Nvidia unveils AI data-center blueprints, OpenShell architecture and robotics tools at GTC Berlin

The first data-center offering, DSX POD, is a reference architecture that hardware partners can use to build AI clusters inside containers. Each container houses Rubin graphics processing units, Vera central processing units, networking equipment, cooling gear and power distribution modules. Nvidia said the components are integrated and tested at the factory, removing the need for customers to assemble systems themselves. The company estimates DSX POD can cut hardware deployment times from six months to 14 days. Dell Technologies Inc. and Hewlett Packard Enterprise Co. are among more than a dozen hardware makers planning to adopt DSX POD, and Nvidia expects the first systems to ship early next year. “An AI factory creates value when it starts producing tokens — not when construction is complete,” Huang said.

Nvidia introduced DSX POD alongside a second data-center reference architecture designed to help customers deploy OpenShell, an AI-agent toolkit the company released last month. OpenShell runs AI agents in sandboxes that isolate them from sensitive infrastructure, while a supervisor module monitors traffic leaving the sandboxes for cybersecurity risks. The architecture combines Vera CPUs, Rubin GPUs and BlueField-4 accelerators, with the latter chips offloading certain auxiliary computing tasks. It also allows users to run OpenShell environments at the edge and supports Nvidia’s DGX Station workstations, DGX Spark miniature desktop and laptops using its RTX Spark system-on-chip. “As enterprises prepare to deploy billions of AI agents, safety and control must be built into the computing architecture,” Huang said. “Nvidia created OpenShell to give agents the freedom to do useful work within clearly defined boundaries — contained, monitored and under control.”

On robotics, Nvidia offers Isaac, a software toolkit with prepackaged foundation models, a robot simulator and related tools. The company launched a new version of Isaac’s simulator with an enhanced set of application programming interfaces that developers can use to define details such as the sensors a virtual robot should use to navigate a simulated factory. The update is rolling out with three template workflows for customizing GR00T, a robot-optimized foundation model included in Isaac. GR00T powers Agile ONE SN, a 5.6-foot-tall humanoid robot that also made its debut. The system was developed by Agile Robotics Inc., a startup that counts Nvidia as an investor. Nvidia said customers can use Agile ONE SN as the starting point for humanoid robot projects and that it takes less time to set up than the hardware engineers typically use for the task.

Huang’s keynote also emphasized Nvidia’s role in the European Union’s AI ecosystem. Mercedes-Benz will use the chipmaker’s Hyperion computing and sensing module to build autonomous vehicles, while Deutsche Telekom AG plans to expand its Industrial AI Cloud, an Nvidia-powered service that hosts open-source frontier models for customers. Forrester Vice President and Principal Analyst Paul Miller said Huang positioned Nvidia as “a strong supporter of — and enabler for — Europe’s sovereign tech ambitions,” according to SiliconANGLE. Miller added that Nvidia has invested almost $10 billion into Europe in the past year, including 14 investments in European startups during Huang’s current trip.

Editor's Summary

Nvidia used GTC Berlin to introduce DSX POD and OpenShell reference architectures intended to shorten AI data-center deployment and secure AI-agent workloads, along with updated Isaac robotics tools and a new humanoid robot. Dell and HPE plan to adopt DSX POD, with first shipments expected early next year. The announcements also tied Nvidia more closely to European AI, telecommunications and automotive projects.