NVIDIA Offers Massive AI Tokens to Attract Its Engineers

NVIDIA and the AI Token Initiative

Jensen Huang, the CEO of NVIDIA, recently announced an ambitious initiative to allocate each engineer in the company an AI token budget equivalent to half of their annual salary. With a workforce of 42,000 employees, this measure represents a colossal investment of nearly $2 billion.

According to Huang, AI tokens are quickly becoming a sought-after recruitment tool in Silicon Valley. He stated, "I can easily imagine that every engineer in our company will need an annual token budget. They will earn a few hundred thousand dollars a year in base salary. I will probably give them half of that again, in the form of tokens, so that they are amplified tenfold." For a software or AI engineer with a salary between $200,000 and $300,000, this translates to an additional annual computing credit of between $100,000 and $150,000, or more than $100 per day of available inference.

OpenCLAW and the Rise of Open Source

At the 2026 GPU Technology Conference (GTC), Jensen Huang also introduced OpenCLAW, an open-source agent framework. He considers this project to be the fastest adopted in history, surpassing the adoption curve of Linux over thirty years in just a few weeks. In a work environment where agents autonomously break down problems, generate sub-agents, execute tools, and run experiments overnight without human intervention at every step, an engineer with a pre-allocated token budget can launch a research task and return to the results. In contrast, an engineer who must validate each computing expense in advance can no longer work in this manner.

Tokens, a New Asset in Job Offers

AI tokens are beginning to feature in hiring discussions, alongside salary, bonuses, and stock options. Candidates are now interested in access to computing resources, which is becoming an essential competitive resource. However, to establish a token budget per employee, each company must know in advance which models its teams will use, at what volume, and for what purposes. Prices can vary from one to three times depending on the provider, model, and volume. Without an internal governance framework, a token budget can quickly lead to uncontrolled overconsumption, much like an unlimited data plan.

NVIDIA and the Future of AI Agents

Jensen Huang envisions that NVIDIA will reach 75,000 human employees accompanied by 7.5 million AI agents over the next ten years, resulting in a ratio of 100 agents for each employee. However, if an engineer with a token budget produces ten times more, companies no longer need to hire ten times more engineers. Goldman Sachs estimates that AI could automate 25% of work hours in the United States, with a displacement of 6 to 7% of jobs during the adoption period. However, about 80 to 85% of AI projects have failed since 2018, raising doubts about the feasibility of a massive deployment of agents at this scale.

NVIDIA reported $215.9 billion in revenue for fiscal year 2026, up 65% year-over-year, with $62.3 billion coming from data centers alone. Each token consumed by an engineer runs on GPUs, and nearly all of these GPUs bear the NVIDIA logo. Jensen Huang is thus simultaneously selling a recruitment model, a promise of productivity, and a market for his own chips.

Additionally, NVIDIA has invested hundreds of billions of dollars in chip infrastructure in the United States over four years, which is the physical infrastructure that its own agents will prioritize consuming.