Uber Caps Claude Code Spending at $1,500 After Blowing AI Budget

Key Takeaways

• Uber now limits employees to $1,500 per month per AI coding tool after exhausting its annual AI budget in four months
• 10% of Uber's codebase is now generated by AI agents, with 25% of code commits coming via Claude Code last quarter
• Microsoft is also pulling back on Claude Code, telling employees to switch to GitHub Copilot by June 30

Uber has introduced a $1,500 monthly cap on employee spending for AI coding tools like Anthropic's Claude Code and Cursor. The restriction comes after the ride-hailing company burned through its entire annual AI budget in just four months of 2026.

The news, first reported by Bloomberg, signals a broader reckoning for enterprises that adopted AI coding assistants aggressively. What started as productivity experiments have turned into financial headaches as token-based pricing scales with usage.

How the Cap Works

Under Uber's new policy, each employee can spend up to $1,500 per month in tokens for each AI coding tool. If an engineer maxes out their Claude Code allowance, they can still use Cursor or other approved tools. Employees track their spending through an internal dashboard and can request approval to exceed the limit if needed.

“We think this is all a pretty straightforward way to responsibly encourage agentic AI adoption and experimentation at scale across the company.”

— Uber spokesperson to Bloomberg

The company is framing the caps as guardrails rather than restrictions. Engineers still have meaningful budgets to work with. But the message is clear: unchecked AI spending is no longer acceptable.

The Numbers Behind the Problem

Uber CTO Praveen Neppali Naga told The Information in April that the company was "back to the drawing board" after blowing past its full-year AI budget. The culprit was a surge in usage of AI-powered coding tools, particularly Claude Code.

The adoption numbers explain the cost explosion. According to Neppali's recent posts on X, 1,800 code changes per week are now written by Uber's internal background coding agent. 95% of engineers use AI tools every month. CEO Dara Khosrowshahi revealed that roughly 10% of the company's code is now submitted and built by AI agents.

That's a lot of tokens. And tokens cost money.

The ROI Question No One Can Answer

Here's where it gets interesting. Uber COO Andrew Macdonald raised a pointed question on a recent episode of the Rapid Podcast: is all this AI usage actually delivering better products?

“You sometimes go and talk to your senior engineering leaders and say, okay, how many projects that were on the cutting room floor got moved above the line because of the productivity gains? Because 25% of our code commits were via Claude Code last quarter. That link is not there yet, right?”

— Andrew Macdonald, COO at Uber

Macdonald acknowledged that more code is getting shipped. But drawing a direct line between AI-assisted commits and better consumer features remains difficult. The productivity gains are real. Whether they translate to business outcomes is an open question.

Microsoft Is Also Pulling Back

Uber isn't alone. Microsoft began telling employees to wind down Claude Code usage and shift to its own GitHub Copilot CLI, according to The Verge. The company rolled out Claude Code to thousands of employees in December 2025 but has now set June 30 as the last date for using Anthropic's tool.

Microsoft's move is partly about promoting its own product. But it also reflects the same cost pressures Uber faces. When every API call has a price tag, usage scales faster than budgets.

The Bigger Pattern

AI coding tools operate on consumption-based pricing, similar to cloud infrastructure. The more tokens your engineers burn, the higher your bill. This model worked fine during pilot programs with small teams. It breaks down when thousands of developers start using these tools daily.

Enterprises are now entering what you might call operational maturity for AI. The experimentation phase is over. Finance wants predictable costs. IT wants governance. And engineering leaders are being asked to justify the spend.

Discussion on Hacker News focused on whether $1,500 is generous or restrictive. Some engineers argued the cap could discourage use of the most efficient tools. Others called it a necessary step for fiscal responsibility. The debate reflects real tension between developer productivity and cost control.

Frequently Asked Questions

Why did Uber cap Claude Code spending at $1,500?

Uber burned through its entire annual AI budget in just four months due to rapid adoption of AI coding tools. The $1,500 monthly cap per tool helps control costs while still allowing engineers to use AI assistants.

Can Uber employees still use AI coding tools after hitting the limit?

Yes. The $1,500 cap applies per tool, so engineers who max out Claude Code can still use Cursor or other approved tools. They can also request approval to exceed the limit through an internal dashboard.

How much of Uber's code is written by AI?

CEO Dara Khosrowshahi says about 10% of Uber's code is now submitted and built by AI agents. Last quarter, 25% of code commits came through Claude Code specifically.

Is Microsoft also limiting Claude Code usage?

Yes. Microsoft is telling employees to wind down Claude Code usage by June 30, 2026, and shift to its own GitHub Copilot CLI.

Are AI coding tools actually improving productivity?

Uber's COO Andrew Macdonald acknowledges more code is shipping but says it's hard to draw a direct line between AI usage and better consumer features. The ROI question remains open.

Source: mint / Aman Gupta