Your AI Budget Is Too Small

At NVIDIA's GTC 2026, Jensen Huang sat down with the All-In Podcast hosts and dropped a number that should make every enterprise leader uncomfortable: a $500,000 engineer should be consuming at least $250,000 worth of AI tokens annually. Not as a perk. Not as an experiment. As baseline infrastructure.

"If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed."

Think about that for a moment. We're not talking about getting Copilot licences for employees here. We're talking about arming knowledge workers with superhuman powers and setting expectations to effectively force them to create and manage multiple AI agents at scale.

Is your organisation treating AI spend as a line item to minimise? Does it sit between software licences and office supplies? Huang is making a completely different argument: AI tokens are core infrastructure. The same category as compute, not stationery. And if your people aren't consuming them at scale, they're operating with their hands tied behind their backs.

He went further. When Anthropic CEO Dario Amodei projected AI could generate hundreds of billions in revenue by 2027/28 and reach a trillion dollars by 2030, Huang called those numbers "very conservative." His reasoning: every enterprise software company will eventually become a value-added reseller of tokens from the likes of Anthropic and OpenAI. The token economy isn't emerging. It's already here.

The $250K figure, let's call it 50% of an employee's salary, is provocative. But it's directionally right for knowledge workers in engineering, design, and product. CAD tools didn't replace chip designers. They made the ones who adopted them dramatically more capable, and made the ones who didn't basically unemployable. The same pattern is playing out right now across every knowledge discipline.