Uber’s chief technology officer said the company’s spending on Claude Code ran through its planned budget by April, offering a new example of how quickly AI coding tools can become expensive at enterprise scale.
The remarks, reported by The Information, suggest that even when a company is enthusiastic about adopting AI software for engineering work, usage patterns can push costs far beyond initial forecasts. Claude Code, Anthropic’s coding-focused tool, is designed to help developers write, debug and navigate software more quickly. But Uber’s experience indicates that broad use inside a large organization can create a much bigger bill than expected.
The detail was presented as a cautionary tale about the economics of AI assistants in real-world business settings. Companies often trial these tools with a limited number of users, then expand access if the productivity gains appear promising. That can make early spending look manageable. But once usage spreads across teams and workloads increase, the cost structure can change rapidly.
Uber has not disclosed how much it budgeted for Claude Code or what it ultimately spent after the budget was consumed. The available reporting also does not say how many employees were using the tool or how it was deployed across the company. Still, the message from the CTO’s comment was clear. AI coding products can deliver value, but their variable usage-based pricing can lead to unexpectedly high expenses.
The episode also underscores a broader challenge for companies trying to bring generative AI into internal workflows. Leaders are under pressure to show that AI can speed up development, reduce repetitive work and help teams ship software faster. At the same time, they must manage new and often difficult-to-predict operating costs. For tools that charge by usage, a successful rollout can itself become the driver of higher spend.
That tension is becoming a central issue for enterprise AI adoption. A product that looks affordable in a small pilot may not stay that way once it is used daily by hundreds or thousands of workers. For finance and engineering teams, the challenge is not only deciding whether the tool is useful, but also determining how to limit cost growth without reducing its effectiveness.
Uber’s experience may resonate with other companies evaluating similar coding assistants. It suggests that before rolling out AI tools broadly, firms may need stronger controls around permissions, usage limits and budget monitoring. Without those safeguards, a tool that is intended to improve productivity can quickly become a line item that is hard to contain.
The reporting does not indicate that Uber is pulling back from AI coding tools altogether. Instead, the company’s early experience appears to be serving as a practical lesson in how quickly enthusiasm for agentic software can collide with budgeting reality.
For businesses weighing similar deployments, the takeaway is straightforward. AI coding tools can be powerful, but their costs may scale just as fast as their benefits.