Four months. That's how long it took Uber to exhaust its entire annual AI tools budget, according to a TechCrunch report. By April 2026, the company's CTO had revealed the full-year allocation was gone. Uber has since capped individual spending at $1,500 per employee per month on agentic coding tools (AI assistants that write and edit code on your behalf), with Claudee Code](/tools/claude-code/) and Cursor named as the primary tools driving the overage.
How Uber Made This Problem for Itself
The budget blowout wasn't random. Uber had been actively pushing employees to maximize their AI usage, tracking consumption on internal leaderboards that ranked who was using these tools most heavily. The company gamified AI adoption - then got surprised when people fully adopted it.
COO Andrew Macdonald framed the awkward position honestly: "it's very hard to draw a line between AI usage and new consumer features." That's essentially an admission that the output from Claude Code and Cursor is real enough to be showing up in shipped product. Engineers aren't burning tokens on experiments. They're using these tools as part of their core workflow, at a volume nobody budgeted for.
The $1,500/month cap isn't punitive - that's $18,000/year per engineer if someone consistently hits the ceiling, which is substantial tool access. Exceptions can be granted with approval, and an internal dashboard tracks each employee's spend against the limit in real time.
The Broader Budget Reckoning
Uber's situation is playing out at companies across the industry. The "try everything and see" phase of enterprise AI adoption - unlimited expensing policies, exploratory usage, no guardrails - is ending. Finance teams that modeled cautious adoption rates are dealing with enthusiastic ones.
The harder problem: despite soaring usage numbers, concrete ROI from AI tooling remains difficult to isolate and quantify. Macdonald's comment captures it - when AI tools are woven into the development process, their contribution to any given feature becomes nearly impossible to separate out. That's a headache for every CFO being asked to justify expanding the AI tools line item after it already overran.
For freelancers and small businesses using Cursor or Claude Code independently, none of this directly affects your access. But it signals that large corporate clients are about to start managing AI tool procurement much more tightly, which will shape conversations about approved integrations and what tools they'll fund in vendor contracts.