Wistron opened its first US manufacturing facility in Fort Worth on July 21, 2026, a 324,000-square-foot greenfield plant that will produce the Nvidia GB300 Grace Blackwell Ultra Superchip and, later, the Vera Rubin Superchip. The Taiwanese contract manufacturer put $700 million behind the site and has already hired more than 500 workers, with a plan to reach 1,000 by the end of the year. The plant, known internally as the D1 facility, is scaling to produce tens of thousands of boards per month.
Nvidia founder and CEO Jensen Huang appeared alongside Wistron Chairman Simon Lin at the opening, and Nvidia used the event to anchor a much larger figure: a commitment to manufacture up to $500 billion in advanced AI platforms in the United States. Fort Worth is one of the first physical footprints of that pledge. The site currently runs two manufacturing cells, one for GB300 and one preparing to build Vera Rubin superchips as that platform ramps.
Huang framed the opening as a reindustrialization moment rather than a product launch. Wistron builds the systems that ship inside Nvidia's rack-scale AI supercomputers, and moving that step onshore matters both for supply-chain resilience and for the economics of the buildout Nvidia's customers are financing.
Key facts
- 01Wistron opened a 324,000-square-foot Fort Worth plant on July 21, 2026, representing a $700 million investment in US manufacturing.
- 02The facility already employs more than 500 workers, with plans to reach 1,000 by year-end.
- 03The plant is scaling to produce tens of thousands of Grace Blackwell and Vera Rubin superchip boards per month.
- 04Each Nvidia GB300 Grace Blackwell Ultra Superchip contains 1.5 million parts, weighs two tons, and costs $4 million.
- 05The Fort Worth facility is part of Nvidia's broader $500 billion commitment to manufacture AI platforms in the United States.
The scale of a single unit explains why volume production is hard. Each GB300 Grace Blackwell Ultra Superchip contains roughly 1.5 million parts, weighs about two tons, and carries a bill of materials Huang put at $4 million. These are not lightweight accelerators shipped in blister packs — they are room-scale systems, and Wistron is now assembling them at a cadence Nvidia describes as phone-like.
The economics behind that cadence are what make the Fort Worth plant matter. Nvidia's Vera Rubin NVL72 platform, which enters production this year, is projected to deliver roughly 10x the tokens per megawatt of prior generations — and every one of those systems has to be assembled somewhere. Building them in Texas rather than shipping finished racks from Taiwan compresses lead times for hyperscaler customers already competing for delivery slots.
Huang described the GB300-based system as the most powerful AI supercomputer in the world and emphasized the operational reality of shipping them in volume. Nvidia is treating the assembly step, not just the silicon, as a bottleneck to solve.
“It has one-and-a-half million parts inside — two tons, 4 million dollars — and we're producing them like phones, right here in Wistron, cranking them out in volume because the world needs all of these machines to drive the intelligence infrastructure.”— Jensen Huang, Nvidia founder and CEO
Wistron built the plant itself using Nvidia's own simulation stack before construction began. Engineers modeled the full facility in a digital twin using Nvidia's Omniverse and Metropolis libraries along with the Nemotron and Cosmos frontier models and the open PhysicsNeMo framework. Line layouts, production flows and worker training were validated virtually before any physical system came online — an example of Nvidia's software being used to accelerate the buildout of the factories that produce Nvidia's hardware.
Lin framed the opening in broader terms, arguing that AI infrastructure now shapes both national economies and daily life. His pitch was less about supply chain economics than about what the buildout produces downstream.
The Fort Worth site sits inside a wider Texas manufacturing cluster tied to the AI buildout, including Coherent's groundbreaking on a new facility in Sherman in June. The jobs footprint spans plumbing, electrical, construction and skilled manufacturing labor, which is politically useful for Nvidia as regulatory scrutiny of AI infrastructure siting intensifies. The bet is that concentrating AI-adjacent manufacturing in a single US region compounds — suppliers, workforce and utility capacity all reinforce each other.
There are real constraints on how quickly this scales. Tens of thousands of boards per month is meaningful, but Nvidia's customer demand — spanning hyperscalers, sovereign AI projects and enterprises — currently exceeds any single facility's output. Wistron will need to hit its 1,000-worker target on schedule, and the Vera Rubin production cell has to come online cleanly for the site to hold its role in Nvidia's 2026 delivery cadence. Any delay pushes bottlenecks back onto Nvidia's Taiwan-based capacity.
The Fort Worth opening is a marker in Nvidia's shift from chip vendor to full-stack infrastructure orchestrator. The company is no longer just selling GPUs — it is directing where they are assembled, financing the digital twins that design the factories, and setting the pace at which AI compute enters the market. For competitors trying to break into rack-scale AI systems, the message is that catching Nvidia now means catching an integrated manufacturing base, not just a chip roadmap.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
