Uber Ends Tokenmaxxing After Burning 2026 AI Budget In Weeks
KEY POINTS
- •Uber CTO Praveen Neppalli Naga announced their tokenmaxxing era is ending after quadrupling AI tool use this year.
- •By August 2026, Uber lowered AI costs through better prompt caching, default models, and open-weight model experiments.
- •COO Andrew Macdonald and CFO Balaji Krishnamurthy agreed productivity gains from AI are improving, but cost justification remains tough.
In a tale straight from Silicon Valley’s 'spend like there's no tomorrow' playbook, Uber CTO Praveen Neppalli Naga announced in April 2026 that the company ceremoniously nuked its Anthropic Claude Code budget by ‘tokenmaxxing’ wild AI usage—quadrupling frontier AI tool adoption but making CFO Balaji Krishnamurthy worry about the math as tokens rained like confetti. By August 9 at the HumanX Conference, Naga shifted gears praising ‘prompt caching’ and ‘open-weight models’ like deluxe cost-cutting magic, while COO Andrew Macdonald admitted it’s getting harder to justify the AI token party as productivity gains play hard to get. Meanwhile, the industry scrambles to go from reckless to refined smarter spending—even Coinbase juggling 'model-switching' like a circus act. Tokenmaxxing: once Uber’s favorite pastime, now a cautionary Silicon Valley bedtime story.
Share the Story
(1 of 3)Source: Businessinsider | Published: 8/6/2026 | Author: Aditi Bharade