October 21, 2026Alesha Thomas
5 min read
In Focus
- Jensen Huang urged Europe to produce AI using its industrial expertise and proprietary data
- NVIDIA expanded tools for enterprise agents, combining company knowledge with security controls
- Robotics, autonomous vehicles, healthcare and telecommunications demonstrated the reach of physical AI
NVIDIA CEO Jensen Huang used his GTC Berlin 2026 keynote on October 21 to outline how European businesses can turn industrial expertise into AI capabilities.
The address linked infrastructure, open models, and enterprise agents to applications across manufacturing, healthcare, telecommunications, and autonomous mobility.
“To be part of this AI economy, you don’t want to just be consuming the AI; you want to be producing the AI,” Huang said. NVIDIA’s keynote positioned access to computing, company knowledge, and specialized tools as the foundations for that transition.
Skip the endless scrolling. Get the day’s top tech stories straight to your inbox. Subscribe Now
How Is NVIDIA Preparing Enterprises for AI Agents?
Enterprise agents were central to the NVIDIA GTC Berlin 2026 keynote. Huang described systems that reason, use tools, and act on information, creating new security and operational-control requirements.
NVIDIA introduced its OpenShell Reference Architecture to help businesses build, isolate, monitor, and govern agents. The architecture integrates Vera Rubin infrastructure for inference, Vera CPUs for agent orchestration, and BlueField-4 processors running NVIDIA Sentry for policy enforcement.
The reference architecture is now available, with companies including SAP, ServiceNow, and Microsoft working with NVIDIA on integrations. For business leaders, the significance is the ability to define and enforce boundaries as agents gain access to enterprise systems.
NVIDIA also expanded its Agent Toolkit through the Agent Knowledge Foundation. The toolkit brings together retrieval models, structured-data models, and accelerated libraries to help agents interpret company information and support decision-making. Palantir, Databricks, and Snowflake are among the companies using these capabilities.
What Role Do Open Models Play in Enterprise AI?
NVIDIA presented open models as a foundation businesses can adapt with their own data and industry knowledge.
Its expanded portfolio includes Nemotron 3.5 models forlanguage and vision, Kumo models for structured business data, Cosmos world models, and Alpamayo 2 Edge for autonomous vehicles. These address a range of needs, from retrieving information and forecasting outcomes to understanding physical environments.
The business proposition is greater scope to customize AI and deploy it on infrastructure organizations control. Availability varies across the portfolio: Nemotron 3.5 Super VL is available, while Alpamayo 2 Edge is expected later this month.
Where Is Physical AI Moving into Industry?
Huang highlighted applications that integrate AI reasoning with simulation, sensors, and physical systems.
In manufacturing, NVIDIA’s Factory Operations Blueprint is now in early access, supporting agents for tasks such as inspection, root cause analysis, and coordination of factory operations. Siemens is integrating blueprint-built agents into Intelligence Center X, while Audi is using digital twins to evaluate production concepts and robotics workflows.
Robotics announcements included Agile Robots’ Agile ONE SN reference humanoid, built on NVIDIA Isaac GR00T and Jetson Thor and scheduled for early 2027. NVIDIA also introduced Isaac OS and made Isaac Lab 3.0 generally available to support robot development and training.
The keynote expanded on the physical AI theme, covering adaptive radiotherapy with Siemens Healthineers and autonomous mobility via NVIDIA Hyperion. Huang also highlighted telecommunications, citing a commercial phone call on T-Mobile’s network using NVIDIA technology with Nokia.
What Does the Keynote Mean for Industry Leaders?
Infrastructure remained part of the address, including the introduction of DSX POD modular AI factories. NVIDIA plans to deliver the systems through partners worldwide in early 2027.
Across the announcements, NVIDIA’s strategy is to connect computing capacity with the models, knowledge, controls, and tools businesses need to operate AI.
For industry leaders, the practical takeaway is to evaluate those requirements together. An agent’s usefulness depends on the company information and tools it can access, while its deployment requires clear permissions, integration, and oversight. The keynote framed Europe’s industrial expertise as an advantage; the execution challenge remains converting that expertise into measurable operational results.
AIData CenterManufacturingSoftwareDeepTechInformation Technology
