Factory-assembled data centers
Nvidia has unveiled DSX POD, a standard design for AI data centers assembled in containers at a factory and deployable on-site in 14 days. The aim is to shorten AI data center deployment times and accelerate investment recovery and revenue generation. Technology industry observers say Nvidia is seeking to establish standards by defining not only chips and servers but also how data centers are built.
Nvidia CEO Jensen Huang unveiled DSX POD at GTC Berlin in Germany on October 21, local time, saying, "This product will change how AI data centers are built." DSX POD is based on Nvidia's latest AI accelerator, Vera Rubin, and will be available from early 2027.
DSX POD is supplied in containers. Inside are networking equipment, cables, liquid-cooling systems and other components manufactured and performance-tested at the factory. Nvidia explained, "It only needs to be connected at a site where power, cooling and other infrastructure are ready, reducing AI data center deployment time from the conventional six months to as little as 14 days." To produce these modular data centers, Nvidia has established partnerships with 17 server, power and cooling companies, including Dell Technologies, HPE, Lenovo, Foxconn and Vertiv.
Faster deployment and revenue generation
Nvidia's introduction of modular data centers that can be assembled quickly on-site reflects strong activity in the AI industry. Demand for AI infrastructure is currently surging. Bringing a data center under construction into operation even a day sooner can generate additional revenue. Huang said, "The value of an AI factory, or data center, arises not when construction is completed but when it starts producing tokens," linking rapid commissioning directly to profitability. According to Morgan Stanley, the Vera Rubin NVL72 server rack, which combines 72 Nvidia GPUs and 36 CPUs, is estimated to cost about $7.8 million, or 10.3 billion won, per unit. The article argues that companies are better off putting equipment costing 10 billion won into revenue-generating service within two weeks than buying it and leaving it unused for half a year.
Standards across AI data center infrastructure
Technology industry observers view the unveiling of the modular data center as part of Nvidia's strategy to establish standards for the AI industry. Nvidia accounts for more than 80% of the global AI accelerator market. Building on this position, it has been introducing a succession of designs tailored to its own products.
At CES in January 2026, Nvidia established liquid cooling using water at 45 degrees Celsius as the standard for its latest AI accelerator, Vera Rubin. Following this move, shares in other cooling equipment suppliers, including Vertiv and Schneider Electric, fell sharply. In September, Nvidia also launched DSX Ready, a certification program for power and cooling equipment. Under the program, Nvidia directly assesses and certifies whether energy storage systems (ESS) and coolant distribution units (CDUs) meet its design requirements. A technology industry representative said, "Nvidia is effectively establishing 'Nvidia standards' at every level of AI data centers, from chips to server racks, power and cooling equipment, buildings and operating software."
Original Korean article: Chosun Ilbo
English brief by Korea Wind Intelligence, koreawindintel.com
