Uber spends entire 2026 AI budget in 4 months, sees no ROI

Anthropic is likely generating at least 35% more revenue than OpenAI, but enterprise AI buyers are increasingly doubtful about throwing money after “tokenmaxxing.” The latest company to indirectly ask vendors for evidence rather than promises? Uber.

Uber president and COO Andrew Macdonald told the Rapid Response podcast that the company cannot yet draw a clear connection between rising Claude Code token consumption and the additional useful consumer features it has shipped.

The math is merciless. Earlier this year, Uber CTO Praveen Neppalli Naga also said that Uber burned through its annual AI budget within the first four months of 2026.

After conversations with senior engineering leaders, Macdonald now admits he doesn’t see higher token usage translating into a proportional increase in useful consumer features.

“If you're not actually able to draw a direct line to how many useful features and functionality you’re shipping, that trade becomes harder to justify,” said Macdonald.

“That link is not there yet.”

On the same pod, he quotes Airbnb CEO Brian Chesky, who was asked why his company wasn’t rushing out AI-led interfaces for customers.

“He said it was very hard for them to make the transition from their traditional UI to an AI-led UI without killing conversion. You can destroy billions of dollars of value in the short term,” said Macdonald.

“By the way, it’s not obvious it’s better for the consumer.”

In other words, it looks like doubt is slowly creeping in, and executives are discussing how to get a handle on token consumption costs.

AI firms such as Anthropic and OpenAI are increasingly shifting to token-based pricing that charges customers based on usage, rather than a subscription-based model, reflecting a broader reset in AI economics and raising costs for enterprises.

There are more examples of businesses openly asking for evidence of AI’s usefulness and making hard pivots.

Starbucks, which 9 months ago launched an AI-powered inventory tool that employees used to automate the counting of milk and beverages in stores, has since scrapped it after finding it often miscounted and inaccurately identified items.

US retail giant Target also said this week it would likely reassess how AI is deployed across the corporation, with the firm’s India head, Andrea Zimmerman, observing: “It’s about the intentional use and integration of AI rather than deploying it everywhere.”

At the National Restaurant Association Show this week, restaurant tech executives were offering warnings about AI overdependence left and right, it seems.

Have thoughts about this topic? Others do, too. Join them in the discussion.

Brendan Sweeney, CEO and cofounder of digital marketing and ordering platform Popmenu, told Restaurant Dive that overadoption of LLMs rests on shaky fundamentals.

That’s because the cost of AI tokens required to operate coding services can exceed the salaries of the workers the technology purports to replace. At the same time, the slow pace of data structure construction and rising cost of electricity make it unlikely that AI services will see a decline in cost.

Unlock more exclusive Cybernews content on YouTube.