Jensen Huang pushes "Tokens in paychecks" — is it generosity or a strategic trap?

The pitch

At Nvidia (英伟达) CEO Jensen Huang's GTC keynote, he reportedly sketched a radical re-think of engineering pay: add large annual "Token" budgets on top of base salary to fuel continuous AI agent runs. Huang said an engineer earning $300,000 might receive about $150,000 in Token credits; a $500,000 engineer might get roughly $250,000. The logic is simple: Tokens are the new fuel for AI work, and giving them to engineers multiplies output — at least according to the company line. It has been reported that Nvidia is targeting roughly $2 billion a year in Token subsidies as part of that push, even as the firm projects massive orders for its next-generation chips.

Critics cry foul

But not everyone sees this as a win for employees. TechCrunch ran a blunt critique asking: are Tokens a new sign-on bonus or just a transfer of operating cost? Reportedly, former investor–turned-CFO Jamaal Glenn warned Tokens are not an asset the employee can carry to the next job — they vanish once spent and don’t appear on offers or resumes. Journalistic reporting from the New York Times and others has also flagged internal "Token leaderboards" and instances of engineers being celebrated for burning huge amounts of compute — a practice dubbed "Tokenmaxxing." The concern is twofold: Tokens can normalize ever-higher consumption as a proxy for productivity, and they make firms’ payrolls look bigger in labor terms while shifting cash compensation pressure into variable compute budgets.

China is already experimenting

It has been reported that major Chinese tech groups are piloting similar moves. Tencent (腾讯) reportedly rolled out Token packages worth about ¥220,000 per employee; Alibaba (阿里) is bundling internal AI tool access and reimbursements; ByteDance (字节) and cloud providers such as Alibaba Cloud (阿里云) and Tencent Cloud (腾讯云) are selling low-cost "Coding Plan" subscriptions to blunt per-Token billing anxiety. But there is a difference: domestic packages typically bind usage to each vendor’s models and clouds, while Huang’s vision effectively channels demand toward GPUs — and thus toward Nvidia’s hardware footprint.

Why this matters geopolitically and economically

This is not just an HR debate. If Tokens become a normalized pillar of compensation, they reshape procurement incentives and could accelerate demand for GPU capacity — a tailwind for Nvidia, which already controls an outsized share of the AI chip market. It also raises regulatory and geopolitical questions: export controls, supply concentration and national strategies for AI infrastructure become more salient when compensation systems effectively lock engineering organizations into particular suppliers. So is this a bold way to supercharge engineering, or a strategic maneuver that shifts costs and binds ecosystems? The answer will shape hiring, procurement and corporate accounting for years to come.