Simon Willison — Uber Caps Usage of AI Tools Like Claude Code¶
Source: Simon Willison's Blog
Uber blew through its entire 2026 AI budget in just 4 months. The company's new response: cap all employees at $1,500 per month per AI coding tool (Cursor, Claude Code).
The Numbers¶
Simon Willison breaks down whether this is reasonable or draconian:
- At $1,500/tool × 2 tools = $36,000/year per engineer
- Median Uber engineer compensation: ~$330,000/year
- AI tooling costs represent ~11% of total comp — significant but not absurd given productivity multipliers
Willison's own AI coding tool usage runs ~$1,000/month, though subsidized plans cost him just $100/month personally.
A Rational Response¶
Willison characterizes the cap not as anti-AI sentiment but as a rational response to an unbudgeted variable cost. Enterprises budget annually. When AI tool spend grows 10× faster than anticipated, something has to give.
Key dynamics at play:
- Discoverability — engineers are still learning what AI can do for them, each new use case adds to the baseline.
- Habit formation — once an engineer builds a workflow around AI code generation, revocation feels like a productivity loss.
Broader Lesson¶
The Uber cap reveals something important: tokenmaxxing is real. When organizations first adopt AI coding tools, usage explodes as engineers explore boundaries. The bill arrives month 4.
For the broader industry: AI tooling needs predictable pricing models — enterprise seats, consumption caps, or committed-use discounts — before it can be treated as infrastructure rather than an experiment.