Claude ate up Uber's full year AI budget in 4 months? Here is what really happened

Uber’s aggressive push into AI coding tools appears to have backfired financially, with the company reportedly burning through its entire 2026 AI budget in just four months after giving engineers broad access to AI tools primarily Claude.

According to reports, the company saw AI adoption explode internally after rolling out Claude Code to engineers in December 2025. What began as a productivity-focused experiment quickly turned into a massive expense problem that even Uber's leadership did not fully anticipate. Uber in fact has gone on record to say the company exhausted its entire annual AI budget in four months.

Uber has not disclosed what this annual AI budget is in terms of exact numbers but since the company is known to pool it under Research and Development (R&D), it is likely around a rough ball figure of $3.4 billion which was the amount it spent on R&D in 2025. Again, there is no way of knowing if this figure is more or less than $3.4 billion in 2026 and how much of it was actually allocated for AI spending, but at least it gives us something to work with. Regardless of what the exact numbers are, the headline grabbing piece of course remains how Claude – and possibly other AI – bills are going out of control.

Uber Chief Technology Officer Praveen Neppalli Naga confirmed in the Rapid Response podcast that the company exhausted its AI coding budget by April 2026. The issue was not because the tools failed, or employees misused them. In fact, Uber engineers used Claude Code exactly the way it was intended to be used — for writing backend code, running automated tests, refactoring large codebases, and managing multiple AI agents in parallel. Some tokenmaxxing was involved as well, reports suggest. Tokenmaxxing is a trend where employees are encouraged to use as many AI tokens as possible to the extent they compete on leaderboards. Companies like Amazon are now – finally – cracking down on tokenmaxxing after rising cost with senior employees telling employees not to use AI just for the sake of it.

At Uber, the problem seems to be that of scale. Internal adoption reportedly jumped from 32 percent of engineers using AI in February to 84 percent by March. By spring, around 95 percent of Uber engineers were using AI tools every month, while nearly 70 percent of committed code was being generated with AI assistance. A report from Forbes claims that about 11 percent of live backend updates were being written entirely by AI agents without human intervention.

AI adoption inside Uber grew faster than expected

The financial impact became visible almost immediately. Uber reportedly spent between $150 and $250 per engineer every month on average, while heavy users generated bills ranging from $500 to $2,000. Naga himself said he once spent $1,200 during a two-hour demo session.

What made the situation more difficult was the pricing structure. Unlike traditional enterprise software that works on predictable per-seat subscriptions, Claude Code operates on token-based billing. That means costs change depending on how much computing power an engineer consumes. Someone using simple autocomplete features costs far less than a developer running several AI agents simultaneously across a massive code repository.

Uber now doubting AI spending?

Uber President and COO Andrew Macdonald recently admitted that the company is still struggling to clearly measure whether the massive AI spending is directly leading to more useful consumer-facing features.

“That link is not there yet,” Macdonald said during an interview on the Rapid Response podcast. “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.’”

The situation at Uber shows a growing challenge for companies rushing into AI adoption. While the cost of AI models themselves may fall over time, overall enterprise spending can still rise because advanced AI agents consume far more tokens and computing resources per task.

Anthropic itself recently moved toward more usage-based pricing for Claude’s agent tools, moving away from flat subscription models. OpenAI CEO Sam Altman has also described a future where AI works like utilities such as electricity or water, with businesses paying based on consumption.

Uber is not the only tech giant rethinking AI spending. Reports suggest Microsoft has already started reducing direct Claude Code licenses internally and is instead moving engineers toward GitHub Copilot tools. Other business leaders have also become more cautious after initially embracing AI with strong optimism.

Despite the budget overrun, Uber is not slowing down its AI ambitions. CEO Dara Khosrowshahi recently said around 10 percent of the company’s committed code is now being written by autonomous AI agents. He also claimed AI tools are giving employees “superpowers” across departments including engineering, legal, and marketing.