NVIDIA CEO Huang Renxun estimated that $50–60 billion is needed for 1 GW of AI infrastructure at the G20.

ChainCatcher report: On Wednesday, U.S. local time, NVIDIA CEO Jensen Huang appeared at the G20 Ministerial Meeting on Innovation, engaging in a fireside chat with U.S. Secretary of Commerce Luotonick. Huang stated that AI is evolving into a national-level economic infrastructure akin to electricity and the internet, and that the greatest risk countries face is failing to invest in and adopt AI sufficiently, ultimately being left behind by the industrial revolution. He noted that building 1 gigawatt of AI infrastructure currently requires an investment of approximately $50 to $60 billion, and projected that by the end of this decade, total construction scale could reach around 100 gigawatts. He also referenced the “five-layer cake” model—energy, chips, infrastructure, models, and applications—emphasizing that every country must build its own AI infrastructure and decide which layers it wishes to participate in. Huang anticipates that AI will largely achieve AGI within the next few years, with the next phase transitioning from large language models to agents and embodied intelligence. He believes AI is more likely to replace tasks rather than entire jobs, with operational tasks such as writing and information processing gradually becoming automated.