Nvidia wants AI data centers to ship fully preassembled in containers. At GTC Berlin, the company unveiled DSX POD, a reference design for modular data centers built around Nvidia's Vera Rubin systems. Partners such as Dell, HPE, Lenovo, and Schneider Electric install the computers, networking gear, cabling, power supply, and liquid cooling in containers at the factory and test them there.
On site, the units only need to be hooked up to the power, cooling, and network of a prepared location. Nvidia says this cuts setup time from six months to as little as 14 days. The company argues that the sooner a data center runs AI models, the sooner its operator starts making money.
Nvidia also says the broader DSX platform can use up to 100 percent less cooling water because its closed-loop cooling systems release heat into the air without evaporation. The company admits that some climates will still need extra cooling during extreme heat. The first units are set to ship worldwide in early 2027.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.