Nvidia announced a new product called the DSX Pod AI Factory, a containerized system housing seven Vera Rubin AI racks designed to be transported on standard roads via truck. The company stated that the unit weighs approximately 28,000 kilograms and is roughly the size of a standard shipping container, with all cabling and liquid cooling components assembled at the factory prior to shipment.
Ian Buck, Nvidia's VP of hyperscale computing, told reporters that the deployment process at the customer site involves placing the pod in its location, connecting it to other containers, and completing final validation and testing. Nvidia claims this approach will reduce the time required for a new AI data center to begin serving tokens from six months to 14 days. The company noted that the pod is intended for customers worldwide and could help address deployment challenges faced by European clients.
The DSX Pod is a reference design that will be shipped from Nvidia's manufacturing partners, including Foxconn, Gigabyte, Lenovo, HPE, Quanta, and Pegatron. Nvidia expects deliveries to commence in early 2027. While the company did not announce a price for the system, analysts indicated that a single Vera Rubin rack can cost over $7 million.
This release marks a further evolution in Nvidia's product lineup, which has expanded from single GPU chips to servers with four or eight GPUs, then to entire racks, and now to containerized building blocks with multiple integrated racks. Jensen Huang, Nvidia's CEO, recently discussed the increasing cost of data center infrastructure on a podcast. He stated that the value offered per gigawatt of power usage has risen from approximately $18 billion to $20-something billion per gigawatt, and now to $40-plus billion per gigawatt, suggesting that this trend will continue.