In a major win for domestic advanced manufacturing, Taiwanese server giant Wistron officially opened its US$700 million D1 Smart Factory in Fort Worth, Texas.
Unveiled alongside Nvidia CEO Jensen Huang and Wistron Chairman Simon Lin, the 324,000-square-foot facility marks the first time Nvidia’s flagship GB300 Grace Blackwell Ultra Superchip systems are being mass-produced on American soil.
The launch is a concrete step in Nvidia’s pledge to manufacture up to $500 billion in AI platforms within the United States over four years, transitioning critical assembly away from single-source hubs in East Asia.
What Just Dropped?
Wistron’s new Texas hub isn't a typical assembly plant—it’s a high-tech facility built explicitly to handle two-ton AI supercomputer cabinets packed with Nvidia’s most demanding hardware.
The factory has already generated over 500 jobs across engineering, construction, and electrical trades, with workforce headcount expected to reach 1,000 by the end of 2026.
"The largest infrastructure buildout in history is underway," said Nvidia CEO Jensen Huang at the opening ceremony. "These AI factories must be produced everywhere. Together, Nvidia and Wistron are restoring US advanced manufacturing capacity in Texas."
Specs & Production Matrix
| Feature / Detail | Wistron D1 AI Smart Facility |
| Location | Fort Worth, Texas |
| Total Investment | $700 Million USD |
| Facility Footprint | 324,000 sq. ft. (~30,100 sq. meters) |
| Active Primary Output | Nvidia GB300 Grace Blackwell Ultra Superchip |
| Future Production Roadmap | Nvidia Vera Rubin Superchips |
| Target Output Capacity | Tens of thousands of computing boards / month |
| Workforce Scale | 500+ current employees (Expanding to 1,000 by late 2026) |
| Digital Simulation Tech | Built using Nvidia Omniverse, Cosmos, and Nemotron models |
Under the Hood: Building Factories in the Virtual World First
Before ground was broken in Fort Worth, Wistron engineers built and simulated the entire D1 facility inside a digital twin powered by Nvidia’s Omniverse and Cosmos AI models.
1. Digital Twin Workflow Optimization
By modeling physical robotics, assembly line layouts, and automated transport vehicles in a virtual space, Wistron optimized factory flow and trained technicians on standard operating procedures before a single real-world machine was installed.
2. High-Level L6 Server Assembly
Building an AI supercomputer isn't like assembling consumer electronics. A fully populated GB300 rack contains roughly 1.5 million individual parts, weighs nearly two tons, and costs up to $4 million per unit.
3. Energy and Operational Intelligence
The facility runs automated inspection tools built on Nvidia Metropolis frameworks and uses physical AI models to balance electrical grid loads during peak testing phases, keeping factory power draw predictable and efficient.
Real-World Utility & Verdict
For enterprise tech leads, cloud providers, and the broader AI market, Wistron’s Texas expansion offers clear operational advantages:
- Shortened Delivery Timelines: Assembling and testing server infrastructure domestically reduces transit times for North American data center operators from weeks to days.
- Supply Chain Resilience: Adding domestic production capacity reduces vulnerability to geopolitical friction and regional shipping bottlenecks across the Pacific.
- Paving the Way for Rubin: Establishing automated lines for Blackwell today creates the exact physical foundation needed to seamlessly transition to Nvidia’s next-generation Vera Rubin platform without requiring factory overhauls.
- Easing the Compute Shortage: While D1 currently accounts for roughly 5% of Nvidia's total global manufacturing output, every additional board leaving the Fort Worth floor helps address the persistent shortage of enterprise AI compute.
0 Comments