The Pulse
Germany Plans Hundreds of Millions for Nvidia AI Pact
Germany is preparing a state-backed industrial AI partnership with Nvidia, with federal funding in the hundreds of millions of euros. The plan would support industrial AI models and physical AI for robotics and autonomous vehicles, using co
AI.info Team · Written with AI ·
“What runs on the systems should be provided here in Germany. That way, you control your own data and your ability to act.”
Jensen Huang, CEO of Nvidia, speaking to ZDFheute
Germany is preparing a state-backed industrial AI partnership with Nvidia that would fund computing capacity and model development with hundreds of millions of euros. Handelsblatt first reported the plan on Thursday, Oct. 15, citing people familiar with it. Government sources confirmed the planned deal to ZDFheute on Friday, Oct. 16, as Nvidia CEO Jensen Huang discussed the proposal in a Berlin interview. The agreement is expected to be presented at the government’s “Invest in Germany” summit, scheduled for Oct. 19 and 20 in Berlin.
Robots, vehicles and industrial models
The initiative has two stated aims: build a foundation for “physical AI” used to train robots and autonomous vehicles, and develop powerful AI models for industrial applications. Physical AI systems interact with the real world rather than producing only text or images. The proposal would draw on Nvidia’s Nemotron and Cosmos model families, Handelsblatt reported.
The pact is planned around SOOFI, short for Sovereign Open Source Foundation Models, a German consortium launched in September 2025 by research organizations, universities and AI companies. The Federal Ministry for Economic Affairs has already provided SOOFI with €20 million, which funded training for its first model on Deutsche Telekom’s Industrial AI Cloud. The new initiative would support further model development and prepare the systems for use in industry and smaller businesses.
Federal funding would buy compute time
The federal government is prepared to invest a three-digit million-euro sum, Handelsblatt reported, primarily to buy computing capacity and train new models. The funding could come from money reserved for the federal government’s Hightech Agenda, the report said. Nvidia is expected to contribute engineers, while the roughly dozen German companies involved are expected to include Siemens, Audi and Black Forest Labs.
Using Telekom’s cloud gives the project an existing place to run its training workloads, rather than requiring a new computing site. But the proposed arrangement is not limited to infrastructure: the government’s money would help pay for compute and model training, while Nvidia would supply technical staff and its model families. The public agreement’s final budget, partners and terms have not yet been announced.
Huang’s data-control promise faces a practical test
Huang told ZDFheute that Nvidia supplies the computer systems, but that software and applications should be hosted in Germany so companies retain control of their data. He said Nvidia already works with Siemens, Audi and Mercedes on systems that train and control robots and industrial equipment. His argument draws a line between relying on foreign hardware and retaining local control over data and applications.
That distinction matters to Berlin’s stated goal of strengthening digital sovereignty, but it does not erase dependence on Nvidia’s chips and software. Local hosting can determine where industrial data is processed; it does not by itself make the underlying computing stack replaceable. The pact remains a plan rather than a publicly signed agreement, with its funding and governance terms still to be disclosed at the Berlin summit.