Uber CTO Says ‘Tokenmaxxing’ Era Is Ending After AI Budget Burnout
In April, Uber revealed it had already exhausted its entire AI budget planned for 2026. The company’s AI programming costs had surged faster than expected due to the increased adoption of Claude Code. Subsequently, Uber reportedly limited AI usage to try to curb expenses. Now, eight months into the year, Naga states that the company is no longer squandering money on AI; instead, Uber has become more efficient, using AI tokens more intelligently and keeping costs under control. In fact, Naga views this as a sign that the era of “tokenmaxxing” is drawing to a close.
In a post shared on X following the second-quarter earnings call, CTO Praveen Neppalli Naga noted that the company’s AI strategy is shifting: moving away from massive token consumption to focus on more efficient usage. “As our CFO mentioned today during the earnings call, we are seeing very interesting trends in AI costs,” Naga wrote. “I think it’s another sign that we are nearing the end of the so-called ‘tokenmaxxing’ era.”
Naga states that Uber is now experiencing a very different trend. Although the number of employees using advanced AI tools has more than quadrupled since the beginning of the year, the company’s AI costs—he reveals—are actually decreasing. “You would expect costs to rise as adoption accelerates. However, we have seen the opposite,” he wrote. “The next phase, whatever we call it, won’t be defined by who spends the most tokens, but by how people use them most efficiently.”
As our CFO @_balaji_km mentioned at earnings today, we’re seeing some very interesting trends on AI costs. I think it’s another signal that we’re coming to the end of the so-called ‘tokenmaxxing’ era.
Here’s what’s been happening behind the scenes.
Since the beginning of the…
— Praveen Neppalli (@praveenTweets)
August 5, 2026
Uber claims to have controlled AI spending through efficiency
Naga reveals that the reduction in AI costs is not the result of restricting access to the technology. Instead, Uber has focused on helping employees use AI more efficiently. The company has cut down on unnecessary AI requests, improved prompt reuse, given engineers greater visibility into AI usage and spending, and tested different models to find the most cost-effective option for each task.
“The next phase, whatever we call it, won’t be defined by who consumes the most tokens, but by how people use them most efficiently,” he added.
Naga’s comments come months after Uber revealed—during its first annual earnings call—that it had already exhausted its 2026 budget for Anthropic’s Claude Code following a surge in AI adoption among its engineering teams. The incident also served as a wake-up call for the tech industry, highlighting the true financial cost of deeply integrating AI into developers’ daily workflows. It also prompted companies like Uber and Microsoft to rethink whether higher AI spending actually translates into greater productivity.
However, Uber claims to have found a way to maximise AI usage without costs spiraling out of control. Rather than simply encouraging employees to use AI more, the company is now focusing on achieving better results while spending less.
Uber is also taking AI beyond the realm of engineering. According to a Business Insider report, the company has begun integrating AI-expert engineers into non-technical teams—such as finance, legal, human resources, marketing, and customer service—to redesign entire workflows around this technology. According to Naga, the initiative has already yielded significant benefits, reducing the duration of certain processes from hours (and, in some cases, days) to just a few minutes.