Earlier this week, a discussion that’s been circulating in Silicon Valley came into focus: AI tokens as part of compensation. The idea is simple — aside from providing salary, equity, and bonuses, companies would also allocate AI tokens to engineers. These tokens, which power tools like Claude, ChatGPT, and Gemini, could be used to run tasks and automate coding. The premise is that having more compute resources boosts engineer productivity, thereby increasing their value.
Jensen Huang, CEO of Nvidia, sparked interest by suggesting at the GTC event that engineers should receive tokens worth half their salary. Based on his calculations, top engineers might use up to $250,000 annually in AI compute. He described it as a recruitment incentive and predicted it would become standard in Silicon Valley.
The origins of this concept are unclear. Tomasz Tunguz, a well-known VC in the Bay Area focused on AI and data startups, discussed it in February. He noted that startups were adding inference costs as a “fourth component” of compensation. He cited data showing a top software engineer’s salary at $375,000, with an additional $100,000 in tokens bringing the total package to $475,000.
Agentic AI has been rapidly gaining traction, with OpenClaw’s release in January fueling the discussion. OpenClaw, an open-source AI assistant, can execute tasks autonomously, contributing to the shift toward “agentic” AI, where systems act independently over time.
Consequently, token consumption has surged. An engineer might use millions of tokens daily while running multiple agents in the background.
By weekend, the New York Times highlighted the tokenmaxxing trend, noting engineers at companies like Meta and OpenAI compete on token consumption leaderboards. Generous token allocations are becoming a standard perk, similar to dental insurance or free lunches. One Ericsson engineer mentioned he likely spends more on Claude than his salary, though his employer covers the cost.
While tokens might become a key component of compensation, engineers should be cautious. More tokens could equate to immediate power but may compromise job security. A large token allocation comes with considerable expectations. If companies fund significant compute resources per engineer, the implicit pressure is to deliver at an increased rate.
Furthermore, when token expenses per employee rival their salary, CFOs may reconsider headcount. As compute handles tasks, the necessity for human coordination becomes less clear.
Jamaal Glenn, a former VC and CFO, points out that offering tokens may inflate compensation value without increasing cash or equity, which benefit the employee over time. Tokens don’t vest or appreciate like salaries or equity and might not impact future negotiations.
For companies, this is advantageous. However, whether it benefits engineers depends on questions they may not yet be able to fully answer.