Enterprises pull back on AI spending as token bills outpace productivity gains

Companies that spent the past year pushing employees to use AI tools as aggressively as possible are now struggling to manage the costs.  

CFOs are now demanding to see measurable returns on the ever-increasing API bills, threatening growth projections at OpenAI, Anthropic, and other large language model providers.  

Why are companies suddenly cutting back on AI spending? 

Companies are now dialing back their AI spending as CFOs demand justification for ballooning API bills. This reversal marks the end of what the industry has dubbed “tokenmaxxing,” and the correction is hitting fast.

Amazon recently dismantled an internal leaderboard that tracked employee AI usage after leadership concluded the system was producing more AI-powered busywork than useful output. “Please don’t use AI just for the sake of using AI,” an Amazon SVP told staff. 

Uber burned through its entire 2026 AI coding budget in four months, and Meta sent an internal memo to roughly 6,000 employees flagging what it called an “exponential increase” in AI usage, warning the company faced billions in internal AI costs. Uber has since imposed a $1,500 monthly spending cap per employee on AI coding tools.

Consulting giant Accenture previously warned employees they could “risk losing out on promotions” if they failed to adopt AI tools. Now, Accenture is trying to stop staff from using AI on trivial tasks. 

Leaked audio from an internal meeting captured an Accenture executive saying that AI spending is “becoming very unpredictable.” The same executive said that leadership at the “CFO, COO, and CIO level are still asking the question of whether they’re getting value from what we’re spending.”

International Business Machine’s (IBM) Adam McDaniel and Markus Eisele argued in a recent analysis that token minimization is just as bad as tokenmaxxing because both make token consumption the main goal rather than focusing on business outcomes. 

IBM advocates for what it calls “valuemaxxing,” which focuses on measuring completed tasks, time saved, and rework avoided rather than tokens consumed.

Is the reduction in AI spending already affecting companies?

OpenAI and Anthropic built their growth plans on the idea that enterprises would keep consuming more and more tokens. 

OpenAI crossed $25 billion in annualized revenue earlier this year, while placing its own valuation at $1 trillion, while Anthropic is valued a few billion dollars less. Both companies are burning through cash on compute, research, and hiring while hoping enterprise adoption will make them profitable. 

But enterprises are already reserving expensive flagship models for complex work and using smaller, cheaper alternatives for routine tasks. Some are moving workloads onto open-source models that run on their own infrastructure without per-token charges.

The International Data Corporation (IDC) predicts that by 2028, 70% of leading AI-driven enterprises will use multiple models rather than relying on a single provider. That would turn AI into a commodity where providers compete on price rather than just capability. 

The money thing is not going anywhere anytime soon, though. Even OpenAI’s CEO, Sam Altman, has acknowledged that the cost of AI has become a “huge issue” for customers this year.

The smartest crypto minds already read our newsletter. Want in? Join them.