Wistron and NVIDIA open $700 million Fort Worth plant to build AI superchip systems

Wistron has opened its first U.S. manufacturing facility in Fort Worth, a 324,000-square-foot greenfield plant built to produce NVIDIA AI systems and related “superchips” used in large-scale artificial intelligence infrastructure. NVIDIA founder and CEO Jensen Huang and Wistron Chairman Simon Lin took part in the opening ceremony, positioning the project as part of a broader effort to expand advanced manufacturing capacity in the United States for the AI era.

According to NVIDIA’s announcement, the plant, known as Wistron’s D1 facility, is currently running two manufacturing cells. One cell produces the NVIDIA GB300 Grace Blackwell Ultra Superchip, which is designed for AI data center performance, while the second cell is set to produce the NVIDIA Vera Rubin Superchip.

NVIDIA said D1 is scaling up during 2026 to produce “tens of thousands of boards per month” of the most advanced technology. At the ceremony, the company also unveiled the first NVIDIA GB300 Grace Blackwell Ultra Superchip produced on site, which Huang signed, tying the milestone to a specific production start rather than a future build-out.

The project is also being framed as a jobs and workforce initiative. NVIDIA described the plant as a $700 million commitment to advanced manufacturing in the U.S., creating more than 500 new jobs across the Texas facility, with plans to expand employment to 1,000 by the end of the year. Huang said the work will span roles such as plumbing, construction, manufacturing, and electrical jobs, reflecting the scale of industrial operations required to assemble and test high-end hardware.

A key operational theme in the announcement is how the factory was planned and brought online. NVIDIA said Wistron designed and fully simulated the facility before construction began using a “digital twin” built on NVIDIA’s open platform, including Nemotron and Cosmos frontier models, Omniverse and Metropolis libraries, and open frameworks including PhysicsNeMo. In plain terms, the digital twin approach uses software replicas to model factory layouts and production steps so teams can validate processes and train staff before physical equipment is fully running.

NVIDIA added that engineers used the simulation environment to validate assembly line layouts, optimize production processes, and train workers on standard operating procedures virtually before any system came online in the real factory. That workflow, the company said, is intended to help the plant ramp faster and adapt as products evolve at the chip and systems level.

NVIDIA’s Fort Worth push is part of a wider manufacturing strategy. The company said it has committed to manufacturing up to $500 billion in advanced AI platforms in the United States, and the Wistron project is one of the efforts making that number concrete. Huang described manufacturing capacity as a core “infrastructure” layer for AI, comparing it to established societal infrastructure such as roads, rail, and power generation.

Huang also characterized the systems being built around the GB300 superchip as among the most powerful AI supercomputers available. He said the integrated supercomputer built around the GB300 Grace Blackwell Ultra Superchip contains about 1.5 million parts, and he used a manufacturing analogy to underscore volume and precision, saying the company is producing them in volume “like phones,” right in Wistron’s facility.

While NVIDIA’s announcement outlines the plant’s capabilities and early production targets, it does not provide additional details that investors and industry observers often seek, such as the length of the ramp period to full capacity, specific customer contracts, yields, or how much of the broader AI systems bill of materials will be produced in Fort Worth versus assembled elsewhere. The company also did not disclose how quickly the plant will shift from producing the GB300 superchip cell to higher volumes for the Vera Rubin Superchip cell beyond the statement that D1 is set to produce it.

Looking ahead, attention will likely focus on whether D1 reaches its stated output ramp, how the factory’s digital twin workflow translates into production stability, and whether similar manufacturing expansions follow in other U.S. locations as NVIDIA’s AI supply-chain buildout progresses. With the plant already producing a first on-site GB300 unit and scaling board output, the next milestones to watch are updated volume figures and confirmation of ramp plans for the Vera Rubin Superchip production cell.