When NVIDIA announced its U.S. manufacturing push in April 2025, the story was mostly a roadmap.
By 2026, part of that roadmap had become real production.
NVIDIA now says Blackwell wafers are being manufactured in volume at TSMC’s Arizona facility. The company also lists U.S. partner manufacturing activity across chips, packaging, optics, systems and AI-server assembly.
What has actually moved onshore?
TSMC is producing NVIDIA Blackwell wafers in Arizona. NVIDIA says Foxconn is working on AI-server manufacturing capacity in Houston and Wistron is investing in facilities in Dallas.
The wider partner network also includes companies working on packaging, optical components, fiber, power systems and other parts of the AI-infrastructure supply chain.
That is more concrete than the original 2025 announcement. It is still not the same as saying an NVIDIA system is made entirely from U.S.-origin components.
“American-made” needs a supply-chain map
An AI server contains far more than a GPU. Fabrication, advanced packaging, memory, networking, optics, boards, power equipment, cooling and final system assembly can all happen in different places.
So the useful question is not whether NVIDIA can point to U.S. factories. It can. The useful question is how much of each stage has actually moved onshore and at what production scale.
The $500B number is still a target
NVIDIA continues to say it plans to produce up to $500 billion of AI infrastructure in the U.S. with partners. That is a forward-looking production target, not a completed investment total.
The Robius takeaway: the 2025 announcement was not empty. Some manufacturing is now operating. But “U.S. manufacturing” describes a growing network of production stages, not a single switch from foreign-made to fully domestic hardware.
Sources: NVIDIA U.S. manufacturing update; NVIDIA April 2025 announcement.
Robius.news — Dubai, UAE — 2026 | Built to be first. Built to be trusted.
