$100,000 to $150,000 in AI compute credits, handed to every engineer, on top of a base salary that already ranges from $200K to $300K. That's the new compensation model Jensen Huang laid out at Nvidia's GTC 2026 keynote last week.
The idea: give engineers an annual "inference budget" they can spend running AI models and agents to multiply their own output. Huang framed it as a "10x productivity amplifier," arguing that in the age of AI agents, an engineer's output depends less on hours worked and more on access to compute.
What This Actually Looks Like
A token is the basic unit AI models consume when generating text - roughly three-quarters of a word. Companies like OpenAI charge about $15 per million tokens. So $150K in credits buys a staggering amount of AI processing power for a single engineer to deploy.
The practical effect: Nvidia engineers can spin up AI agents to handle code generation, testing, documentation, data analysis, and other tasks that would otherwise eat their working hours. Huang's vision is that each human engineer manages a fleet of AI agents doing the grunt work.
And that fleet could get very large. Huang projected that within a decade, Nvidia could employ around 75,000 human workers (nearly double its current 42,000) alongside 7.5 million AI agents. That's a 100-to-1 ratio of agents to people.
The Recruiting Angle
This isn't just a productivity play. Huang explicitly positioned token budgets as a recruiting tool, saying compute credits are becoming "one of the recruiting tools in Silicon Valley." The logic is straightforward: if you're an engineer choosing between two $250K offers, the one that comes with $150K in AI compute on top gives you capabilities the other job doesn't.
Sam Altman picked up the thread separately, floating the idea that AI tokens could eventually function as a form of universal basic income - though that's a much longer-term and more speculative argument.
The Other Side of the Equation
Huang insists AI agents won't replace workers but will handle tasks humans don't need to do. The broader data tells a more complicated story. A recent survey found 98% of C-suite executives expect AI to lead to headcount reductions within two years, and about 65% of executives anticipate redeploying or reskilling 11% to 30% of their workforce.
The tension is obvious: Nvidia is simultaneously saying "we'll nearly double our headcount" and "each person will have 100 AI agents working for them." Both of those things might be true for Nvidia specifically, a company riding the biggest infrastructure boom in tech history. For companies that buy Nvidia's chips rather than sell them, the math could look very different.
Still, the compensation model itself is worth watching. If token budgets become standard at major tech companies, it changes what engineers optimize for. Instead of writing every line yourself, you're managing AI agents effectively - a fundamentally different skill set. The engineers who figure out how to deploy $150K in compute credits productively will be worth far more than those who don't.