Uber burned through its entire 2026 AI budget in just four months and is now capping employee usage of agentic coding tools — Claude Code and Cursor — at $1,500 per person per month. The company says 95% of its engineers use AI tools every month, and roughly 10% of all code submitted at Uber is now written by AI agents, not humans.
Bloomberg reported the Uber AI spending cap — $1,500 per employee per month — on June 2. TechCrunch confirmed it and added the key detail: Uber’s CTO disclosed to The Information that the company had consumed its full-year AI budget by April — four months into 2026.
The Numbers That Forced Uber’s Hand
Uber had previously encouraged engineers to use AI tools “as much as possible” — the company even ran internal leaderboards ranking employees by AI usage. That aggressive adoption worked. 95% of Uber’s engineers now use AI tools monthly, and ~10% of all code is submitted by autonomous agents.
Then the bill arrived. Four months of uncapped usage wiped out an entire year’s budget. Uber’s response was to impose hard caps, build an internal dashboard for tracking spending per employee, and create an override process for teams that legitimately need more.
The Uber AI Spending Cap: $1,500 Per Month, Per Tool
The cap applies separately to each agentic coding tool. That means an engineer could hit $1,500 on Claude Code and another $1,500 on Cursor in the same month — though in practice, exceeding either limit now requires manager approval.
For context: Claude Code billed at token usage can run thousands of dollars per month for developers running parallel autonomous coding agents at scale. The new limit is a meaningful ceiling for power users, especially on large-scale refactors or autonomous PR generation workflows. See our coverage of GitHub Copilot’s new AI credits billing model for how other platforms are handling the same pressure.
The ROI Problem Nobody Wants to Talk About
Uber COO Andrew Macdonald said during a podcast that “it’s very hard to draw a line” between AI spending and specific new consumer features. That’s not a confident statement about returns on what became an out-of-control budget line.
A June 2026 Bain survey found that AI is delivering less cost reduction than most enterprises predicted. Uber’s situation — near-total adoption, blown budgets, and ambiguous productivity attribution — is quickly becoming the canonical data point on the gap between AI adoption and AI ROI. The cost-per-output improvements in Claude Opus 4.8 and competing models may ease some of this pressure, but not eliminate it.
Frequently Asked Questions
Why did Uber cap AI tool spending?
Uber consumed its entire 2026 AI budget by April — just four months into the year — due to uncapped usage of agentic coding tools like Claude Code and Cursor. The company introduced per-employee monthly limits to control costs going forward.
What is Uber’s AI spending limit per employee?
Uber caps AI tool spending at $1,500 per employee per month, applied separately to each tool. An employee using both Claude Code and Cursor could theoretically spend up to $3,000 per month combined before needing manager approval to exceed limits.
Which AI tools did Uber cap?
The caps confirmed by Bloomberg apply to Anthropic’s Claude Code and Cursor, the two dominant agentic coding tools in enterprise use. Both require separate $1,500 monthly allocations per employee.
How much of Uber’s code is written by AI?
Approximately 10% of code submitted at Uber is written and submitted by AI agents — not by humans manually prompting tools, but by autonomous agents operating independently within workflows. Around 95% of Uber engineers use AI tools at least once per month.
Is Uber cutting AI usage entirely?
No. The $1,500 monthly cap still allows significant AI usage — it’s a governance measure, not a ban. Employees can request to exceed the limit with manager approval, and Uber has built an internal dashboard to track spending across tools.
The caps took effect immediately. If you’re watching AI’s enterprise moment, this is the clearest public signal yet that the “use AI as much as possible” phase is giving way to something more deliberate. For more on the enterprise AI landscape, including Microsoft’s push to lower developer AI costs with its own models, see our recent coverage.
Last Updated: June 2026
