Uber Technologies Inc. is facing a complex quandary over its artificial intelligence (AI) strategy. A top executive warns the company’s skyrocketing AI expenditures are failing to deliver clear, tangible benefits for consumers.
In a recent interview on the Rapid Response podcast, Uber President and Chief Operating Officer Andrew Macdonald sounded the alarm on the rideshare giant’s AI spending. Macdonald revealed that despite massive adoption of automated engineering tools, it remains remarkably difficult to connect rising technology costs to innovations meant to serve end users.
“That link is not there yet,” Macdonald said. “Maybe implicitly there’s 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 like 25% more useful consumer features.’”
The executive’s candid remarks follow a “head-exploding moment” last month when Uber Chief Technology Officer Praveen Neppalli Naga disclosed that the firm had exhausted its entire 2026 AI coding tools budget in just four months.
The budgetary blowout was fueled by an internal leaderboard ranking teams by their total AI tool usage, which heavily incentivized employees to adopt Anthropic’s Claude Code. Usage among Uber engineers surged from roughly one-third in February to 84% in March. While the average per-engineer bill ran between $150 and $250 a month, heavy users generated up to $2,000. In one hands-on demonstration, Neppalli Naga himself burned through $1,200 worth of tokens in just two hours.
What is more, during a recent earnings call Uber CEO Dara Khosrowshahi noted that 95% of the engineering workforce has adopted AI coding tools, with autonomous agents now authoring more than 10% of the company’s committed code. To offset ballooning costs, Khosrowshahi announced that Uber would scale up its AI budget while slowing down human hiring.
However, Macdonald indicated that sacrificing headcount for unproven technology is becoming difficult to defend, noting that while engineers treat AI tools as free, “somebody’s paying the bill.”
“If you’re not actually able to draw a direct line to how much useful features and functionality you’re shipping to your users, that trade becomes harder to justify,” Macdonald warned.
Uber’s friction points reflect a broader macroeconomic shift. A recent study by research firm Gartner forecasts that enterprise spending on AI agent software will jump 139% to nearly $207 billion this year. Yet, Gartner warns that cheaper per-token costs will not translate to corporate savings, because complex, agentic models require far more computing power per task.
“Uber’s reassessment marks the point where AI developer tooling moves from experimental budget to governed P&L line item. Token consumption metrics scale with usage; business outcome attribution does not, and engineering leaders are being asked to defend spend they cannot tie to deliverable productivity,” said Mitch Ashley, vice president and practice lead for Software Lifecycle Engineering and AI-Native Software Engineering at The Futurum Group.
“Adoption metrics and seat counts no longer satisfy as ROI evidence,” Ashley said. “Engineering organizations need observability-native solutions to understand how AI consumption translates to measurable productivity, and vendors selling developer AI will face procurement conversations grounded in outcome attribution.”
Consequently, Silicon Valley is quietly recalibrating. Microsoft Corp. reportedly began canceling direct Claude Code licenses for most engineers, diverting them to GitHub Copilot CLI to manage costs. Duolingo CEO Luis von Ahn reversed a policy that tied employee evaluations to AI usage after staff complained the metric rewarded blind tool adoption over actual outcomes.
At the same time, Anthropic and OpenAI are shifting toward usage-based, metered pricing models to capture the immense compute demands of autonomous agents.
Despite the near-term financial friction, Uber is not abandoning its tech-forward ambitions.
Financial disclosures show Uber spent $951 million on research and development in the first quarter of this year alone, a 17% hike year-over-year. Macdonald emphasized that the company remains fully committed to autonomous driving, predicting the technology will become so ubiquitous over the next two decades that the next generation will never need a driver’s license.
