This week, a topic that has been circulating in Silicon Valley moved into the spotlight: AI tokens as compensation. The idea is straightforward. Rather than giving engineers only salary, equity, and bonuses, companies would also provide them with a budget of AI tokens. These are the computational units that power tools like Claude, ChatGPT, and Gemini. Engineers could spend these tokens to run AI agents, automate tasks, or process code. The pitch is that access to more computing power makes engineers more productive, and more productive engineers are more valuable. It is framed as an investment in the person.
Jensen Huang, the CEO of Nvidia, captured widespread attention when he suggested at the company’s annual GTC event that engineers should receive an additional amount roughly equal to half their base salary in the form of tokens. He estimated that his top engineers might use two hundred and fifty thousand dollars worth of AI compute annually. He described this as a recruiting tool and predicted it would become standard across Silicon Valley.
The origin of the idea is not entirely clear. Tomasz Tunguz, a venture capitalist focused on AI, data, and SaaS startups, was discussing it in mid-February. He wrote that tech startups were already adding inference costs as a fourth component to engineering compensation. Using compensation data, he illustrated that a top-quartile software engineer salary of three hundred seventy-five thousand dollars, plus one hundred thousand dollars in tokens, results in a total of four hundred seventy-five thousand dollars. This means roughly one dollar in five is now allocated to compute.
This trend aligns with the rise of agentic AI. The release of OpenClaw in late January accelerated the conversation considerably. OpenClaw is an open-source AI assistant designed to run continuously, working through tasks and spawning sub-agents autonomously. This shift toward systems that take sequences of actions over time has a practical consequence: token consumption has exploded. While someone writing an essay might use ten thousand tokens in an afternoon, an engineer running a swarm of agents can use millions in a day automatically.
By this weekend, major publications were examining the so-called tokenmaxxing trend. Reports found that engineers at companies like Meta and OpenAI are competing on internal leaderboards that track token consumption. Generous token budgets are quietly becoming a standard job perk, similar to dental insurance or free lunch once were. One engineer in Stockholm noted he likely spends more on AI compute than he earns in salary, with his employer covering the cost.
Perhaps tokens will become the fourth pillar of engineering compensation. However, engineers might want to consider the implications before embracing this as a straightforward win. More tokens may mean more power in the short term, but it does not necessarily mean more job security. A large token allotment comes with large expectations. If a company is funding a second engineer’s worth of compute on your behalf, the implicit pressure is to produce at twice the rate.
A deeper issue exists. When a company’s token spend per employee approaches or exceeds that employee’s salary, the financial logic of headcount changes for its finance team. If the compute is doing the work, the question of how many humans are needed to coordinate it becomes harder to avoid.
Financial analysts point out that what may seem like a perk can be a way for companies to inflate the apparent value of a compensation package without increasing cash or equity. These are the things that actually compound for an employee over time. A token budget does not vest, appreciate, or factor into future offer negotiations like a base salary or equity grant does. If companies successfully normalize tokens as pay, they may find it easier to keep cash compensation flat while pointing to a growing compute allowance as evidence of investment.
That is a good deal for the company. Whether it is a good deal for the engineer depends on questions most engineers do not yet have enough information to answer.