Nvidia VP: Compute costs now exceed employee salaries — Uber burned 2026 AI budget in 4 months
The company building AI chips says chips cost more than people — and Uber just proved it.
Nvidia's VP of applied deep learning dropped a bombshell: for his team, the cost of compute now exceeds what they pay their people. This isn't a startup struggling with GPU bills — it's the company that builds the chips powering the entire AI industry. The admission signals that even at the source, AI inference and training costs are ballooning faster than headcount expenses.
Uber's CTO independently confirmed the math. The company's entire 2026 AI coding budget was burned through by April — just four months into the year. Individual engineers were racking up $500 to $2,000 per month in token costs alone (prompts, not licenses or hardware). With AI agents and coding assistants becoming ubiquitous, token consumption is exploding. The question now: Can pricing models shift before AI adoption hits a cost wall?
- Nvidia's VP says compute costs now exceed employee salaries at the chipmaker itself.
- Uber exhausted its full 2026 AI coding budget by April, spending $500–$2,000 per engineer monthly on tokens.
- Token pricing models face pressure as AI usage scales — prompts alone are outpacing traditional software costs.
Why It Matters
AI cost structures may force a pricing revolution — enterprise budgets are already breaking under token consumption.