NVIDIA Unveils DSX POD AI Factories at GTC Berlin
NVIDIA unveiled DSX POD modular AI factories at GTC Berlin, cutting deployment from six months to 14 days. Vera Rubin systems use less water and less power.
NVIDIA has unveiled DSX POD, a new reference design that turns complete AI factories into modular container units partners can build, ship, and plug in at customer sites. Announced at GTC Berlin on October 21, 2026, the design cuts AI factory deployment from about six months to as little as 14 days, and it runs on NVIDIA's new Vera Rubin systems. Cloud providers including CoreWeave, Google Cloud, Lambda, Nebius, and Nscale plan to use the technology.
What DSX POD actually is
A DSX POD is a complete Vera Rubin AI factory packed into modular containers by NVIDIA's partners. Computing, networking, cabling, rack-level power, and liquid cooling are integrated and validated at the manufacturing facility, not at the customer site. When the containers arrive, they connect out of the box to a prepared site's power, cooling, and networking infrastructure. "An AI factory creates value when it starts producing tokens - not when construction is complete," said Jensen Huang, NVIDIA's founder and CEO. "With NVIDIA DSX POD, we're bringing factory-built precision and speed to AI infrastructure construction - reducing deployment time from months to weeks so customers can scale capacity and bring revenue-generating production online faster."
Why it matters for the AI buildout
Building AI data centers has been slow, expensive work: months of planning, specialized labor, and error-prone onsite installation. Factory-built containers remove much of that friction, and NVIDIA says they also reduce exposure to component delays. The units use less cooling water and up to 30 percent less power than a traditional build, which matters in Europe, where power constraints and sustainability rules are tightening. The announcement pairs with the broader NVIDIA DSX AI Factory Platform: Europe's industrial and energy companies, including AVEVA, Dassault Systemes, Siemens, ABB, Eaton, GE Vernova, Hitachi Energy, Schneider Electric, Siemens Energy, Trane Technologies, and Vertiv, are adopting DSX to design and run AI factories that minimize water use and adapt power consumption to support grid reliability. NVIDIA also open-sourced DSX OS software for managing infrastructure at scale.
A safety layer for the agent era
NVIDIA also launched the OpenShell Reference Architecture at GTC Berlin, a full-stack design for running autonomous AI agents safely on enterprise infrastructure. It combines NVIDIA Vera CPUs and BlueField DPUs with open-source OpenShell software that monitors agent behavior and enforces policies in silicon. Companies including Anthropic, OpenAI, Google Cloud, Meta, Microsoft, Dell Technologies, and SAP are working with the architecture as enterprises prepare to deploy billions of AI agents.
Sources
GlobeNewswire (NVIDIA announcements) - https://www.globenewswire.com/news-release/2026/10/21/3385267/0/en/nvidia-introduces-dsx-pod-ai-factories-to-accelerate-europe-s-infrastructure-deployment.html Seeking Alpha - https://seekingalpha.com/news/4655023-nvidia-targets-faster-ai-infrastructure-rollouts-with-new-dsx-pod-design