Nvidia and TSMC produce the first Blackwell wafer made in the U.S. — chips still need to be shipped back to Taiwan to complete the final product
11 hours ago
11
(Image credit: Nvidia)
Nvidia and TSMC on Friday announced they had reached a significant milestone with the manufacturing of the first production Blackwell wafer at Fab 21 near Phoenix, Arizona. The achievement of making one of the most complex chips in history in the U.S. has strategic, symbolic, and political importance for both companies, but there is a major catch.
"This is a historic moment for several reasons," said Jensen Huang, chief executive of Nvidia, at the commemorative event. "It is the very first time in recent American history that the single most important chip is being manufactured here in the United States by the most advanced fab, by TSMC, here in the United States. This is the vision of President Trump of reindustrialization — to bring back manufacturing to America, to create jobs, of course. Still, also, this is the single most vital manufacturing industry and the most important technology industry in the world."
Production of Nvidia Blackwell B300 silicon (we presume these are the B300 chiplets, though as it is unconfirmed) in the U.S.— one of the most complex chip ever created — using TSMC's 4N fabrication process (a custom 4nm-class node for Nvidia) means that TSMC's Fab 21 is capable of making such large chips presumably with good yields (which apparently took some time to achieve), which may indicate that the fab is indeed as capable as ones that TSMC runs in Taiwan (something that we have already heard from TSMC's own comments, but those were made about smaller chips).
Strategically, it fulfills a key goal of the U.S. government's industrial policy: bringing more advanced semiconductor manufacturing to American soil. Therefore, Nvidia can now claim that it produces one of its most essential products in the U.S. and avoid tariffs if they are imposed on Taiwan-made goods. The same applies to TSMC as it spreads its production across the world, reducing risks for its business from potential interruptions caused by China. Eventually, TSMC is set to build Fab 21 phases that will produce chips on N3, N2, A16, and A14 process technologies in Arizona.
For decades, Nvidia — the world's most valuable company with a market capitalization of $4.5 trillion — has produced its chips exclusively in Taiwan with a brief production of the NV40 silicon at IBM's microelectronics fab in New York. Producing its key AI GPUs in the U.S. has a deep symbolic meaning for Nvidia, which is an American company.
For the U.S., the achievement has a deep political meaning too. For decades, the world's most sophisticated chips were developed in the U.S. and then almost exclusively made in Taiwan, which created heavy dependence on a nation located in a high-risk geopolitical region. Producing Blackwell — the most advanced and demanded AI GPU — in the world, in the U.S., now gives Washington a tangible outcome from years of subsidies and incentives under the CHIPS Act (and before it), and pressure from the Trump administration. To some degree, this even gives Washington political leverage when negotiating with partners.
While Nvidia's Blackwell silicon is now built in the USA, to make an actual Nvidia B300 out of them, they must be transported back to Taiwan and then integrated using CoWoS-L advanced packaging technologies with HBM3E memory at a TSMC facility. This not only makes these chips more expensive than those produced in Taiwan, but to a large degree, this strikes strategic and political advantages, leaving only the symbolic part.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Still, this strategic dependence on advanced packaging in Taiwan is not going to last long. Both TSMC and Amkor are building advanced packaging facilities in the U.S., so when they come online, sometimes towards the end of the decade, that dependence will get significantly lower. Also, Micron and SK hynix are building DRAM production (Micron only) and HBM packaging facilities in the U.S., which will mark another significant step towards onshoring of production of strategically important components.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.