N環球 
Nvidia unveils OpenShell and DSX POD at GTC Berlin: AI factories in 14 days
CEO Jensen Huang's October 21 GTC Berlin keynote introduced the OpenShell Reference Architecture to contain and monitor AI agents, plus DSX POD factory-built AI factories cutting deployment at a prepared site from six months to as little as 14 days, with global deliveries in early 2027.
Why it matters
At his October 21 keynote at GTC Berlin (October 20-22), Nvidia CEO Jensen Huang rolled out two headline announcements: the OpenShell Reference Architecture, designed to keep the billions of AI agents enterprises are about to deploy contained and under control, and DSX POD, factory-built AI factories that move data center construction from the building site into the factory. "An AI factory creates value when it starts producing tokens — not when construction is complete," Huang said.
Full report
OpenShell is a hardware-and-software recipe: models run on Vera Rubin systems, agent harnesses and tool use run on Vera CPUs, while BlueField-4 chips run Nvidia's Sentry software to watch agent behavior from outside the main system and enforce policy in hardware. "Agents can think on Rubin, act on Vera and be governed on BlueField," Huang said. It extends to desk-side machines such as DGX Spark and DGX Station. Nvidia lists Anthropic, AWS, Google Cloud, Meta, Microsoft, OpenAI, Perplexity, SAP and SpaceX Super Intelligence among those working with the architecture, alongside cybersecurity firms including CrowdStrike and Palo Alto Networks.
There are caveats. The Register noted OpenShell requires Nvidia hardware, the open-source secure runtime still has open security issues on GitHub, and it collects telemetry by default unless users opt out. Banks BBVA and Santander are reviewing the spec.
DSX POD packs a complete Vera Rubin AI factory into modular containers: partners complete rack-to-rack cabling, piping and functional validation inside the factory before shipping, so that at a prepared site — power, cooling and networking ready — deployment falls from about six months to as little as 14 days. Global computing and infrastructure partners including Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv are working toward worldwide delivery in early 2027. Nvidia's economic rationale: building a 1-gigawatt AI data center costs about $50 billion, and at an assumed $40 billion in annual revenue, a three-month delay wipes out $10 billion in revenue opportunity — construction time is money.
💬 Comments
No comments yet. Be the first!