Nvidia unveils DSX POD containers to cut AI factory deployment from months to weeks

Nvidia describes DSX POD as a reference design for complete Vera Rubin AI factories that partners assemble inside modular containers and validate before delivery. Onsite deployment can shrink from six months to 14 days once site utilities are ready. Nvidia also says its DSX platform can cut cooling water use by up to 100% and deliver the same AI output with up to 30% less power. The launch came at GTC Berlin, where Jensen Huang opened by pointing to European companies already building with Nvidia and urged the region to accelerate its AI buildout. Coverage of the event put a one-gigawatt data center at about $50 billion once land, power, buildings, compute, storage and cooling are counted. For founders, capacity timing often matters as much as chip access. A prebuilt, validated module that compresses deployment could shape where neoclouds and model builders place their next clusters and how quickly they add inference capacity. Teams planning multi-site growth should track which partners ship these pods first.