Silicon Valley is buzzing with a new idea for compensating tech talent: AI tokens. Instead of just salary, equity, and traditional bonuses, companies are considering offering engineers a budget of AI tokens. These tokens represent computational power, essential for running advanced AI tools like Claude, ChatGPT, and Gemini. The concept hinges on the belief that providing engineers with ample compute resources will significantly boost their productivity, making them more valuable assets to a company.

The notion gained significant traction after Jensen Huang, CEO of Nvidia, suggested at the company’s GTC event that engineers should receive compute tokens equivalent to roughly half their base salary. He posited that top engineers might utilize upwards of $250,000 annually in AI compute. Huang framed this as a crucial recruiting advantage and predicted it would soon become a standard practice across the tech industry, transforming how engineering talent is attracted and retained.

While Huang’s endorsement brought the idea into the mainstream, venture capitalist Tomasz Tunguz was discussing similar concepts earlier this year. In mid-February, Tunguz wrote about tech startups already incorporating AI inference costs as a core component of engineering compensation. He cited data suggesting that when adding an estimated $100,000 in tokens to a top engineer’s $375,000 salary, the total compensation package reaches $475,000, with compute costs making up about a fifth of the total.

This trend seems closely tied to the rise of agentic AI, systems designed to perform tasks autonomously. The recent release of OpenClaw, an open-source AI assistant capable of continuous operation and task delegation, has amplified these discussions. Such tools can manage to-do lists and even spawn sub-agents, highlighting the potential for AI to fundamentally change how work is done and how engineers interact with technology.

The core question remains whether these AI tokens represent a genuine enhancement to compensation, akin to a signing bonus, or simply an operational cost that companies will eventually absorb as standard business expenses. As agentic AI capabilities expand and compute becomes more accessible, the integration of AI tokens into compensation packages is likely to evolve rapidly, shaping the future of tech employment and productivity.