a half years ago, he noted, the company's heaviest user consumed about 100,000 tokens a month, a figure that is now roughly the global average per person. To his own embarrassment, Altman added, OpenAI's internal leader is not even the biggest spender in the world.

Some of that consumption has tipped into excess. The trend dubbed "tokenmaxxing" has produced internal leaderboards that reward employees for heavy usage, and Tom's Hardware reported that some Amazon staff admitted to running AI agents on unnecessary tasks just to climb those rankings. In one extreme case, OpenClaw creator Peter Steinberger's team ran through $1.3 million in OpenAI API tokens in a single month, totaling 603 billion tokens, according to Tom's Hardware.

Companies are pumping the brakes

The backlash is already underway. Some of the largest companies have started rationing access to AI tools, tracking usage more aggressively, and steering employees toward cheaper options as costs rise faster than expected, IBTimes reported.

Uber is the clearest example of the squeeze. The company burned through its entire 2026 AI coding budget in the first four months of the year, CTO Praveen Neppalli Naga told The Information in April, leaving him "back to the drawing board" on what the budget needed to be (as we covered here). Before the company stepped in, individual engineers were running up monthly bills of $500 to $2,000 in token consumption, and roughly 95% of Uber engineers were using AI tools every month. In response, Uber set a $1,500 monthly cap per employee for each agentic coding tool, including Anthropic's Claude Code and Cursor, and gave staff a dashboard to track real-time spending, Bloomberg reported. Uber's own leadership has questioned the payoff: COO Andrew Macdonald said on the Rapid Response podcast that the link between rising Claude Code use and better products "is not there yet."

Microsoft has reportedly cut back on Claude Code licenses over cost as well, and Walmart capped access to its internal assistant, Code Puppy, after usage far outran expectations.

The math is forcing a more selective approach. Matan Grinberg, CEO of coding automator Factory, summed up the new mindset to the Wall Street Journal: "If your daughter needs tutoring in algebra, you can probably find someone cheaper than Albert Einstein." That logic is pushing firms to triage which tasks actually need a frontier model, with some building homegrown tools to cut their dependence on outside providers.

Altman's pitch: more value, more usage

Altman framed the moment as a problem OpenAI intends to solve rather than a ceiling on demand. "We are continuing to push on that more with models. I think we'll have a lot of ways we can help people get more value for less spend," he said, per Tom's Hardware, adding that he wants customers to "use AI and never worry about it being great and affordable."

He is still betting usage climbs from here. Altman pointed to "constant running proactive AI," or autonomous systems that act without being prompted, as the next phase companies should prepare for, according to Axios. That shift would reshape how firms plan around AI usage and security, but it also threatens to push token consumption higher, since agents can rack up costs quickly. Goldman Sachs has estimated that agents could increase token demand by as much as 24 times, Tom's Hardware reported.

The tension underneath

The optimism collides with a stubborn dynamic. Even as OpenAI, Anthropic, and Google all offer cheaper versions of their flagship models, agentic systems are consuming tokens faster than the labs are driving down the per-token price. Some companies are now finding it more expensive to run AI models than to hire people for the same work, Tom's Hardware noted.

That is the bind behind Altman's "huge issue." The industry's growth case assumes prices fall faster than usage rises. For the finance chiefs who watched a Q1 experiment swallow a full year's budget, that assumption is no longer theoretical.