The traditional formula of “salary + bonus + shares” in Silicon Valley has been supplemented with a new element — AI inference. Now companies are luring IT talent with access to computing power.
The traditional formula of “salary + bonus + shares” in Silicon Valley has been supplemented with a new element — AI inference. Now companies are luring IT talent with access to computing power.
As AI tools become an integral part of software development, the cost of running the underlying models — known as inference — is becoming a productivity driver and a budget line that executives can no longer ignore, Business Insider writes .
AI engineers and researchers at tech giants have long fought over access to GPUs. That computing power is strictly allocated to the highest-priority projects. Now, job candidates are openly asking how much AI budget they will get if they join the company.
“In interviews, candidates increasingly ask me how much dedicated inference capacity they will have to develop with Codex,” Thibaut Sottio, OpenAI Codex’s engineering lead, recently wrote in X.
He added that the intensity of resource use by a single user is growing much faster than the total number of users, indicating that AI computing is becoming increasingly scarce and valuable.
This shortage is changing how engineers perceive their work and pay. OpenAI President Greg Brockman put it bluntly: “The amount of inference power available to you will increasingly determine your overall software development productivity.”
In other words, access to AI could soon be as important as a high salary or a hefty stock package. For a developer in the AI era, not having access to large-scale computing means risking producing far less than their peers, jeopardizing their career prospects.
Hakim Shibli, a data scientist at Levels.fyi, recently noticed an interesting detail in an engineer’s income statement: a subscription to Copilot was listed as part of the compensation package. It’s a small but symbolic step toward making access to AI a standard perk.
The developer community predicts the emergence of new hiring standards.
According to Peter Gostev, head of AI capabilities at startup Arena, giants like OpenAI and Anthropic should launch their own job search services. Each vacancy should include, in addition to the salary range, a token limit available to the future employee.
Investors have already taken notice. Tomasz Tunguz of Theory Ventures noted that companies are effectively adding AI inference as a fourth component of engineer compensation: salary, bonus, stock, and now tokens.
Tokens are the economic language of generative AI. Models break down words and other inputs into numerical tokens for easier processing and understanding. One token is about ¾ of a word. They are also used to calculate the cost of running AI models using a commonly accepted standard — the cost per million tokens.
“Will you be paid in tokens? In 2026, it is likely that you will start receiving them,” Tunguz concluded.
For CFOs, this potentially huge expense item needs to be tracked as meticulously as personnel costs, Tunguz noted. “It’s already becoming a reality,” he said, as AI use by employees increasingly impacts the overall velocity of spending across the company.
Since top developers earn about $375,000 per year, Tunguz calculates that if you add $100,000 to pay for AI power, the company's total costs for such an employee will be $475,000. Thus, a fifth (21%) of an engineer's "cost" is no longer money in his hands, but payment for his computing resource.
If this trend continues, 2026 could be the year when engineers negotiate pay not only in dollars and shares, but also in tokens.
Recall that Nvidia CEO Jensen Huang was one of the first CEOs to propose tokens as a strategy to attract specialists .
