Uber president and chief operating officer Andrew Macdonald has publicly stated that the company is finding it increasingly difficult to justify its growing investments in artificial intelligence. In an interview with a business podcast, Macdonald revealed that Uber exhausted its annual AI budget just four months into 2026, yet the expected return in terms of meaningful consumer features has not materialized.
"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 acknowledged that while certain underlying metrics, such as token consumption for AI tools like Claude Code, are trending in "a really astronomical direction," the connection to measurable productivity gains remains opaque.
Uber's spending on research and development reached $3.4 billion in 2025, representing a 9 percent increase over the previous year. Earlier this month, CEO Dara Khosrowshahi noted that the company is offsetting rising AI costs by hiring fewer human employees. Macdonald echoed this sentiment, emphasizing a new calculus: "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 remarks come amid a broader industry debate about the return on investment from AI. While tech giants and startups alike have poured billions into large language models, coding assistants, and generative AI tools, concrete improvements in user-facing features have been difficult to quantify. Uber, which operates in ride-hailing, food delivery, and freight, has integrated AI into areas such as route optimization, dynamic pricing, and driver matching. However, the company's leadership is now grappling with whether those enhancements translate to a tangibly better experience for riders and drivers.
Uber's experience is not unique. Across the tech sector, companies are facing pressure to demonstrate that AI spending is more than a competitive necessity—that it actually boosts revenue, customer satisfaction, or operational efficiency. Analysts have pointed to a mismatch between the hype around AI and the reality of deployment: while models can generate impressive demos, integrating them into stable, secure, and scalable products remains challenging. The concept of "token consumption"—the cost of processing prompts through AI models—has become a new metric that CFOs and operations leaders watch closely, often comparing it to traditional costs like employee salaries.
Macdonald's comments highlight a shift from the earlier "move fast and break things" era of AI adoption to a more mature phase where ROI is scrutinized. Uber's AI budget being exhausted by April suggests that internal demand for AI services was higher than anticipated, but the payoff in terms of shipped features has not kept pace. This echoes difficulties experienced by other firms that have deployed AI coding assistants: while developers report productivity gains in certain tasks, the overall impact on product delivery cycles is ambiguous.
For context, Uber has been investing in AI for years, from acquiring autonomous vehicle startups to building machine learning platforms. In 2025, the company launched an updated version of its AI-powered routing system that promised faster pickups and reduced wait times. Yet, Macdonald's skepticism indicates that even after such investments, the broader contribution of AI to the consumer experience remains uncertain. He suggested that over the coming quarters or years, the link may become clearer, but for now, the data does not support a straightforward conclusion.
The debate is part of a larger conversation about the economics of AI. While tokens—the units of text processed by language models—are relatively cheap, the volume of usage required to improve features can drive costs exponentially. Uber's use of Claude Code, an AI assistant for software development, is a prime example. Developers may feel more productive, but that productivity may be spent on refactoring code rather than shipping new capabilities.
Macdonald also touched on the implications for the workforce. With AI enabling fewer hires, the company is effectively substituting capital (in the form of compute costs) for labor. This trade-off is ethical as well as financial: if AI spending does not lead to better features, shareholders may question why labor costs are not being used more effectively. Meanwhile, employees and labor advocates worry about job displacement without clear productivity gains to show for it.
In the ride-hailing industry, competition is intense. Lyft, Uber's main rival in the US, has also increased AI spending, but with a smaller budget. Meanwhile, Waymo and other autonomous vehicle companies are pursuing a different AI path—full autonomy—which requires even larger capital outlays. Uber, having sold its own self-driving unit, now focuses on integrating third-party autonomous vehicles and enhancing its platform with AI. The question of whether these investments are visible to end users remains open.
Macdonald's candid remarks may signal a recalibration. Companies that once viewed AI as an unqualified boon are now starting to demand evidence. This could lead to more disciplined AI procurement, tighter budgets, and a renewed focus on applications that directly improve the core product rather than experimental uses. For now, Uber president provides a sober assessment: AI spending is getting harder to justify, and the industry should take note.
Source: The Verge News