After blowing its entire 2026 AI budget in months, Uber CTO says ‘We’re coming to the end of the so-called ‘tokenmaxxing’ era’

Uber's Chief Technology Officer Praveen Neppalli Naga revealed that the company has hit a turning point regarding its AI spending strategy. In an interview with The Information, Naga admitted that the rideshare giant had overestimated its AI investment in the early months of 2026. The company encouraged employees to use its AI tools extensively, even ranking software engineers on usage levels, which led to a phenomenon known as "tokenmaxxing."

However, Naga believes this era is coming to an end. To address the issue, Uber increased the number of employees using frontier AI tools, which resulted in a lower cost per token. The company improved prompt caching, adjusted default model settings, evaluated new models for efficiency, and allowed engineers to monitor their AI usage and costs per hour.

Naga explained that companies are treating efficiency as an engineering problem rather than a budget problem, leading to reduced costs. While AI's rising ROI stakes are increasing for companies, the threat of Jevons paradox looms. This paradox suggests that spending on a resource, in this case tokens, increases even as its cost decreases.

Despite this, Naga emphasized a shift in company philosophy to prioritize quality over quantity in token spending, with the next phase of enterprise AI not focused on who spends the most tokens but on how people use them efficiently.