NVIDIA introduces DSX POD for modular AI factory deployment

DSX POD is designed around NVIDIA’s Vera Rubin platform and rack-scale systems.[2] NVIDIA says the reference design describes a complete AI factory, rather than treating a compute enclosure as a standalone unit.[2] Partners build the factory in modular containers and validate it before delivery.[1] NVIDIA says the initial target includes large-scale, greenfield, liquid-cooled AI infrastructure, including NVIDIA Cloud Partners and other organizations building at scale.[3] Standardized container interfaces are intended to provide a repeatable unit for planning and adding capacity.[2]

NVIDIA says DSX POD shifts integration and installation of computing, networking, cabling, rack-level power and liquid cooling to the manufacturing facility.[1] Partners complete rack-to-rack cabling, wiring and piping, then functionally validate the systems before shipment.[1] The containers arrive ready to connect to a prepared site’s power, cooling, networking and other supporting infrastructure.[1] Because factory work can proceed separately from site preparation, NVIDIA says the approach shortens the time needed to bring capacity online.[1] Its claimed deployment-time reduction is from six months to as little as 14 days.[1]

DSX POD does not eliminate supporting facility infrastructure, and NVIDIA says the reference design does not make containerization mandatory for every AI factory.[2] The design uses a standard 800 VDC power input, while partners engineer the upstream power-generation, conversion and regional electrical solution for each customer site.[2] NVIDIA says factory testing reduces exposure to component delays and costly, error-prone onsite work.[1] These details mean the shorter timeline depends on coordinating factory delivery with site readiness; the supplied design does not replace that site preparation.[1][2]

NVIDIA says DSX POD AI factories are to be delivered worldwide in early 2027.[1] The company named Anduril, Armada, Bull, Crusoe, DayOne, Dell Technologies, Delta, Eaton, Foxconn (Ingrasys), GIGABYTE, HPE, Lenovo, Pegatron, Quanta Cloud Technology, Schneider Electric, Vertiv and Wistron as delivery partners.[1] NVIDIA frames the business milestone as production: it defines time to first token as when a new AI factory begins running models for production workloads, and says an AI factory creates value when it starts producing tokens rather than when construction is complete.[1] It says customers can therefore begin generating tokens and earning revenue sooner.[1]