Nvidia unveiled prefabricated data center containers called DSX pods at GTC Berlin, and The Register's Tobias Mann described the plan on October 21 as Nvidia's pitch for speeding up deployments, comparing the containers to trailer homes. Every figure below is an Nvidia claim.
What a pod contains
Starting in 2027, Nvidia and partners plan to offer Vera Rubin rack systems pre-plumbed and pre-wired, ready for deployment in these pods. Each pod is said to include compute, interconnects, electrical systems and plumbing, and the customer places it on a prepared site and connects power, networking and cooling. A pod weighs 28 metric tons and holds seven Vera Rubin NVL72 racks, with 504 GPUs and 252 Vera CPUs. The first Vera Rubin DSX pods are expected to leave the production line early next year.
The speed claim
Nvidia says deployment time falls from six months to as little as two weeks, because every fitting, cable and component is validated at the factory before shipment. Buyers can choose vendors, such as HPE over Dell or Eaton over Schneider Electric, and add capacity by adding pods. Named partners are HPE, Dell and Schneider Electric, plus others the article does not identify.
Waterless cooling
Nvidia also laid out new DSX specifications for waterless data centers that use closed-loop coolants and dry coolers instead of water-hungry cooling towers. The article ties this to earlier work on squeezing more performance from less power and letting facilities scale down workloads when grid conditions change.
The Register's skepticism
Mann frames the plan as a sales pitch, presents the two-week timeline as something the company says and not a verified result, and suggests the effort is aimed at addressing supply chain problems before they affect revenue. The article contains no direct quotes from Nvidia executives and does not mention pricing or customer deployments.
Source: The Register. This is an AI Business Watch summary of the original report.
