Uber’s AI bill got so high that it is now limiting Claude Code usage to $1,500: Report

Uber has introduced spending limits on AI-powered coding tools used by employees, such as Anthropic’s Claude Code and Cursor, in an effort to manage costs after it blew through its annual AI budget in just the first four months of the year, according to a report by Bloomberg.

The ride-hailing giant has capped employees at $1,500 per month in token spending for each AI coding tool, an Uber spokesperson told Bloomberg. This means that even if employees spend their AI tokens for one model, they can still use other coding software.

Employees can reportedly track their AI spending through an internal dashboard and seek approval if they need to exceed the allocated budget.

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

Notably, 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 following a surge in usage of AI-powered coding tools, particularly Claude Code.

The news of Uber limiting its spending on AI-powered coding tools comes just a day after Microsoft also reportedly began telling employees to wind down usage of Claude Code and instead shift to its GitHub Copilot CLI. As per a report by The Verge, Microsoft began rolling out Claude Code to thousands of employees in December last year but has now set 30 June as the last date for using Anthropic’s AI-based coding tool.

AI usage inside Uber:

In a post on X, Neppali revealed that agentic software engineering adoption is rapidly growing at the company, with 1,800 code changes per week now written by the company’s internal background coding agent and 95% of its engineers using AI every month.

Meanwhile, Uber CEO Dara Khosrowshahi revealed last month that around 10% of the code inside the company was now submitted and built by AI agents.

However, Uber Chief Operating Officer Andrew Macdonald questioned in a recent episode of Rapid Podcast whether the use of additional AI-powered tools was leading to better features being delivered to customers.

“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,” Macdonald said.

“That link is not there yet, right? I think 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, right?” he added.