AI compute costs now exceed employee salaries at major tech companies — Uber burned its full 2026 AI budget on token costs alone

Uber's CTO exhausted the entire 2026 AI budget due to token costs (charges for each unit of data an AI processes), according to The Information. Bryan Catanzaro, vice president of applied deep learning at Nvidia, confirmed the trend: 'For my team, the cost of compute is far beyond the costs of the employees.' Anthropic and OpenAI have both raised prices in response to surging demand.

The shift is reshaping how companies view AI ROI (return on investment). Enterprises now face pressure to prove that spending on AI subscriptions and infrastructure delivers measurable productivity gains or cost savings — otherwise, heavy AI spending becomes a liability rather than a strategic advantage. This means AI projects will face the same quarterly earnings scrutiny that employee headcount does.

For business leaders, this signals that AI is no longer cheap or experimental. If your company is piloting generative AI agents or large-scale language models, expect token costs to compound faster than headcount would. The calculus on whether to hire humans or automate with AI just flipped: the automation option is now the expensive one.