Nvidia, a dominant force in artificial intelligence hardware, has strategically relocated some of its manufacturing processes to the United States. The company disclosed that it has commenced the production of its cutting-edge Blackwell AI graphics processors (GPUs) at the TSMC facility in Arizona. Additionally, the packaging and testing of these sophisticated chips are being managed by Amkor and SPIL within the state.
Based in Taiwan, TSMC (Taiwan Semiconductor Manufacturing Company), the preeminent global chip manufacturer, recently announced a substantial $100 billion investment plan for bolstering chip production in the U.S. Since January, the Arizona plant has been successfully fabricating chips using the 4nm process. The company has ambitious plans to introduce more advanced 2nm technology to the U.S. market before the decade’s end, although Taiwanese facilities are expected to adopt this technology sooner.
While Nvidia has not specified which Blackwell chip models are being produced, its website indicates that these chips utilize TSMC’s 4NP production process. It remains uncertain if the production includes the Blackwell Ultra GB300, Nvidia’s most powerful model announced in recent months.
Nvidia’s initiatives in the U.S. extend beyond chip production. The company is collaborating with electronics powerhouses Foxconn and Wistron to construct two distinct supercomputer manufacturing facilities in Texas. According to the announcement, mass production at these sites (potentially for DGX Spark and DGX Station) will intensify over the next 12 to 15 months.
Nvidia’s strategic shift aligns with the U.S.’s objective to reduce reliance on overseas semiconductor technologies. Amidst ongoing trade tensions with China, the importance of cultivating domestic critical technology infrastructure has increased. The Trump administration, in particular, is exerting pressure to repatriate manufacturing processes to the U.S., which may inadvertently lead to increased costs and higher global prices.
Yorum Yap