Uber president and chief operating officer Andrew Macdonald has cast doubt on the value of the company’s massive artificial intelligence investments, saying it is getting “harder to justify” the spending. In an interview with the podcast Rapid Response, Macdonald acknowledged that Uber exhausted its annual AI budget just four months into 2026, yet the company cannot clearly connect rising token consumption with better user features.
“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 said. He added that while underlying metrics are “trending in a really astronomical direction,” the connection between AI usage and productivity remains elusive.
Uber spent $3.4 billion on research and development in 2025, a 9 percent increase over the previous year. Earlier this month, CEO Dara Khosrowshahi said the company is offsetting rising AI costs by hiring fewer human employees. Macdonald echoed that sentiment, emphasizing that the trade-off between token consumption and headcount is becoming central to internal debates.
“We’re going to have to start talking about token consumption and the associated cost versus headcount,” Macdonald said. “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 remarks from a top executive at one of the world’s largest ride-hailing and delivery platforms highlight a growing skepticism about the returns on AI investments across the tech industry. While companies from Google to Microsoft have poured billions into AI infrastructure and model training, questions about measurable business impact are mounting.
Uber’s AI efforts have centered on tools like Claude Code, an AI coding assistant developed by Anthropic. The company has been aggressively adopting the tool to accelerate software development. However, Macdonald’s comments suggest that despite soaring token usage, the actual output of useful features has not kept pace.
Token consumption is a key metric in the AI industry. A token is a unit of text that an AI model processes; the more tokens consumed, the more computing power and costs incurred. Uber’s AI budget being exhausted by May indicates an incredibly rapid adoption rate, presumably driven by engineers using Claude Code for code generation, debugging, and documentation.
But Macdonald warned that without a clear correlation to user-facing improvements, the spending spree may not be sustainable. “I think over the coming quarters and years, maybe that will become clearer, but I think today it’s hard,” he said.
The issue is not unique to Uber. Across the tech sector, companies are grappling with how to measure the ROI of generative AI. For example, some enterprises report productivity gains in software development, but translating those into revenue or user satisfaction has proven difficult. A recent survey by Gartner found that while 80% of executives believe AI will increase productivity, only 15% could provide concrete examples of measurable improvements.
Uber’s situation is particularly instructive because the company is a platform business with millions of drivers and riders. Its core product—matching supply and demand for rides and deliveries—depends on reliable, fast algorithms. AI has long been central to Uber’s operations, from surge pricing to route optimization. Yet the new wave of generative AI tools represents a different type of investment, one that Macdonald suggests is harder to justify in traditional business terms.
The company’s R&D spending of $3.4 billion in 2025 is substantial but not out of line with peers. By comparison, Apple spent about $30 billion on R&D in fiscal 2025, while Meta spent $38 billion. However, Uber’s revenue base is smaller, making the relative weight of AI costs more significant. In 2025, Uber reported revenue of $45 billion, meaning R&D represented about 7.5% of revenue—a typical proportion for a tech firm.
Still, the comments from Macdonald and Khosrowshahi signal a potential shift in strategy. If Uber cannot demonstrate clear returns from AI spending, it may be forced to slow its adoption or redirect resources. This could have ripple effects for AI startups like Anthropic, which depend on large enterprise contracts to sustain their growth.
Another factor is the broader economic environment. With interest rates remaining elevated and venture capital tightening, tech companies are under pressure to prioritize profitability over growth. Uber’s own path to profitability has been a long one; the company only reported its first full-year profit in 2023. Any large cost center that does not directly contribute to earnings is likely to face scrutiny.
Macdonald’s comments also touch on a philosophical question: Should AI tools be measured by consumption (tokens used) or by outcomes (features shipped)? He suggests the latter is more important but harder to quantify. This debate mirrors earlier discussions about cloud computing, where companies initially struggled to justify cloud costs but eventually found ways to measure agility and time-to-market.
Uber’s experience may serve as a cautionary tale for other organizations rushing to adopt generative AI. Without clear metrics linking AI usage to business objectives, executives may find themselves in the uncomfortable position of defending escalating costs. Macdonald’s advice seems to be: proceed with caution, measure carefully, and be ready to make hard trade-offs.
The ride-hailing giant is not abandoning AI, but its top leaders are demanding more accountability. As Macdonald put it, if the line between token consumption and useful features remains blurry, the spending becomes “harder to justify.” For an industry obsessed with AI’s potential, that is a sobering message.
In the coming months, all eyes will be on Uber’s quarterly earnings to see whether AI spending begins to yield tangible results—or whether the company will be among the first to publicly pull back.
Source: The Verge News