Uber's $3.4B AI Budget Depleted Amid Surge in Claude Code Usage, CTO Says
Key Takeaways
- ▸Uber's annual AI R&D budget of $3.4 billion was exhausted within months of 2026 due to unexpectedly high usage of Anthropic's Claude Code
- ▸AI agents now account for approximately 11% of Uber's live backend code updates, demonstrating significant productivity gains
- ▸The company is pursuing a future vision of fully autonomous "agent engineers" that would handle coding, testing, and deployment with AI supervision rather than human engineers
Summary
Uber has exhausted its planned AI research and development budget just months into 2026, despite allocating $3.4 billion annually, according to Chief Technology Officer Praveen Neppalli Naga. The budget shortfall was driven by unexpectedly high adoption of Anthropic's Claude Code among engineers, who were actively encouraged to use AI coding tools and ranked on internal leaderboards by usage. The surge in AI tool consumption has forced Uber to return to the drawing board on its AI spending strategy.
The cost explosion comes as AI is beginning to deliver measurable productivity gains at the ride-sharing giant. Approximately 11% of Uber's live backend code updates are now written by AI agents, a sharp increase over just a few months. Naga outlined a longer-term vision of "agent engineers"—autonomous AI systems capable of handling coding, testing, and deployment with minimal human oversight. While Uber continues to test additional tools like OpenAI's Codex, the company faces mounting pressure on R&D expenses, which rose 9% to $3.4 billion in 2025 with further increases expected.
- Rising AI tool costs are emerging as a major expense driver alongside traditional R&D, with further increases anticipated
Editorial Opinion
While Uber's rapid depletion of its AI budget highlights the genuine productivity gains these tools can deliver, it also exposes a critical blind spot in enterprise AI planning. Companies are discovering that the operational costs of AI adoption can quickly spiral when usage incentives outpace financial forecasting. This cautionary tale suggests that organizations need more sophisticated cost-management frameworks and usage governance models before aggressively encouraging AI tool adoption at scale.
