Uber Blows Through Its AI Budget — And Has Little to Show for It
Uber reportedly exhausted its entire annual AI budget by April 2026 — just four months into the year — and the company's own president is now asking whether any of it was worth it.
In a candid interview on the Rapid Response podcast, Uber president and chief operating officer Andrew Macdonald admitted that the ride-hailing giant isn't seeing a clear connection between its surging AI spending and better products for consumers.
"That link is not there yet," Macdonald said. "I think maybe implicitly there is 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 25 percent more useful consumer features.'"It's a rare moment of public candour from a major tech company executive at a time when the industry has largely been cheerleading AI investment without much scrutiny.
The Claude Code Problem
Macdonald specifically called out Uber's use of Claude Code — Anthropic's AI-powered coding assistant — as a major driver of token consumption costs. The company has been using it to speed up software development, but the cost-versus-output equation isn't adding up the way leadership expected.
This raises a broader question that more companies are starting to ask quietly: is AI actually making teams more productive, or is it just generating more activity without more results?
For Uber, a company that operates at massive scale across ride-sharing, food delivery, and freight, even marginal efficiency gains matter. But when your AI tools are burning through budget at that pace, marginal gains aren't enough.
A Wider Industry Reckoning
Uber isn't alone. Across the tech world, the narrative around AI has shifted from breathless optimism to cautious scrutiny. Early adopters flooded into AI coding tools and large language model integrations expecting transformative productivity boosts. What many found instead was an expensive experiment with uneven results.
The challenge is that token consumption — the metric that drives most AI pricing — doesn't directly map to business value. A developer using Claude Code might generate ten times as many lines of code, but if those lines require heavy review, create new bugs, or solve the wrong problems, the productivity math falls apart.
Macdonald's comments suggest Uber is grappling with exactly that disconnect. The company is spending more, consuming more compute, and shipping something — but can't definitively say that something is meaningfully better for riders or drivers.
What This Means for the AI Gold Rush
For investors and analysts watching the AI sector, Uber's situation is a useful reality check. Billions of dollars are flowing into AI infrastructure, tooling, and enterprise licenses. The assumption underpinning all of it is that productivity gains will materialize at scale and justify the costs.
If a company as large and tech-sophisticated as Uber can't draw that line, it raises real questions about the ROI timeline for the broader industry.
That doesn't mean AI is a bust — but it does suggest the easy wins may already be claimed, and the harder, more expensive gains are what's left.
Macdonald's willingness to say the quiet part out loud may signal a broader shift: after years of hype, companies are starting to demand receipts.
Source: The Verge
