Germany prepares state-funded Nvidia partnership to advance “physical AI”
Key Takeaways
- •Germany plans to commit hundreds of millions of euros to a partnership with Nvidia and about a dozen domestic companies to fund computing power for training robots and autonomous vehicles, with an unveiling reportedly planned for next week.
- •The initiative pursues two goals: supplying compute for physical AI used in robotics and self-driving vehicles, and building industrial AI models capable of competing with international counterparts.
- •Siemens, Audi, and Black Forest Labs are among the roughly twelve German participants, while Nvidia will assign engineers to play an active role in the project.
- •The program extends the Soofi consortium's approach, which received €20 million in federal funding and trained its first model on Deutsche Telekom's Industrial AI Cloud, a Munich facility built around roughly 10,000 Nvidia Blackwell GPUs under a €1 billion deal.
- •Nvidia CEO Jensen Huang is scheduled to meet Chancellor Friedrich Merz in Berlin on October 19 during a European tour that included talks with French President Emmanuel Macron and an additional $2 billion commitment to European AI.
Germany is laying the groundwork for a state-backed partnership with Nvidia (NASDAQ: NVDA) and roughly a dozen homegrown AI laboratories to fund “physical AI” for use in robots and autonomous vehicles. According to German business daily Handelsblatt, the government is preparing to commit funding in the range of hundreds of millions of euros to the effort, with an unveiling reportedly planned for next week.
Two goals: train robots, train industrial models
The initiative is divided into two main categories. The first is to put plans in place for physical AI, the computing capacity used by AI labs to train robots and self-driving vehicles. The second is to build industrial AI models with the strength and capacity to compete with their international counterparts.
Based on the reports, the money is intended to cover expenses for compute and model training. Nvidia is the go-to provider for work of this nature: the company’s chips and matched software train the majority of today’s AI systems, and frontier models from the likes of Anthropic and OpenAI both run on Nvidia hardware. State funding for that hardware layer is what sets the effort apart, placing the compute underpinning robotics and autonomous-vehicle development behind public backing.
A dozen firms and a reception in Berlin
About a dozen German companies are set to take part in the partnership, with Siemens, Audi and Black Forest Labs holding a front-row seat. Nvidia’s contribution includes assigning engineers who will play an active role in the project. The lineup brings together industrial, automotive and AI-lab players, mirroring the initiative’s two tracks — physical AI for robots and autonomous vehicles, and competitive industrial models.
The unveiling is reportedly slated for the “Invest in Germany” summit. Nvidia CEO Jensen Huang is scheduled to arrive in Berlin on October 19 for a reception hosted by Chancellor Friedrich Merz at Kraftwerk Berlin. A private meeting will take place, after which a joint meeting with Mercedes CEO Ola Källenius will be held.
Built on top of the Telekom cloud deal
The partnership builds on existing projects and takes the Soofi initiative to the next level. Soofi, short for Sovereign Open Source Foundation Models, is a consortium of universities, research bodies and AI firms that was created in September 2025. The Federal Ministry for Economic Affairs and Energy invested €20 million into Soofi, which was used to train its first model on Deutsche Telekom’s Industrial AI Cloud. The new program would extend that formula — public money flowing into models trained on domestically hosted compute — from foundation models into physical AI.
Telekom’s industrial AI cloud is itself an Nvidia project. Deutsche Telekom and Nvidia signed a €1 billion deal, announced on November 4, 2025, covering the operation of a Munich facility that is set to become the world’s first industrial AI cloud. The AI cloud is built around roughly 10,000 Nvidia Blackwell GPUs and is expected to increase Germany’s AI computing capacity by approximately 50%. The facility would make use of more than 1,000 Nvidia DGX B200 systems and deliver 0.5 exaflops of compute, to be applied across robotics simulation and digital twins.
One stop on a European tour
Germany is just one part of a larger picture. Huang has spent the better part of this week traveling through Europe’s largest economies. He met French President Emmanuel Macron in Paris on Tuesday after backing four French startups, while committing an additional $2 billion to European AI on top of the $ billion spent earlier. The Berlin schedule puts Nvidia’s CEO in direct contact with both the French president and the German chancellor in a single week, underlining how centrally AI infrastructure now sits in Europe’s economic diplomacy.
The success of the partnership will largely depend on the terms of the agreement. The amount of funds pledged, the scope of the partnership, and the agencies that sign onto the agreement will play a key role in determining whether the initiative succeeds — with those specifics expected to come into view at next week’s unveiling.
