The era of blank-check artificial intelligence (AI) spending has hit a sudden, unexpected wall.
In a series of startling revelations at a Tuesday enterprise event, OpenAI CEO Sam Altman warned that the tech industry is experiencing a severe pullback on AI spending as corporate clients aggressively implement cost-cutting measures to restrain unsustainable buildout costs.
According to Altman, the dramatic shift in corporate sentiment has caught the AI sector by surprise. He noted that at the beginning of 2026, AI budgeting “never came up” and clients were entirely comfortable with their investments. Just months later, however, balancing the AI budget has suddenly transformed into a “huge issue” threatening profitability for hyperscalers and end-users alike.
This is especially troubling for OpenAI and Anthropic, both of whom expect to go public this year and will soon face pressure from Wall Street and investors to show profits.
Financial anxiety gripping the broader market stands in stark contrast to the internal culture at OpenAI, where employees heavily consume their own product. Altman revealed that six and a half years ago, OpenAI’s top “token burner” used 100,000 tokens a month — a figure that matches today’s global per capita average. Today, OpenAI’s top internal user burns an astonishing 100 billion tokens monthly.
Anthropic President Daniela Amodei said the steep cost of creating AI models is forcing companies like hers to seek capital from public markets. Anthropic recently raised $65 billion in its latest funding round as it preps for an IPO. It filed the paperwork this week.
Other instances of extreme usage have emerged. Peter Steinberger, creator of OpenClaw, once racked up a $1.3 million bill by burning 603 billion tokens in 30 days. The New York Times reported an OpenAI staffer consuming 210 billion tokens in a single week.
Altman admitted OpenAI discovered an external user who outpaced everyone inside the company, calling the finding a personal “embarrassment.”
This high-rolling consumption is actively gamified within OpenAI via internal leaderboards, with employees frequently boasting about their usage on social media.
Outside the AI labs, the financial reality of this consumption has triggered immediate executive pushback. Major enterprises including Amazon.com Inc., Uber Technologies Inc., Walmart Inc., and Microsoft Corp. are actively rethinking their AI integration strategies due to surging token usage.
Uber has reportedly enforced strict token caps after its chief operations officer noted that the spiraling costs were becoming impossible to justify; Amazon went as far as shutting down its own internal token leaderboard.
Andy Sen, chief technology officer at AppDirect, said there is “an important balance to strike to manage AI spend properly, while still enabling an AI-forward work environment.”
“A lot of companies are turning on AI features everywhere, riding the productivity wave until they are blindsided by a huge bill,” he said. “So how do companies manage their AI costs without turning it off entirely? The key is understanding the difference in cost between some of these models, because there are some that can be 100 times more expensive than others.”
PolyAI CEO Nikola Mrkšić argues the pilot stage of AI spending is over, and corporations are starting to demand ROI on the massive budgets they’ve allotted to new AI technologies. Specifically, he believes companies should prioritize AI architectural efficiency over brute force. For example, PolyAI uses lean, domain-specific models built from the ground up to deliver efficiency and savings to its customers.
Acknowledging a popular industry meme — “My company spent my entire 2026 budget in Q1, can you make this more efficient?” — Altman conceded that OpenAI must adapt. He assured investors and clients that the company is aggressively pushing its models to deliver “more value for less spend” as the industry faces its first true economic reality check.
“Many companies are starting to see that AI can’t replace skill or experience. Even when skilled people use AI to augment their work, generating consistent, quality results still takes an extraordinary amount of tokens, and when the bill comes due, can they actually pay for it?” said tech strategist Jennifer Jacobson. “Was it worth it? Did they get good results? If the answer is no, these companies will have to limit their use if AI, or phase it out.”
