Nvidia CEO Jensen Huang has recently put forth a novel concept: offering a new form of employee benefit to the company's engineers – AI tokens. These tokens, potentially worth tens of thousands of dollars, are envisioned to become a form of 'currency' and a significant incentive tool in the future AI-driven workplace.
Huang unveiled this concept during the company's annual GPU Technology Conference. He pointed out that AI tokens, as the fundamental unit for AI systems to perform tasks, are increasingly becoming a new method for Silicon Valley to attract talent. This vision aligns with his outlook on future work scenarios: future employees will be able to manage vast teams of AI agents, software programs capable of independently completing complex, multi-step tasks.
Huang predicts that Nvidia's current workforce of 42,000 employees will eventually be surpassed by 'hundreds of thousands' of 'digital employees.' In this transformation, data centers will be what Huang calls 'AI factories,' operating similarly to traditional factories producing goods, except they will produce AI tokens.
He vividly likened it: 'If compute is the printing press, then tokens are the real currency of the AI era.' He further explained that computing power is akin to current income, forming the basis for generating tokens, and without tokens, the development and application of AI will stagnate.
Technological Iteration and Cost Considerations
Regarding Nvidia's leap in chip technology, Huang stated it was a strategic decision made while the existing Hopper architecture was still performing exceptionally well. He lauded the Grace Blackwell platform as a reliable cornerstone for enterprises in private cloud or global deployments. He emphasized that as companies build gigawatt-scale data centers, their power capacity is largely fixed, making efficiency improvements crucial. He concluded: 'Your workload is inference, your tokens are the commodity, and compute is your revenue. Every company will be focused on token efficiency.'
Looking ahead, Huang predicts that AI services will trend towards a tiered pricing model, ranging from free entry-level services to premium packages costing up to $150 per million tokens.
Concurrently, tech giants are actively strategizing. Alibaba recently restructured parts of its business, establishing the 'Taotian Group' led by CEO Eddie Wu, aiming to integrate all AI products and focus on token construction, delivery, and application.
Challenges and Opportunities Coexist
However, this transition is not without its challenges. A survey indicates that 98% of executives believe AI will ultimately lead to job cuts, yet 54% still identify recruiting qualified talent as the biggest hurdle. For startups, the cost of AI applications is beginning to manifest, with each AI task completion consuming tokens, and several mainstream AI service providers have recently increased their prices by 5% to 30%.
Bruno Guicardi, President of IT company CI&T, pointed out that the current trend is for engineers to drive computers with natural language commands rather than writing code. In this environment, understanding when and how to effectively utilize AI and derive genuine value from it will become paramount.
(Note: This content is compiled from public information and does not constitute investment advice.)