Uber burned its entire 2026 AI budget in four months. Now executives question the returns

Uber exhausted its full-year 2026 AI budget by April, a spending spree driven primarily by enterprise-wide adoption of Anthropic's Claude Code and Cursor across its engineering organization. According to The Information, CTO Praveen Neppalli Naga acknowledged that "the budget I thought I would need is blown away already." Monthly API costs per engineer ranged from $500 to $2,000 as adoption skyrocketed. Briefs reported that 95% of Uber engineers now use AI tools monthly. The speed of consumption caught leadership off guard, transforming what was meant to be a year-long allocation into a quarterly burn.

This isn't a case of cautious piloting. Uber's engineering culture embraced AI coding assistants with minimal friction, and the costs scaled accordingly. The company now faces the uncomfortable reality that enthusiasm outpaced financial planning, a pattern repeating across Silicon Valley.

Uber president and chief operating officer Andrew Macdonald told the Rapid Response podcast that connecting AI spending to customer value has become increasingly difficult. "It's very hard to draw a line" between rising costs and useful features, he said, according to Fortune and Business Insider. Macdonald's comments carry weight because Uber is already one of the most AI-dependent companies in tech. Its pricing algorithms, route optimization, and predictive features are deeply embedded in the product. If Uber can't clearly demonstrate ROI, the warning resonates across less AI-mature enterprises.

Macdonald emphasized he's not rejecting AI investment outright. Rather, he's demanding clearer accountability for what these tools actually deliver. This marks a shift from the experimental phase of enterprise AI adoption toward a more skeptical, ROI-focused evaluation that prioritizes measurable outcomes over speculative potential.

Microsoft offers a parallel case that reinforces Uber's concerns. Livemint reported that Microsoft cancelled most of its direct Claude Code licenses and redirected engineers toward GitHub Copilot CLI, just six months after opening access across thousands of developers. The reversal came after adoption proved "too swift" and costs escalated beyond projections. Microsoft's experience suggests that even companies with deep AI integration expertise can misjudge consumption patterns and total cost of ownership.

The Microsoft and Uber cases together signal a broader enterprise reckoning. Initial AI tool pricing often masks true costs at scale. Per-user or per-token fees that seem reasonable in pilot phases multiply dramatically when engineering organizations adopt tools as default infrastructure rather than occasional assistance.

Despite massive spending, the productivity gains remain elusive. Business Insider noted Macdonald's observation that increasing AI costs aren't yielding proportional output improvements. The Verge highlighted his broader point that there's "no clear connection between AI usage and productivity." This captures a growing frustration across tech: tools that promise 10x developer velocity often deliver incremental convenience at premium prices.

Uber's situation is particularly telling because its use cases are genuine, not speculative. If a company with real AI infrastructure can't justify costs, the challenge for firms still searching for applications becomes far more acute. The gap between tool adoption and business value creation has emerged as the central enterprise AI problem of 2026.

The Uber and Microsoft experiences suggest enterprise AI budgeting needs fundamental restructuring. Companies are discovering that traditional annual IT allocations collapse under consumption-based AI pricing models. Projectflux noted that Uber's $3.4 billion R&D budget provided insufficient cushion against engineering-driven AI tool adoption. Procurement teams now face pressure to implement usage caps, department-level chargebacks, or outcome-based vendor contracts.

Vendors like Anthropic and Cursor may need to evolve pricing to retain enterprise customers. If buyers increasingly demand proof of productivity impact before renewing, the current land-and-expand model faces resistance. The next 12 months will likely see more explicit ROI requirements baked into enterprise AI contracts, with finance teams gaining veto power over engineering tool choices that previously flew under the radar.

Macdonald's public skepticism matters beyond Uber. As Startupfortune observed, his comments represent "the harder question every company will soon face." When a leading AI adopter publicly questions value, it validates finance executives across industries who have harbored similar doubts. The Cleantechnica framing, that someone "poked the bubble," captures the sentiment shift from uncritical enthusiasm toward measured scrutiny.

This doesn't indicate AI investment is stopping. Uber continues spending, just more anxiously. The inflection is psychological, from growth-phase optimism to maturity-phase accountability. Companies that navigate this transition successfully, tying AI costs tightly to revenue or efficiency metrics, will gain durable advantage over competitors still chasing capability without clarity.