Jensen Huang says AGI is already here in many real-world tasks, as Nvidia plans 400,000 AI agents

Nvidia CEO Jensen Huang said artificial general intelligence, or AGI, has already been achieved in many practical tasks, according to remarks he made on the Lex Fridman podcast and during the company’s Q2 2026 earnings call. Huang argued that AGI should be measured by what he called “profitable tokens” — the real economic output generated by AI systems — rather than by traditional cognitive benchmarks. He also disclosed that Nvidia is building a working model in which more than 400,000 AI agents operate alongside roughly 40,000 human employees across engineering, operations, and business functions. Separately, the company plans to deploy more than 1.5 million GPUs to build what Huang described as “AI factories,” framing intelligence production as an industrial process. The report said Huang’s framing differs from OpenAI’s formal definition of AGI, which describes it as systems that outperform humans on most economically valuable tasks. That definition also carries contractual implications tied to corporate governance.