In close cooperation with Europe
A memorandum of understanding with the German federal government, a reference humanoid from Munich, and suppliers ranging from Schaeffler to SICK all on one platform: NVIDIA demonstrated in Berlin its vision for Physical AI from Europe. It remains to be seen who will ultimately own the result.
23 Oct 2026Share
On Monday, October 19, the Ministry of Economic Affairs and the Ministry of Research signed a memorandum of understanding with NVIDIA at the “Invest in Germany” investment summit; Federal Chancellor Friedrich Merz welcomed Jensen Huang at the Kraftwerk in Berlin. On Wednesday, the NVIDIA CEO took the stage at the GTC event at STATION Berlin. At the heart of the agreement is an initiative to develop open, specialized models for Physical AI—that is, for robots, autonomous vehicles, and automated systems—with two main focuses: training robots and vehicles, and models for simulating entire production processes. NVIDIA is to make its models available to small and medium-sized enterprises and research institutions and send its own engineers to Germany; according to Handelsblatt, the federal government is prepared to contribute a sum in the hundreds of millions. The agreement builds on the Industrial AI Cloud announced in June 2025.
A Reference Humanoid from Munich
The keynote was followed by the hardware presentation. Agile Robots, a spin-off from the German Aerospace Center (DLR) in 2018 and a unicorn since 2021, introduced Agile ONE SN, a reference platform for humanoid research: 1.40 meters tall, 40 kilograms, 31 degrees of freedom, three kilograms payload per arm, two five-fingered hands with 20 degrees of freedom each, Jetson Thor as the computer, Isaac GR00T as the software, with delivery scheduled for early 2027. According to the company, Stanford and ETH Zurich have expressed interest.
Robotics on the Verge of Its ChatGPT Moment
NVIDIA itself expanded its Isaac platform to include the Isaac OS runtime environment, the Isaac Lab 3.0 learning framework, and a library with more than 2,000 simulation-ready factory and robot objects. According to Huang, robotics is on the verge of its ChatGPT moment. The fact that NVIDIA simultaneously announced investments in Agile Robots and three other German startups—Avelios Medical, Mirelo, and Reflex Aerospace—fits right in: Huang had already promised approximately two billion dollars in new investments in Europe on October 13 in Paris, in addition to eight billion from the previous year.
The links in the supply chain are interlocking
The real news lies in the list of partners. ABB Robotics connects Omniverse with RobotStudio HyperReality for virtual commissioning, while KUKA uses the libraries in Visual Components FactoryLens. Schaeffler contributes its actuator technology as an exclusive core component to the Newton physics engine. Sereact trains its Cortex platform with GR00T on arms from Universal Robots, while Franka translates voice commands into movements of the FR3 Duo. Toyota Material Handling Europe and SoftServe report that commissioning times for autonomous industrial trucks have been cut by up to half; SICK is testing the Halos safety system, and TÜV SÜD and Accenture are working on certification. NEURA Robotics from Metzingen, which had unveiled its multimodal base model Neuron AI on October 14—combining 3D vision, touch, audio, robot status, and speech along with memory in a single model—had its 4NE1 humanoid sort laundry and load a washing machine and dryer for eleven minutes at its booth without any human intervention.
A Major Accolade for German Mechatronics
One could view this week as a major accolade for German mechatronics; Huang himself cited Europe’s manufacturing and mechatronics base as the foundation for a leadership role in humanoid robotics. It can also be seen as what it is from Santa Clara’s perspective: the development of an ecosystem in which actuators, safety technology, and factory data come from Europe, while models, simulations, and computing power come from NVIDIA. The German government speaks of technological sovereignty while simultaneously relying on the provider that controls the stack.
So that industrial data remains where it is generated
This is not a contradiction, as long as the models actually remain open and the industrial data remains where it is generated. For decision-makers, this means: The platform question is largely decided before the first application goes live. Anyone investing now in simulation, data pipelines, and security architecture is doing so on a foundation dictated by NVIDIA. This makes it all the more important to identify what sets Europe apart in this ecosystem—actuator technology, sensor technology, functional safety, and process expertise. It is likely no coincidence that Schaeffler, SICK, and TÜV SÜD are mentioned in NVIDIA’s announcement.
Related Exhibitors
Interested in news about exhibitors, top offers and trends in the industry?
Browser Notice
Your web browser is outdated. Update your browser for more security, speed and optimal presentation of this page.
Update Browser