Silicon Valley has a new idea for keeping engineers happy: pay them in AI tokens. TechCrunch AI reports that the concept has been gaining serious momentum, with Nvidia CEO Jensen Huang floating the idea at GTC that engineers should receive roughly half their base salary in compute tokens, meaning his top people might burn through $250,000 a year in AI compute alone.
The pitch sounds simple. Give engineers a budget of computational units for tools like Claude, ChatGPT, and Gemini. Let them run agents, automate tasks, crank through code. More compute equals more productivity, and more productive engineers are worth more. That’s the theory, anyway.
📊 The numbers are already real
VC Tomasz Tunguz of Theory Ventures flagged this trend back in mid-February, according to TechCrunch AI. He noted that startups were already treating inference costs as a “fourth component to engineering compensation.” Using data from Levels.fyi, he put top-quartile engineer pay at $375,000. Add $100,000 in tokens and you’re at $475,000 fully loaded. One dollar in five is now compute.
The New York Times found that engineers at Meta and OpenAI are competing on internal leaderboards tracking token consumption. One Ericsson engineer in Stockholm reportedly spends more on Claude than he earns in salary, with his employer picking up the tab.
🤖 Why this is happening now
Agentic AI changed the math. The release of OpenClaw in late January, an open-source AI assistant that runs continuously, spawning sub-agents and working through tasks autonomously, accelerated token consumption dramatically. An engineer writing an essay might use 10,000 tokens in an afternoon. An engineer running a swarm of agents can blow through millions in a day, automatically, without typing a word.
This isn’t a hypothetical trend. It’s already reshaping how companies think about the cost of a productive engineer.
⚠️ The uncomfortable flip side
What stands out here is the tension nobody in the “tokens as compensation” camp wants to talk about. TechCrunch AI highlights several sharp counterpoints:
- Higher expectations, not just higher budgets. If a company funds a second engineer’s worth of compute on your behalf, the implicit pressure is to produce at twice the rate. Token budgets aren’t free: they come with performance assumptions baked in.
- The headcount math gets ugly. When a company’s token spend per employee approaches or exceeds that employee’s salary, finance teams start asking a different question: how many humans do we actually need coordinating this compute?
- Tokens don’t vest. Jamaal Glenn, a former VC turned financial services CFO, makes the sharpest point: your token budget doesn’t appreciate, doesn’t compound, and won’t show up in your next offer negotiation. Companies could use token budgets to keep cash comp flat while pointing to growing compute allowances as “investment in their people.”
That’s a good deal for the company. For engineers, it’s murkier.
🔮 What comes next
Expect token budgets to become standard at AI-forward companies within 12-18 months. The trend has too much momentum, and too much financial incentive for employers, to stall. But smart engineers should treat this the way they’d treat any compensation component: negotiate the things that compound (salary, equity) separately, and view tokens as a tool, not a trophy.
The broader signal is worth watching. When companies start measuring engineer value by how much compute they consume rather than what they ship, the relationship between human workers and AI systems shifts in ways that favor the machines. That’s not a crisis today. But it’s the kind of structural change that sneaks up on an industry while everyone’s busy tokenmaxxing.
More details on this trend are available in the original TechCrunch AI report.