Nvidia introduced the DSX POD reference architecture, designed to accelerate the deployment of AI clusters in containers. The system integrates Rubin GPUs, Vera CPUs, and networking hardware, potentially reducing setup time from six months to 14 days. Dell Technologies and Hewlett Packard Enterprise are among the partners adopting this blueprint, with initial shipments expected early next year.
The company also unveiled a second reference architecture to support its OpenShell AI agent toolkit. This configuration utilizes Vera CPUs, Rubin GPUs, and BlueField-4 accelerators to run AI agents in secure, isolated sandboxes. It supports deployment on various hardware, including DGX Station workstations, DGX Spark desktops, and laptops equipped with the RTX Spark system-on-chip.
In the robotics sector, Nvidia updated its Isaac simulator with new APIs and three template workflows for the GR00T foundation model. Agile Robotics debuted the Agile ONE SN humanoid robot, powered by GR00T, which Nvidia intends to serve as a development platform. Nvidia also highlighted collaborations with Mercedes-Benz and Deutsche Telekom within the European market.
CEO Jensen Huang noted that Nvidia has invested nearly $10 billion in Europe over the past year. This investment includes 14 separate deals involving European startups. The company continues to focus on providing infrastructure for AI factories and autonomous systems, emphasizing security and control in large-scale AI agent deployments.
