What changed
Nvidia introduced DSX POD, a reference design for modular AI data centers, at GTC Berlin. The system is built around Nvidia's Vera Rubin systems and includes preassembled computers, networking gear, cabling, power supply, and liquid cooling in containers. Partners such as Dell, HPE, Lenovo, and Schneider Electric install and test these components at the factory. On-site, the units only require connection to power, cooling, and network infrastructure.
What this means in practice
The DSX POD significantly reduces the time required to deploy an AI data center, cutting setup from six months to as little as 14 days. Nvidia claims this allows operators to start running AI models faster, thereby generating revenue sooner. The DSX platform also uses closed-loop cooling systems, which can reduce cooling water usage by up to 100 percent. However, Nvidia acknowledges that extreme heat in certain climates may still require additional cooling measures.
Limitations and availability
The first DSX POD units are set to ship worldwide in early 2027. The system is designed for use with Nvidia's Vera Rubin systems, and the full deployment depends on partners installing and testing components in factory settings. The announcement does not specify exact hardware requirements or licensing conditions for the DSX platform.
