The company Uber spent its entire artificial intelligence budget for 2026 in just four months. While American companies are investing hundreds of billions of dollars in AI, expecting the technology to save time and money, Uber has already encountered an unexpected issue — the annual budget for these initiatives was depleted much earlier than anticipated.
One of the Uber managers racked up a bill of 1,200 dollars during a single two-hour coding session. The spike in expenses followed the company’s introduction of the Claude Code tool from Anthropic for thousands of engineers, who began utilizing it far more actively than financial forecasts had predicted.
This situation illustrates how quickly AI costs can escalate when companies shift to pay-per-use tools instead of fixed monthly subscriptions. It also raises a broader question: do the billions invested in artificial intelligence justify the results achieved?
Rapid implementation of Claude Code
Uber rolled out Claude Code within its engineering organization in December 2025. By February, 32% of engineers were using AI programming agents, and by March, that figure had climbed to 84%. By spring, 95% of engineers were using AI tools monthly, with approximately 70% of new code being generated by them, according to Forbes.
Claude Code not only suggests individual code snippets but also allows engineers to assign AI agents to perform multiple tasks simultaneously, refactor large sections of software, generate tests, and create server code, all of which require substantial computational resources.
The average monthly cost of usage reportedly ranged from 150 to 250 dollars per engineer. The most active users incurred bills between 500 and 2,000 dollars, according to Forbes.
Challenges in measuring productivity
Reports indicate that Uber maintained internal rankings that evaluated engineers based on their level of Claude Code usage. This encouraged employees to use the tool more actively, although the teams promoting its implementation were not responsible for budget oversight.
The productivity gains from AI tools remain challenging to quantify. Uber President and Chief Operating Officer Andrew Macdonald later remarked that it is increasingly difficult to justify the rising costs of tokens — the payments for using AI models — without compelling evidence that they contribute to creating more valuable features for customers, according to The Verge.
Source: Yahoo
