Uber technology chief exhausts annual AI budget on soaring token costs

Uber’s Chief Technology Officer has reportedly exhausted his entire 2026 artificial intelligence budget before the year was well underway, with the expenditure attributed entirely to token costs rather than hardware or recruitment.

This extreme case highlights a significant shift in corporate IT departments globally, where the financial burden of processing power is rapidly escalating.

Bryan Catanzaro, Vice President of applied deep learning at Nvidia, informed Axios that for his specific team, the cost of compute has already surpassed the cost of employing staff. This inversion marks a radical change from the economic landscape of just two years ago.

The rising costs have even become a point of public discussion for some industry leaders. Amos Bar-Joseph, CEO of Swan AI, recently shared his substantial Anthropic bill on LinkedIn, framing the high cost as a "badge of honour" in his quest to build an autonomous business.

"We're building the first autonomous business scaling with intelligence, not headcount," he wrote. However, the sustainability of this trade-off remains a subject for investor scrutiny as quarterly earnings calls approach.

According to research firm Gartner, worldwide IT spending is forecast to reach $6.31 trillion in 2026, a 13.5 percent increase from 2025.

This growth is driven by AI infrastructure and cloud services. Brad Owens, vice president at Asymbl, noted that boards are now questioning the "true value of a worker" in this digital age.

The rivalry between OpenAI and Anthropic further complicates matters, as differences in the token economy between tools like Codex and Claude Code begin to impact enterprise margins.