The artificial intelligence honeymoon phase is ending as major providers terminate early-adoption subsidies and force enterprise clients onto expensive token-based pricing models. This structural industry shift is now hitting the balance sheets of the biggest tech companies in the world. On Monday, Uber COO Andrew Macdonald revealed on the Rapid Response podcast that the ride-hailing company is finding it “harder to justify” its continued investments in artificial intelligence.
Macdonald explained that Uber cannot link these high computing costs to proportional productivity gains or new consumer features. The financial strain is acute. According to original reporting by The Verge, the company rolled out Anthropic’s Claude Code to its 5,000 engineers in January.
Just four to five months into the year, Uber CTO Praveen Naga confirmed the company completely exhausted its allocated 2026 AI budget.
The internal budget exhaustion highlights a growing problem with token consumption versus actual headcount, a trend detailed in Business Insider’s coverage of Macdonald’s remarks. When developers write and test code using these AI assistants, they generate enormous volumes of tokens every single day. Under the new variable consumption models, those daily interactions quickly outpace traditional flat-fee software subscriptions.
Why Token-Based Pricing Threatens Enterprise AI Adoption
Uber is not alone in this financial crunch. Microsoft is reportedly dealing with similar internal budget blowouts as companies transition from raw experimentation to demanding strict return on investment, India Today reported on the macro-level industry changes. During the initial AI boom, providers heavily subsidized inference costs.
They often spent thousands of dollars to support a single user paying a standard monthly fee. Now that AI companies face pressure to generate real revenue, enterprise clients are absorbing the true compute costs.
As individual token prices drop, the sheer volume of tokens generated by mass engineering teams has roughly tripled enterprise AI spending. This reality is turning a popular developer tool into a severe financial liability.