A new idea gaining traction among global and Chinese tech companies is putting tokens — the smallest unit of data processed by artificial intelligence models — at the center of compensation discussions, as firms seek to boost productivity in an AI-driven development era.
The concept suggests that, in addition to salaries, bonuses and equity, companies could allocate engineers a dedicated budget of AI tokens, which can be used to run AI agents, execute tasks and accelerate research and development.
The idea recently garnered attention in Silicon Valley. In late March, Charles Lamanna, executive vice-president with Microsoft, said in an interview that when responding to a job candidate who requested a specific allocation of AI tokens as part of their offer, he advised thinking in terms of daily usage ranging from $100 to several hundred dollars.
Nvidia Founder and CEO Jensen Huang said at the company's GTC conference that engineers could receive token allocations equivalent to roughly half of their base salary. Top engineers, he estimated, might consume up to $250,000 worth of tokens annually.
In addition, venture capitalist Tomasz Tunguz, founder of Theory Ventures, said inference costs are already emerging as a "fourth component" of engineering compensation, suggesting computing power should account for roughly one-fifth of a top-tier engineer's package.
Their Chinese counterparts are also moving in a similar direction. Major tech players including Alibaba, Tencent, UCloud and Kunlun Tech have introduced measures such as free token allocations or monthly subsidies to support employees' development needs.
According to Jiang Han, a senior analyst at market consultancy Pangoal, the idea of specifying token quotas in job listings is becoming increasingly feasible in the AI era.
"Leading companies are piloting token allocations as a form of productivity support, showing a shift in computing resources from an IT cost item to part of human capital allocation," Jiang said. "This kind of quantifiable productivity enhancement is evolving into a new employment condition that could be incorporated into recruitment frameworks."
Jiang added that token availability is becoming an important factor in attracting AI talent. "For core roles such as algorithm engineers and developers, sufficient token access means higher innovation and output efficiency. It has gone beyond a typical benefit to be a signal of a company's commitment to technological investment."
Industry practices are already reflecting this shift. In February, AI tech firm Kunlun Tech announced that it would provide all technical staff with access to AI coding tools such as OpenAI Codex and Claude Code, along with a monthly subsidy of $100 per account.
Fang Han, chairman and CEO of Kunlun Tech, said the company currently consumes between 1 trillion and 1.2 trillion tokens per month, with monthly spending on tokens reaching about 1.05 million yuan ($152,360).
Fang said the investment has delivered clear returns. "The output is highly measurable — development efficiency has increased by more than 50 percent, and for architects and team leaders, it can reach three to five times."
Similarly, Yao Jinbo, chairman of information platform 58.com, said in an earlier interview that his company encourages teams to maximize token usage despite cost pressures.
"I tell my team to use as many tokens as possible. Even if the cost pressure is high, we will not hold back on spending," Yao said, adding that the company currently consumes nearly 200 billion tokens per day, with that amount expected to soon exceed 300 billion.
Yao added that AI capability has become a key criterion in the company's talent evaluation. "If someone does not understand AI or lacks a forward-looking perspective, they will not be placed in a management position," he said.
Jiang, the senior analyst, noted that as AI agents become more widely used, competition for computing resources across departments such as R&D, product and operations is intensifying.
"This reflects the growing importance of computing power as a scarce strategic resource," he said, adding that as token costs become part of labor costs, companies need to establish clear frameworks to evaluate returns on such investments and avoid inefficiencies.
Jiang also noted that tokens differ fundamentally from traditional compensation components.
"Salaries and bonuses reward short-term performance, while equity aligns long-term interests. Tokens, however, focus on the efficiency of tool usage — they enhance output and function more as process-based incentives rather than outcome-based rewards," he said.
lijiaying@chinadaily.com.cn