Nvidia has announced plans to manufacture some of its AI chips and supercomputers in the United States, marking a significant shift in its production strategy. The company is partnering with TSMC, Foxconn, Wistron, Amkor, and SPIL to establish manufacturing facilities in Arizona and Texas. TSMC has already begun producing Nvidia's Blackwell chips in Phoenix, Arizona. Nvidia is also constructing supercomputer manufacturing plants in Houston and Dallas with Foxconn and Wistron, respectively, with mass production expected to ramp up in 12-15 months. Amkor and SPIL will handle packaging and testing in Arizona.
Nvidia anticipates investing up to $500 billion in US AI infrastructure over the next four years. This move aims to address the increasing demand for AI chips, strengthen the supply chain, and enhance resilience. The company will leverage its AI, robotics, and digital twin technologies, including Nvidia Omniverse and Isaac GR00T, to design and operate these facilities. This initiative is projected to generate hundreds of thousands of jobs and drive trillions of dollars in economic security. The announcement aligns with the US government's push for increased domestic manufacturing amid ongoing trade tensions and tariffs.
Jensen Huang, Nvidia's CEO, stated that building AI infrastructure in the US would help meet growing demand and bolster supply chain resilience. These US-made AI supercomputers will power a new generation of AI-focused data centres, or "AI factories."