Uber, Walmart Impose AI Token Caps as Enterprise Costs Surge

The era of uninhibited AI spending is over at several major enterprises, and the wake-up call did not come from runaway engineering teams. It came from people converting PDFs into slides.

According to 404 Media, leaked audio from inside Accenture captures Justice Kwak, the company's agentic AI strategy lead, acknowledging a pattern now appearing across large organizations: "It's actually not our engineers that are driving the token consumption. It's a lot of the non-engineers." The culprit is mundane: office workers running documents through AI tools for routine reformatting, at a scale apparently large enough to threaten budgets.

The consequences are already concrete. Uber reportedly capped employee use of AI tools including Claude Code and Cursor after the company's CTO said the firm had blown its entire AI budget in four months. The imposed limit was $1,500 per month per engineer across Claude Code, Cursor, and the GitHub Copilot CLI. Walmart imposed similar restrictions on tokens through its internal AI coding platform, Code Puppy, after adoption surged well past expectations.

Adding structural pressure to the trend: GitHub Copilot moved to usage-based pricing, ending the flat-subscription model large engineering teams had relied on. Accenture's response is to build a product around the problem. Kwak says the company plans to formally launch something called Token IQ, aimed at giving leadership visibility into where AI spending is going and whether it generates adequate return. That pitch, AI FinOps for the enterprise, is likely to attract interest from other large organizations facing the same scramble.

The honest caveat is that 404 Media's reporting relies on leaked audio, so Accenture's characterization represents one internal perspective at one point in time. What the reporting does not give you is a clear picture of whether the cap strategies at Uber and Walmart are producing the expected cost reductions, or whether limiting access creates its own productivity drag. A $1,500 monthly cap is a concrete number, but the ROI calculus behind it is not yet visible.

The broader opportunity belongs to whoever can make AI spending legible, showing organizations not just what they spent but what they got for it. Token IQ is one early entrant in that space. If the pattern holds, a whole category of AI cost-management tooling is about to find a much larger addressable market.