Nvidia’s Blackwell Chips Made in Arizona Still Head to Taiwan for Final Assembly

Estimated reading time: 2 minutes

Nvidia has begun production of its next-generation Blackwell GPUs in the United States, but the company still depends heavily on Taiwan to complete the process, The Register reported.

CEO Jensen Huang marked the milestone on Oct. 17, celebrating the first Blackwell wafer produced at TSMC’s Fab 21 in Arizona. Speaking at an event in Phoenix, Huang praised TSMC’s capabilities while aligning with U.S. President Donald Trump’s “America First” manufacturing agenda.

“This is the vision of President Trump of reindustrialization — to bring back manufacturing to America, to create jobs, of course, but also this is the single most vital manufacturing industry and the most important technology industry in the world,” Huang said.

Despite the domestic production, Nvidia’s top-tier GPUs will still travel to Taiwan for advanced packaging. The company’s Blackwell chips used in the B100, B200, and GB200 accelerators contain two large computer dies and eight stacks of HBM3e memory, all connected using TSMC’s chip-on-wafer-on-substrate (CoWoS) technology.

Currently, all of TSMC’s CoWoS facilities are based in Taiwan. Amkor Technology, a U.S.-based chip assembly and testing company, is constructing an advanced packaging plant in Arizona in partnership with TSMC, but it will probably start operating in 2027 or 2028.

During TSMC’s Q3 earnings call in mid-October, CEO C.C. Wei confirmed that breaking ground on the Amkor site had only just begun. Until it’s ready, Nvidia’s U.S.-made wafers will probably continue to be shipped overseas for packaging.

Not all Nvidia chips require CoWoS technology, however. The RTX Pro 6000 workstation GPU and most RTX gaming cards use simpler designs capable of being manufactured entirely within the U.S.

In the longer term, Nvidia may also leverage Intel’s advanced EMIB and Foveros packaging technologies for future products.