Nvidia announced Wednesday the DSX Pod AI Factory, a modular data center system designed to deploy AI infrastructure in 14 days—a significant reduction from the typical six-month timeline for traditional data centers. The system integrates seven Vera Rubin AI racks into a standard shipping container, weighing approximately 28,000 kilograms, and is engineered for plug-and-play deployment with pre-assembled cabling and liquid cooling.
The DSX Pod is intended for global customers, including those in Europe facing deployment challenges. Nvidia stated that the system arrives pre-validated and requires only connection to power, cooling, and networking infrastructure to become operational. Deliveries are scheduled to begin in early 2027, with manufacturing partners including Foxconn, Gigabyte, Lenovo, HPE, Quanta, and Pegatron. The company did not disclose pricing, though analysts estimate a single Vera Rubin rack costs over $7 million.
Nvidia emphasized the system’s efficiency, noting that its liquid cooling and dry-cooler designs could save up to 54 million liters of water per day compared to conventional data centers. The DSX Pod is positioned as a complementary solution to traditional data centers, enabling customers to begin AI operations while permanent facilities are under construction.
Industry Context and Evolution
Nvidia’s AI systems have evolved from single GPU chips to multi-GPU servers, then to integrated racks, and now to containerized Pods. The company highlighted its shift from chip manufacturing to end-to-end AI infrastructure, with CEO Jensen Huang noting that the cost per gigawatt of data center infrastructure has risen from $18 billion to over $40 billion in recent years. The Vera Rubin-powered Pods represent the latest step in this progression, aiming to address deployment bottlenecks and accelerate AI workload deployment for hyperscale customers.