Why Uber has Already Burned Through its AI Budget | AI Magazine
Article
Artificial Intelligence (AI)
Why Uber has Already Burned Through its AI Budget
By Diya Joseph<br>April 23, 2026
undefined mins
Share
Share
Uber is transitioning towards an agent engineering model where AI systems manage code deployment. Credit: Uber
Leading transportation tech company Uber exhausts its AI budget four months into the year due to rising adoption of AI tools like Claude Code and Cursor
Every good thing comes at a cost and, for Uber, that is the price of its heavy AI adoption. The transportation tech company has already exhausted its AI budget for 2026 due to surging use of coding tools, according to the company’s CTO Praveen Neppalli Naga.
Uber’s total research and development (R&D) expenses went up 9% year-over-year in 2025, hitting US$3.4bn, with AI a key cost driver.
As reported by The Information, engineers are extensively using tools like Claude Code, with about 11% of Uber's live backend code updates being completely written by AI agents.
Praveen Neppalli Naga, CTO at Uber. Credit: Uber
Exceeding company calculations
Uber teams are heading “back to the drawing board” as it counts the costs of surging AI usage, says Praveen.
Engineers were encouraged to adopt tools such as Claude Code and Cursor, even ranking staff on internal leaderboards based on usage. This strategy led to Claude Code becoming the dominant tool, while interest in Cursor plateaued.
This forms part of a wider industry trend known as “tokenmaxxing”. As tokens become a measurable unit of Gen AI usage, some organisations informally track how much engineers spend on these tools, often turning it into a competition within engineering teams.
As per reports, even companies such as Meta use internal dashboards to rank employees based on token usage. However, critics argue that high token consumption does not necessarily translate to better outcomes but wasteful spending.
This also becomes crucial at a time when AI budgets are already under pressure. At Uber, the rising usage is taking company costs to surging heights. To meet the growing need, Uber is now preparing to test OpenAI’s Codex in order to expand its AI stack.
The situation has created a scenario where the AI tools proved too successful to afford at scale as engineers reported monthly API costs between $500 and $2,000 per person.
Will it slow down hiring?
The numbers clearly reflect rising usage of AI tools within engineering teams. About 1,800 code changes implemented every week are written entirely by Uber’s internal AI coding agent without direct human input.
Additionally, nearly 95% of the company’s engineers use AI tools every month and close to 70% of the code that gets committed is generated by these systems.
Uber’s internal AI agent has gone from contributing less than 1% of code changes to roughly 8% in just a few months.
Praveen highlights how, in the long term, he sees a shift toward “agent engineers”, where AI systems don't just assist but fully handle coding, testing and deployment. This would also involve other AI tools supervising the process.
While hiring hasn't slowed down yet at Uber, AI has definitely taken on more of the workload. In an interview on The Diary of a CEO, Dara Khosrowshahi, CEO at Uber, was candid about the general worry regarding this shift. He estimated that AI will replace 70-80% of the work that humans currently do.
Dara Khosrowshahi, CEO at Uber. Credit: Uber
While intellectual jobs are expected to be affected within 10 years, Dara suggests physical roles such as driving and logistics to follow within 15 to 20 years.
With about 9.5 million drivers and couriers on the Uber platform, the CEO also acknowledges that autonomous vehicles will eventually fulfil a majority of those trips in the future.
When asked what those millions of people do next, Dara simply says: “I do not know.”
Company portals
Uber
Executives
Dara Khosrowshahi<br>CEO
Tags<br>UberAI AgentsAI BudgetClaude CodeCodex