Bull and Foxconn to Build NVIDIA Vera Rubin NVL72 in Europe: A New Front for AI Sovereignty

This cross-border flow is designed to give the product legitimate European manufacturing credentials while leveraging Foxconn's global supply chain capabilities and Bull's high-performance computing integration expertise. It's a practical answer to a question that has dogged European AI infrastructure ambitions: how to build the most advanced GPU-dense racks on the continent without depending entirely on assembly lines in Asia or North America.

Bull intends to sell the systems to neo-cloud providers, cloud service providers, and a new generation of AI factories across Europe and other markets. Beyond the hardware, Bull will provide the AI software stack, integrate use cases, and contribute data science expertise — extending the "Made in Europe" proposition from the factory floor through to operational deployment.

The platform: NVIDIA Vera Rubin NVL72

The Vera Rubin NVL72 is NVIDIA's third-generation MGX rack-scale system and the company's most aggressive push yet into agentic AI workloads. The platform integrates 72 Rubin GPUs and 36 Vera CPUs connected via NVLink 6, alongside ConnectX-9 SuperNICs and BlueField-4 DPUs. NVIDIA claims the architecture can train large mixture-of-experts models with one-fourth the GPUs required by its Blackwell predecessor while achieving up to 10x higher inference performance.

NVIDIA confirmed in March 2026 that Vera Rubin had entered full production, with shipments to customers scheduled for the second half of 2026. Over 50 MGX partners are building systems around the architecture. The Bull-Foxconn production timeline aligns with this broader global ramp, and the systems will enter a market where multiple heavyweights — Dell Technologies, HPE, Lenovo, and Supermicro — are already committed to Vera Rubin server builds and are in full-scale production.

Strategic implications

A direct play for European AI sovereignty

The collaboration is explicitly framed around reducing Europe's dependency on non-European manufacturing for high-end AI infrastructure. By assembling and validating the most advanced NVIDIA racks inside the EU, the partners are offering a supply chain narrative that hyperscalers and government-backed AI initiatives can cite when procurement decisions intersect with digital sovereignty requirements. Executives from both companies described the move as a milestone for European industrial resilience, and NVIDIA's EMEA vice president called it the "next chapter of AI in Europe."

An end-to-end European value chain

The model deliberately connects hardware manufacturing with software and services. Foxconn contributes scaled component production; Bull contributes system design, integration, and go-to-market expertise. The result is a value chain that stretches from fabrication in Central Europe to customer deployment in Western Europe. The press release states that this initial focus on the Vera Rubin NVL72 creates a foundation for supporting additional technology architectures in the future, signaling a platform strategy rather than a one-off project.

Competitive dynamics

Bull, historically an Atos subsidiary, now has a direct path to supply next-generation NVIDIA racks to European hyperscalers and emerging AI factory operators. It will compete directly with Dell Technologies, HPE, Lenovo, and Supermicro — all of whom are already building and shipping Vera Rubin-based systems. Foxconn, already named in NVIDIA's official list of production partners, deepens its footprint in European AI infrastructure manufacturing at a moment when the continent is ramping up data center investment.

A scalable beachhead

The joint statement explicitly notes that the collaboration is designed to support additional technology architectures in the future. This suggests the Czech-French production route could extend to future NVIDIA generations and potentially to other chip ecosystems, making it a permanent European manufacturing corridor rather than a temporary project tied to a single platform.

What's next

With Vera Rubin systems already arriving at early adopter data centers and volume shipments expected in the second half of 2026, the Bull-Foxconn production line enters a market that is still forming. European AI infrastructure company Verda, for instance, has announced it will be among the first to deploy Vera Rubin NVL72 at scale, with expected capacity exceeding 100,000 GPUs throughout 2027. The Bull-Foxconn partnership is betting that a meaningful share of that demand — and the demand that follows from sovereign AI programs and European hyperscalers — will want a supply chain that lives on the same continent as the data centers it feeds.