Nvidia wants AI factories built in real factories
Nvidia unveiled DSX POD at GTC Berlin, a containerized data center solution designed to accelerate deployment of AI compute capacity. Each container holds seven Vera Rubin NVL72 racks with 504 GPUs, pre-installed with approximately 20 kilometers of copper cabling and 17 kilometers of networking, and uses no water for cooling. The system weighs around 28 metric tons and contains roughly 10 million components, requiring only site preparation with power, cooling, and network connections before deployment.
Key points
- DSX POD containers ship fully-built and ready for deployment in 14 days versus months for traditional data centers
- Each unit contains 504 Rubin GPUs across seven racks with pre-installed cabling and networking
- Water-free cooling design and modular connectivity reduce on-site installation complexity
Summary and key points written by AI from the headline and a short excerpt, so they can be incomplete or wrong. The New Stack did the reporting; read the original for the full story.
Read the full story at The New Stack
