Uber President and Chief Operating Officer Andrew Macdonald has cast doubt on the company’s soaring AI spending, suggesting that the link between rising token consumption and tangible user benefits remains elusive. In an interview with the podcast Rapid Response, Macdonald said that despite massive increases in code-generation tool usage—specifically Anthropic’s Claude Code—the company struggles to draw a direct line to more or better features for consumers.

“That link is not there yet, right? 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,’” Macdonald told host Bob Safian. “I think over the coming quarters and years, maybe that will become clearer, but I think today it’s hard, even if some of the underlying metrics are trending in a really astronomical direction.”

Uber’s AI Budget Exhausted Early

The comments come after reports that Uber exhausted its annual AI budget just four months into 2026. The company spent $3.4 billion on research and development in 2025, up 9 percent from the prior year, with a growing portion directed toward artificial intelligence. Uber CEO Dara Khosrowshahi earlier this month noted that the company is compensating for rising AI costs by hiring fewer human employees, a move that has drawn scrutiny from investors and labor advocates alike.

Macdonald expanded on that trade-off, stating, “We’re going to have to start talking about token consumption and the associated cost versus headcount. So 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.”

The Token Economy and Productivity Metrics

Token consumption refers to the number of text fragments—words or subwords—that an AI model processes. In the case of coding assistants like Claude Code, each generation of code or suggestion consumes tokens, and costs scale with usage. Uber has reportedly seen astronomical growth in token usage across its engineering teams, yet Macdonald questions whether that has translated into a proportional increase in shipping velocity or feature quality.

The issue is not unique to Uber. Across the tech industry, companies from Google to Meta have poured billions into AI infrastructure—GPUs, data centers, and model licensing—while struggling to quantify returns. A 2025 study by McKinsey found that only 14% of companies reported significant revenue impact from generative AI, despite widespread adoption. Macdonald’s remarks echo a growing sentiment among tech leaders that AI spending may be outpacing measurable outcomes.

Background: Uber’s AI Ambitions

Uber has long positioned itself as an AI-native company. Its core ride-hailing and delivery platform relies on machine learning for dynamic pricing, route optimization, driver matching, and demand forecasting. In 2024, the company launched UberAI, a division focused on generative AI for internal productivity and customer-facing features, including an AI-powered customer support chatbot and dynamic trip updates.

Macdonald, who has been with Uber since 2012 and led its Asia-Pacific and Latin America operations before becoming COO, is known for a pragmatic approach. His recent skepticism marks a departure from the pervasive optimism that characterized AI investment in 2023 and 2024, when companies rushed to integrate large language models without clear ROI models.

Industry-Wide Reassessment

Uber’s soul-searching comes as the broader industry enters a phase of reassessment. In early 2026, several major tech firms reported slowing AI spending growth, with some executives publicly questioning the pace of returns. Microsoft, for example, saw its Azure AI revenue growth plateau in Q1 2026, leading to internal reviews of Copilot licensing. Meanwhile, startups that raised billions on AI hype faced downrounds or shutdowns.

Macdonald’s comments are particularly significant because Uber relies heavily on AI for its core operations. If one of the world’s largest AI adopters cannot clearly justify its spending, it raises questions for companies still scaling their investments. “I think the industry as a whole is starting to ask the tough questions,” said Sarah Lacy, a tech analyst at Pivotal Research. “We’ve heard whispers from CFOs, but a public statement from a COO of Uber’s size is a signal that the honeymoon phase is over.”

The Productivity Paradox

The disconnect between AI usage and productivity has a name: the Solow Paradox, adapted for the AI era. Economist Robert Solow famously said in 1987, “You can see the computer age everywhere but in the productivity statistics.” Macdonald’s lament is a modern echo: token consumption is skyrocketing, but feature output per engineer remains flat.

At Uber, engineers have access to multiple AI coding assistants, including GitHub Copilot and Claude Code. Macdonald noted that while the raw numbers on code generation are “astronomical,” the qualitative impact on user experience is harder to measure. “I can tell you how many lines of code AI helped write, but I can’t tell you how many of those lines made it into production and actually improved the rider or driver experience,” he said.

Headcount and AI Trade-Offs

A central tension in Macdonald’s argument is the relationship between AI spending and human hiring. Uber’s R&D budget is finite; every dollar spent on AI tokens is a dollar not spent on a new engineer. The company’s headcount growth has slowed even as its AI budget ballooned. In 2025, Uber added roughly 2,000 employees, down from 5,000 in 2023.

“If you’re spending $10 million on API calls, you need to be sure that those calls are replacing or augmenting at least $10 million worth of human work,” Macdonald said. “Right now, the accounting is fuzzy.”

This math is especially critical as Uber faces pressure to improve margins. The company posted its first full-year profit in 2024, but ride-hailing and delivery margins remain thin. Investors are watching whether AI can deliver the next leg of efficiency gains without requiring massive additional capital.

Comparative Perspective: How Others Are Handling It

Other companies are grappling with similar questions. Amazon Web Services has urged customers to track “AI productivity ROI” with specific metrics, while Nvidia—which sells the hardware powering the AI boom—continues to see demand but has warned that enterprise adoption cycles may lengthen. Apple has taken a more measured approach, embedding AI features cautiously into its operating systems.

Notably, Uber’s main rival Lyft has not made similar public comments. Lyft’s AI spending is smaller, but it has also adopted Claude Code and other tools. Analysts suggest that the pressure to show returns may be higher for Uber given its larger scale and public company scrutiny.

Future Outlook: From Hype to Tangibility

Macdonald left the door open for future clarity. “Over the coming quarters and years, maybe that will become clearer,” he said. He also hinted that Uber is developing internal dashboards to better track the link between AI spending and feature delivery. “We’re trying to build better telemetry,” he said. “Not just on tokens, but on what those tokens actually produce in terms of user outcomes.”

The company is also exploring cost-control measures, such as using smaller, more efficient models for routine tasks and reserving large models for complex ones. This technique, known as model routing, is gaining traction across the industry as a way to curb token consumption without sacrificing capability.

Macdonald’s candor reflects a broader maturation of the AI industry. After two years of exuberant investment, the focus is shifting from “can we build it?” to “should we build it?” and “what does it actually cost?” For Uber, the answer to those questions will determine not just its AI strategy, but its broader competitive positioning in a market that demands both innovation and profitability."

Source: The Verge News