FTFuture Technology
HARDWARE

The Fort Worth Manufacturing Plant Tells Us a Lot About Where AI Hardware Is Made Now

· 3 min read · By Nath Connell

Key takeaways

  • Wistron, a Taiwanese electronics manufacturer, has opened its first US manufacturing plant in Fort Worth, Texas
  • The facility produces rack-scale NVIDIA AI server systems for hyperscale and cloud customers
  • NVIDIA GPUs are primarily fabbed by TSMC in Taiwan, with high-bandwidth memory from Samsung and SK Hynix in South Korea
  • The CHIPS Act and domestic content requirements for government contracts are key drivers of the reshoring trend
  • A disruption to TSMC Taiwan would halt global AI hardware production regardless of US assembly capacity

When Wistron opened its first United States manufacturing plant in Fort Worth, Texas, to produce NVIDIA AI systems, the announcement was framed mainly as good news for domestic manufacturing. And it is. But there is a more interesting story underneath about why AI hardware production is moving to the United States, how much of it is genuinely domestic, and what this shift means for the global supply chain that the AI boom depends on.

Wistron is a Taiwanese electronics manufacturer, one of the major contract manufacturers that builds finished AI server systems. The Fort Worth facility marks its first US-based manufacturing operation, producing the kinds of rack-scale AI systems that hyperscalers and cloud providers buy in large quantities to power AI infrastructure.

Why Texas, Why Now

Texas has emerged as a genuine hub for AI infrastructure manufacturing, not just data centres. NVIDIA announced earlier this year that it is partnering with Foxconn to build AI servers in Houston. Dell is headquartered in Round Rock. The combination of business-friendly regulation, relatively cheap industrial land, available power infrastructure, and proximity to NVIDIA's Austin-area operations makes the state an attractive location for this kind of facility.

The timing is driven by a combination of policy pressure and commercial logic. US government export controls on advanced AI chips have reshaped global supply chains significantly. Building systems in the United States provides a cleaner route to compliant delivery for US government contracts, defence-adjacent customers, and organisations with domestic content requirements. The CHIPS Act and related industrial policy have also created financial incentives that make domestic manufacturing more competitive than it would otherwise be on pure economics.

It is worth being clear about what "made in America" actually means in this context. The chips inside these systems, NVIDIA's GPUs, are fabbed by TSMC primarily in Taiwan, with some production now ramping in TSMC's Arizona facility. The high-bandwidth memory comes from Samsung and SK Hynix in South Korea. The advanced packaging is done in Taiwan. What happens in Fort Worth is final system assembly and testing, which is a real and valuable manufacturing step but sits at the end of a very global supply chain.

What This Means for Supply Chain Resilience

The strategic logic behind onshoring final assembly is partly about resilience and partly about optics. Having the last manufacturing step on US soil creates a defensible position in procurement discussions and makes the systems more straightforwardly compliant with domestic content rules. It also reduces the shipping logistics involved in getting finished rack systems, which are large and fragile, from Asian manufacturing facilities to US data centres.

The future, in 3 minutes a day. The biggest tech story explained every morning, free. Get the briefing →

But the deeper supply chain dependency, on advanced semiconductor fabrication, remains offshore. A disruption to TSMC's Taiwanese operations would halt AI hardware production worldwide regardless of how many final assembly plants exist in Texas. That is a known risk that the US semiconductor industry is working to address through TSMC Arizona and Intel Foundry, but the timeline to meaningful domestic foundry capacity for leading-edge chips is measured in years, not months.

For NVIDIA, having Wistron operating domestically is useful for multiple reasons. It gives the company a US-based manufacturing partner that can handle government and defence-sector orders more cleanly. It also distributes production geography, which is prudent given the concentration risk of having almost all AI server assembly in Taiwan.

The Broader Reshoring Trend

Wistron's Fort Worth plant is one of many similar announcements over the past eighteen months. The trend is real and it is accelerating. AI hardware, solar panels, electric vehicle batteries, pharmaceutical active ingredients: the list of industries where the United States is actively trying to rebuild domestic manufacturing capacity is long. The economic and geopolitical logic behind it is sound.

Whether it succeeds is a different question. Manufacturing capacity takes years to build and skilled workforces take longer. The incentive structures created by the CHIPS Act and Inflation Reduction Act are helping, but they are competing against decades of optimised global supply chains that will not simply redirect overnight.

For now, Fort Worth has a new factory building some of the most powerful AI systems in the world. That is worth acknowledging, context and all.

Get the briefing, free

The biggest tech story, explained in 3 minutes every weekday. Choose your briefings →

Free. No spam. Unsubscribe in one click.