NVIDIA Unveils DSX POD Modular AI Factories to Accelerate Deployment in Europe

Image: NVIDIA · Source

At GTC Berlin 2026, NVIDIA CEO Jensen Huang introduced DSX POD, a modular AI factory reference design that drastically reduces deployment time, aiming to boost Europe's AI infrastructure and production capabilities.

At the GTC Berlin 2026 event held at the Tempodrom venue, NVIDIA CEO Jensen Huang delivered a keynote emphasizing the rapid advancement and industrial revolution driven by artificial intelligence (AI). Huang highlighted how prominent European companies such as L’Oréal, Revolut, Spotify, Merck, Vodafone, Zara, Porsche, Nestlé, and HSBC are partnering with NVIDIA in advancing AI technologies. The company has committed nearly $10 billion to the European technology ecosystem and is collaborating with over 12,000 partners.

Huang described AI as a multilayered system, beginning with energy powering advanced computers that form an "AI factory," which then generates digital intelligence and integrates it into various applications. He stressed the importance of establishing AI infrastructure early, akin to how the internet and electricity required foundational networks and grids prior to widespread adoption.

In his keynote, Huang introduced the DSX POD — a novel reference design for NVIDIA’s Vera Rubin AI factories. These modular containerized units are pre-integrated and factory validated, enabling rapid onsite deployment. Traditionally taking up to six months, installation time for these AI factories can be cut down to as little as 14 days once site utilities are prepared. This approach shifts complex integration tasks like computing, networking, cabling, and cooling from the field to factory settings, enhancing speed and reducing errors.

The DSX POD platform not only expedites deployment but also improves resource efficiency; NVIDIA claims it can reduce water usage for cooling by up to 100% and lower power consumption by up to 30% while maintaining AI output. This innovation represents a significant shift in AI data center logistics and aims to increase deployment velocity worldwide.

Furthermore, Huang discussed the evolution of AI agents, which combine models, orchestrators, sandboxing, knowledge bases, and tools to think, reason, and act autonomously. He introduced NVIDIA’s OpenShell Reference Architecture for Safe AI Agents, emphasizing security through kernel-level sandboxing and dynamic privilege management to safely govern agent operation within systems.

NVIDIA is collaborating with global technology and infrastructure partners including Dell Technologies, HPE, Lenovo, Schneider Electric, and others to manufacture and deploy DSX POD AI factories globally by early 2027. These efforts signal NVIDIA’s commitment to establishing Europe as a central hub in the growing AI economy by providing scalable, efficient, and secure AI infrastructure solutions.