Huang: 'Do more with more'
Jensen Huang argued at GTC that smart companies should ‘do more with more’—investing in people alongside AI—while also experimenting with token/compute budgets as novel compensation models. The remarks frame a broader debate about whether AI should shrink headcount or amplify skilled teams. (blogs.nvidia.com)
Huang proposed companies give engineers an annual "token budget" roughly equal to half an engineer’s base salary and said that budget could “amplify 10X” an engineer’s output. (sfgate.com)) He described tokens as a payable resource, noting that major providers charge for tokens (OpenAI’s top model costs about $15 per 1 million tokens), which he used to illustrate how token consumption can become a material line-item. (sfgate.com)) Huang tied the token economy to product demand, projecting about $1 trillion in orders for Nvidia’s Blackwell GPUs and the Vera Rubin platform through 2027 to support inference-heavy deployments. (cnbc.com)) Onstage he convened the CEOs of Mistral, Perplexity, Cursor and Thinking Machines Lab as an “open model vanguard” to align on agentic AI, inference pipelines and interoperable tooling for enterprise deployments. (blogs.nvidia.com)) Coverage from industry press framed the keynote as a shift from training-focused hardware to “AI factories” and token economics, with one outlet counting 45 explicit mentions of agentic AI during Huang’s remarks. (eweek.com)) Huang explicitly suggested token allotments will become a recruiting metric — quipping that candidates will ask “How many tokens come along with my job?” and that token pay could become standard alongside laptops and salary. (sfgate.com))
