NVIDIA has introduced DSX POD, a modular, factory-built reference design for deploying complete Vera Rubin AI factories in containerized configurations. Announced at GTC Berlin, the architecture shifts much of the physical integration, cabling, power distribution and liquid-cooling installation from customer sites to manufacturing facilities. NVIDIA says the approach can reduce deployment time at a prepared site from approximately six months to as little as 14 days. Ecosystem partner deliveries are planned for early 2027.

DSX POD integrates NVIDIA Vera Rubin computing systems with networking, rack-level power infrastructure, cabling and liquid-cooling systems. Manufacturing partners complete rack-to-rack connections, electrical wiring, cooling pipes and functional validation before shipment. Containers are transported to customer locations and connected to prepared power, cooling, networking and supporting infrastructure. Site preparation and equipment integration can therefore proceed in parallel.

NVIDIA identified ecosystem participants including Anduril, Armada, Bull, Crusoe, DayOne, Dell Technologies, Delta, Eaton, Foxconn (Ingrasys), GIGABYTE, HPE, Lenovo, Pegatron, QCT, Schneider Electric, Vertiv and Wistron. The announcement emphasizes European AI infrastructure but describes worldwide deliveries starting in early 2027.

Key Highlights

Executive Commentary

Jensen Huang, NVIDIA founder and CEO, emphasized that an AI factory begins creating value when it starts producing tokens rather than when construction finishes.

Analysis

DSX POD addresses deployment constraints beyond GPU availability. Conventional data center builds require interdependent rack installation, high-density electrical distribution, liquid-cooling integration, network cabling, commissioning and system validation. Factory integration enables repeatable configurations while operators prepare the site. NVIDIA’s 14-day figure applies to prepared sites and excludes potentially lengthy land, permitting, construction and utility-power work. The announcement does not establish independently validated installation timelines, detailed power density or total deployment costs.

For networking, DSX POD treats the AI cluster as a preintegrated infrastructure unit. Rack-to-rack cabling and network validation become factory processes rather than entirely onsite activities. External optical connectivity, inter-container bandwidth and scaling limits remain important unresolved technical questions for large AI training and inference deployments.

NVIDIA Company Profile

We’re tracking AI infrastructure developments. Follow ongoing coverage at Converge Digest AI Infrastructure.

Source: NVIDIA press release, October 21, 2026.