NVIDIA DSX AI Factory Architecture and DSX Pod Overview

An overview of NVIDIA DSX, a full-stack reference architecture for AI factories, and the announced DSX Pod. It explains how compute, networking, storage, software, power, and cooling are co-designed and validated in a digital twin before construction, then describes a factory-integrated pod based on Vera Rubin NVL72. The material focuses on architecture, simulation partners, power and cooling approaches, and on-site deployment rather than pricing, performance benchmarks, or independent testing.

■ Design and validation
- Digital-twin workflow involving PTC, Dassault Systèmes, Schneider Electric, Jacobs, Siemens, and Procore
- Virtual validation of performance, power, water, thermal, acoustic, and grid impacts before construction

■ Power and cooling
- DSX MaxLPS control loop for workload-matched power delivery
- Closed-loop liquid cooling, dry coolers, and 800 V DC rack connection

■ DSX Pod deployment
- Vera Rubin NVL72, Spectrum-X cabling, factory integration, and testing
- Transportable containers and estimated site deployment reduced from months to days

Intended for AI infrastructure planners, data center architects, and technology decision-makers evaluating large-scale AI factory designs. It provides a concise view of the announced architecture and deployment model and can help frame follow-up questions on site requirements, cooling, power, and integration.