NVIDIA GB300 Manufacturing: What UK Buyers Need in 2026

What opened in Fort Worth

Covering around 324,000 square feet, Wistron's new D1 AI smart facility marks the company's debut manufacturing base on American soil and stands as the first U.S. site turning out the NVIDIA GB300 Grace Blackwell Ultra Superchip. Staffing has already passed 500, a number Wistron expects to double to 1,000 before year's end, and the opening ceremony brought together Wistron Chairman Simon Lin and NVIDIA CEO Jensen Huang alongside local and Taiwanese government officials.

GB300 output is the plant's starting point, with the NVIDIA Vera Rubin Superchip slated to join the line once that platform is ready. Wistron describes the facility as advanced AI infrastructure production rather than standard server assembly, drawing on NVIDIA's Nemotron and Cosmos models together with the Omniverse and Metropolis toolsets to build digital-twin models of the factory, test workflows virtually, and refine operations before ramping output.

Why a US-based plant matters for UK GPU buyers

On the surface, a Texas assembly plant sounds like a domestic U.S. story. But GB300 systems built in Fort Worth still feed a global allocation pool that UK buyers compete for. More assembly capacity onshore in the U.S. — alongside NVIDIA's broader plan to commission over a million square feet of manufacturing space in Arizona and Texas — signals rising overall throughput for GB300, which is the factor that ultimately determines whether UK orders arrive in weeks or quarters.

That said, UK buyers should not assume proximity translates into priority. U.S. hyperscalers and large domestic AI builders are typically first in line for newly opened capacity, which is exactly why tracking allocation patterns via a server lead time and supply chain tracker matters more now than before this plant existed.

The wider U.S. manufacturing build-out behind GB300

Fort Worth is just one piece of a much bigger push. NVIDIA has confirmed that Blackwell chips are rolling off production lines at TSMC's Phoenix fabs, and that it is standing up supercomputer manufacturing in Texas via Foxconn's Houston site and Wistron's Fort Worth plant. Together with partners TSMC, Foxconn, Wistron, Amkor and SPIL, NVIDIA is targeting as much as $500 billion in U.S.-built AI infrastructure over a four-year span.

Houston, run by Foxconn, turns out the GB300 tray modules, whereas Fort Worth, under Wistron, handles final assembly and testing — the two sites function as linked steps in one supply chain rather than rival production points. For buyers modelling delivery timelines, this matters because bottlenecks can appear at wafer fabrication, tray-module production, or final assembly, not just at the headline factory opening.

How much of this reaches the UK

Recent hyperscaler contracts already show how much of this capacity is bound for the UK. Nscale's build-out spans 300,000 NVIDIA Grace Blackwell GPUs across the U.S., Portugal and Norway, of which 60,000 are earmarked for UK deployment. A further Microsoft-Nscale expansion adds 139,600 GB300 GPUs across the U.S. and Europe — reported elsewhere as an additional 100,000 Nvidia GPUs, a figure that does not match the sum of individual site allocations below — with no fewer than 23,000 Grace Blackwell Ultra GPUs destined for a UK site at Loughton, Essex, a 240MW Texas facility slated to hold 104,000 GB300s, and as many as 12,600 units earmarked for Start Campus in Sines, Portugal.

At $14 billion, the Microsoft-Nscale deal shows just how large a queue UK buyers are up against. Enterprise and mid-market UK buyers sourcing GB300-class capacity outside hyperscale contracts should expect allocation to sit behind these headline commitments, which is one reason to model realistic timelines using the UK server lead times outlook rather than vendor quotes alone.

• •300,000 Grace Blackwell GPUs across US, Portugal and Norway (Nscale)
• •60,000 GPUs being established in the UK
• •23,000+ GPUs earmarked for Loughton, Essex
• •104,000 GB300s planned for a 240MW Texas site
• •Up to 12,600 GB300s planned for Sines, Portugal

Lead times, risk and where GB300 sits versus Rubin

By March 2026, Dell was already shipping desktop-class machines built around GB300 with NVIDIA OpenShell support, and NVIDIA reports that the first DGX Station units running GB300 superchips have landed with developers — a sign the platform has moved well past the announcement stage into genuine deployment. Enterprise-grade availability was signalled even sooner: Dell's own roadmap had its GB300 NVL72-based PowerEdge XE9712 pencilled in for the second half of 2025, well before Fort Worth came online.

Buyers weighing whether to commit to GB300 now or wait should also watch NVIDIA's next platform. The Register reported in April 2026 that supply-chain pressure could delay NVIDIA's Rubin GPUs, with TrendForce warning of later shipping and smaller-than-expected volumes — a useful comparison for anyone deciding between ordering GB300 now or timing a purchase around Blackwell now or Rubin later. NVIDIA's own forced-labour due-diligence disclosure also flagged supplier non-compliance issues around hiring fees, document retention and working hours — a governance factor procurement teams should weigh when auditing any new manufacturing node feeding UK deployments.

What UK buyers should actually do next

Fort Worth expands total GB300 assembly capacity, which is a positive signal for supply over time, but it does not automatically shorten UK lead times in the near term given how much volume is already committed to hyperscale programmes like Nscale's UK and European rollouts. Buyers should size real workload requirements with an AI GPU calculator before committing budget to specific GPU counts, and compare configuration options through server configuration tools rather than assuming GB300 is the only viable path.

For teams evaluating whether GB300 fits current workloads versus established NVIDIA DGX systems or broader GPU accelerators options, the practical move is to lock in supplier conversations early, ask specifically where in the Phoenix-Houston-Fort Worth chain their allocation sits, and build contingency into procurement timelines rather than assuming new U.S. capacity resolves UK-side scarcity on its own.