NVIDIA DSX POD lets partners build a complete Vera Rubin AI factory inside modular containers, test it at the factory and ship it ready to connect to a prepared site. NVIDIA says the design moves the installation of computing, networking, cabling, rack-level power and liquid cooling from the customer's site to the manufacturing plant.
At GTC Berlin, NVIDIA said the factory-built units cut deployment at a prepared site from six months to as little as 14 days. Partners complete rack-to-rack cabling, wiring, piping and functional validation before shipment, separately from site preparation, which NVIDIA says shortens the schedule further. Site teams can prepare land, foundations, power and external connectivity in parallel.
After delivery, according to Premal Savla, a senior director of product management at NVIDIA, the work is placement, connections between containers and hookup to external power, cooling and networking, followed by transport checks, site validation and commissioning. Partners can assemble, flush, pressure-test and validate each container's internal cooling loop at the factory, then connect it on site to prepared water and heat-rejection infrastructure. NVIDIA says the design changes where integration happens but does not eliminate the supporting facility infrastructure.
DSX POD uses a standard 800 VDC input at the core containers, and the reference design covers the power path from there through protection, distribution, grounding, rack connections, controls and telemetry. Partners engineer the upstream generation, conversion and regional electrical setup for each site, since utility conditions and codes differ across regions.
NVIDIA does not make containers mandatory. NVIDIA DSX POD is one of several implementation models under its DSX AI Factory Platform, and conventional data center construction remains an option. The company aims the container route at large deployments where schedule predictability, repeatability and less onsite work are priorities, and says that once the supporting site infrastructure is ready, operators can add capacity through the same container interfaces and validation process.
NVIDIA supplies the validated design, OEMs and ODMs integrate the computing systems, modular-solution providers build the enclosures, and site partners prepare the location, complete installation and support commissioning. The company named 17 partners: Anduril, Armada, Bull, Crusoe, DayOne, Dell Technologies, Delta, Eaton, Foxconn (Ingrasys), GIGABYTE, HPE, Lenovo, Pegatron, Quanta Cloud Technology, Schneider Electric, Vertiv and Wistron. NVIDIA says it is working with them to deliver DSX POD AI factories worldwide in early 2027.
