Uber exhausted its entire annual AI coding tools budget in just four months, its CTO disclosed in April. The company has since instituted a $1,500 monthly cap per employee per agentic coding tool.

The overspend followed an internal push to maximize adoption, including a leaderboard that ranked teams by total AI tool usage. The numbers it produced were striking:

• 95% of engineers using AI tools every month
• 70% of committed code is AI-generated
• $500 to $2,000 monthly API cost per engineer
• 1,800 code changes per week written entirely by AI without human input

Each employee now has an internal dashboard to track usage. Exceptions to the cap require approval.

Uber reported that AI generated at least 10% of its code. But COO Andrew Macdonald was direct about what that number doesn't prove:

‍

"That link is not there yet. Maybe implicitly there's more that is getting shipped, but it's very hard to draw a line between one of those stats and 'Okay now we're actually producing like 25% more useful consumer features.'" — Andrew Macdonald, COO, Uber, on the Rapid Response podcast

‍

Uber's R&D spend hit $951 million in Q1 2026 alone, up nearly 17% year-over-year.

Uber is not the only one pulling back

Walmart has also capped how much its employees can use the company's in-house AI agent. 

Microsoft recently canceled most internal Claude Code licenses, in what AI Weekly called "the clearest enterprise-scale AI spending pullback so far in 2026."

Gartner projects that inference costs will drop 90% by 2030 from 2025 levels, but cheaper tokens won't translate to cheaper enterprise AI. 

Agentic models require far more tokens per task than standard models, and providers are unlikely to fully pass savings to customers. Anthropic has already shifted from flat-fee pricing to a usage-based model.

Gartner forecasts AI agent software spending will reach $207 billion in 2026, up 139% from $86.4 billion in 2025. At that growth rate, the gap between what enterprises spend and what they can actually see, track, and justify is only widening.

Why it matters

The pattern playing out at Uber, Walmart, and Microsoft follows the same arc: encourage adoption, lose visibility, hit a ceiling, introduce controls. The problem is that most enterprises only reach step three after the damage is done.

CloudEagle.ai provides per-user and per-model AI token consumption tracking across tools, giving IT and finance teams the visibility they need before spend becomes unmanageable.

‍

Every Token Has A Cost

CloudEagle.ai helps you understand and optimize every one.

See How CloudEagle Works

Book a Demo