Nvidia has introduced a modular approach to building artificial intelligence data centers that assembles components like Lego blocks, cutting the time from construction to operation to as little as 14 days.
Nvidia Chief Executive Jensen Huang unveiled the design standard, called DSX POD, at GTC Berlin 2026 in Germany on the 21st, according to The Wall Street Journal and other outlets.
Under the approach, seven server racks for Vera Rubin, the company's next-generation AI computing system, along with networking equipment, cabling and liquid cooling systems, are assembled at a factory and shipped to the site in containers. Instead of installing parts one by one on site, workers simply connect the containers to power, cooling and communications infrastructure already in place, shortening a process that previously took six months to as little as 14 days. Nvidia will supply AI data centers built to the design from early next year together with 17 global partners, including Dell.