Are AI tokens the new signing bonus or just a cost of doing business?

This week, a topic on the rise in Silicon Valley came into the spotlight: AI tokens as compensation. The idea is simple enough: instead of just giving engineers salary, equity and bonuses, companies would also give them a budget of AI tokens, the computing units that power tools like Claude, ChatGPT and Gemini. Issue them to manage agents, automate tasks, and deploy code. The premise is that access to more computing power makes engineers more productive, and that more productive engineers are worth more. The idea is that it is an investment in the person who holds them.

Jensen Huang, the leather jacket-wearing CEO of Nvidia, seemed to capture everyone’s imagination when he floated the idea earlier this week at the company’s annual GTC event that engineers should again receive around half of their base salaries – in tokens. His top people could, according to his calculations, burn out $250,000 per year in the field of AI computing. He called it a recruiting tool and predicted it would become standard in Silicon Valley.

It is not entirely clear where the idea first arose. Tomasz Tunguz, a renowned Bay Area venture capitalist who runs Theory Ventures and focuses on AI, data, and SaaS startups — and whose writing on all things data has gained a loyal following over the years — spoke about this in mid-February, writing that tech startups were already adding inference costs as a “fourth part to technical compensation.” Using data from compensation tracking site Levels.fyi, he estimated the salary of a top-quartile software engineer at $375,000. Add $100,000 in tokens and you’re at $475,000 fully loaded – meaning about one in five dollars has now been calculated.

That’s no coincidence. Agentic AI has taken off in a big way, and the release of OpenClaw in late January has accelerated the conversation significantly. OpenClaw is an open-source AI assistant designed to work continuously: cycle through tasks, spawn sub-agents, and work through a to-do list while the user sleeps. It’s part of a broader shift toward “agentic” AI, that is, systems that don’t just respond to cues but autonomously take series of actions over time.

The practical consequence is that token consumption has exploded. While someone writing an essay might use 10,000 tokens in an afternoon, an engineer managing a swarm of agents might blow through millions in a day – automatically, in the background, without typing a word.

This weekend the New York Times put together an article smart appearance in the so-called token maxxing trend, which finds engineers at companies like Meta and OpenAI competing on internal leaderboards that track token consumption. Generous token budgets are quietly becoming a standard job perk, the newspaper reported, as dental insurance or a free lunch once was. An Ericsson engineer in Stockholm told the Times that he probably spends more on Claude than he earns in salary, even though his employer foots the bill.

Maybe tokens really will become the fourth pillar of technical compensation. But engineers may want to toe the line before embracing this as a clear win. More tokens may mean more power in the short term, but given how quickly things are moving, it doesn’t necessarily mean more job security. For starters, a large allocation of tokens comes with large expectations. When a company effectively funds the computing power of a second engineer on your behalf, the implicit pressure is to produce twice as much (or more).

And beneath that lies an even more complicated problem: At the point where a company’s token spending per employee approaches or exceeds that employee’s salary, the financial logic of the workforce begins to look different to the finance team. If the computer does the work, the question of how many people have to coordinate this becomes harder to avoid.

Jamaal Glenn, an East Coast-based Stanford MBA and former VC, similarly became CFO for the financial services industry points out that what may seem like a perk can be a smart way for companies to increase the apparent value of a compensation package without increasing cash or equity – the things that actually increase for an employee over time. Your token budget is not definitively allocated. It doesn’t appreciate it. It doesn’t come up in your next offer negotiation the way base salary or a stock grant does. If companies successfully normalize tokens as compensation, they may find it easier to maintain cash comp, while pointing to growing compute compensation as evidence of investment in their people.

That’s a good deal for the company. Whether it’s a good deal for the engineer depends on questions that most engineers don’t have enough information to answer yet.