NVIDIA Announces European Sovereign AI-Factory Plans and Expands DSX POD Ecosystem
DataNews Editorial Team
NVIDIA has announced plans to develop sovereign AI factories in Europe with BULL in France and Foxconn’s Czech operation, while advancing its DSX POD modular deployment approach. The company is positioning pre-integrated compute infrastructure as a way to compress the on-site integration phase once a facility is ready.
- NVIDIA announced European sovereign AI-factory plans involving BULL in France and Foxconn’s Czech operation.
- DSX POD is positioned as a pre-integrated modular deployment model, with a stated 14-day connection-and-start-up target after site readiness.
- NVIDIA’s AI Factory architecture spans compute, networking, storage, software, power management and liquid cooling.
- Reported supplier plans target major capacity increases in IC substrates, packaging, testing and CPO optics.
- The cited manufacturing expansions remain planned targets rather than confirmed operational capacity.
NVIDIA has announced plans to develop sovereign AI factories in Europe with BULL in France and Foxconn’s Czech operation. According to the article, BULL manufactures racks in Angers, France, while Foxconn provides system manufacturing in the Czech Republic.
The initiative is built around NVIDIA’s full-stack AI Factory architecture, which brings together compute, networking, storage, software, power management and liquid cooling. Rather than treating accelerator deployment as a standalone server procurement exercise, the model emphasizes integration across the infrastructure stack required to deploy AI capacity.
A central element is DSX POD, which NVIDIA describes as a modular, pre-integrated approach for AI-factory deployment. The pods are presented as prepared compute containers with connections, piping and system validation completed in advance. The article says conventional installation—from rack deployment and cabling through piping, validation and operation—can take up to six months. DSX POD targets connection and start-up within 14 days after the site is ready; this is a stated target rather than a universally validated delivery schedule.
NVIDIA is expanding the DSX POD ecosystem with Anduril, Armada, BULL, Crusoe, Dell, Delta, Wistron, Gigabyte, Inventec, Compal, Pegatron, Quanta, Schneider Electric and Vertiv. The partner group spans system manufacturing, infrastructure integration, power and cooling, suggesting that the deployment model depends on coordinated delivery beyond the GPU and server layers.
The article also reports planned supply-chain expansion across IC substrates, advanced packaging, IC testing and co-packaged optics (CPO). Aggregate IC-substrate capacity is expected to increase by 2.5 times over three years. KYEC is planning to double relevant capacity by 2027, while Amkor targets a fourfold increase in interposer-wafer capacity by the end of 2028. SPIL targets a 15-fold increase in CPO optical-engine capacity by mid-2027 and a 30-fold increase by the end of 2028.
These are targets and plans, not confirmed commissioned capacity. If delivered, they could support a greater transition toward pre-integrated modular AI-factory architectures and more localized system production. They also reinforce the strategic importance of high-speed interconnects, packaging, testing, power systems and liquid cooling alongside accelerator supply.
The announcement points to a shift in AI data-center competition from supplying discrete GPUs to delivering standardized, full-stack AI-factory architectures spanning compute, networking, storage, power management and liquid cooling. If supplier expansion plans are executed, substrates, advanced packaging, IC testing and high-speed optical interconnects could become increasingly important constraints and strategic control points in AI-infrastructure delivery.
- TechNews 科技新報 / ChineseSource →
This article was prepared by the DataNews editorial team based on the sources listed above.