In Berlin, NVIDIA and Germany lay out a plan for industrial AI, targeting manufacturers and the Mittelstand

Germany is aiming to turn its industrial strengths into an advantage for the next wave of artificial intelligence, with NVIDIA positioning itself as a technology partner for “physical AI” in factories, vehicles and related sectors. On Monday, German Chancellor Friedrich Merz met with NVIDIA founder and CEO Jensen Huang in Berlin at Kraftwerk, a former power station repurposed as a venue for technology and industry talks, where they discussed how industrial AI could be adopted across the country’s ecosystem.

A central element of the meeting was a joint declaration of intent signed earlier that day at the Invest in Germany Summit. The Federal Ministry for Economic Affairs and Energy and the Federal Ministry for Research, Technology and Space, along with NVIDIA, agreed to work on a joint initiative intended to advance cutting-edge AI models for “physical AI,” or AI systems that can interact with, control, or improve real-world operations. NVIDIA described the effort as a way to formalize and expand its collaboration with Germany’s federal government and the Industry 4.0 platform, an industry-and-policy group that brings together politics, industry, research and trade associations.

The declaration aligns with Germany’s stated goal of systematically bringing industrial AI into widespread use and establishing a state-of-the-art industrial AI ecosystem in Europe. During the conversation, Merz emphasized that integrating AI into Germany’s manufacturing and automotive industries is a strategic priority, arguing that the country can be among the world’s leaders in deploying AI across companies that build and supply machinery and vehicles. He pointed specifically to the “Mittelstand,” referring to Germany’s network of small and medium-sized businesses, and urged them to open their doors so AI tools can use the data already held inside their operations.

Huang reinforced that message by framing AI as an augmenting technology rather than a replacement for industrial know-how. He told the audience that the knowledge held by companies cannot be replaced, and that AI can only extend their capabilities. The two leaders tied this approach to Germany’s differentiators in manufacturing and engineering, describing AI as a multiplier for established strengths in math, science and practical production know-how.

Beyond policy and partnership, the Berlin meeting also put investment and execution on the agenda. Huang said NVIDIA is making nearly $10 billion in commitments across Europe to support companies, universities and research institutions. As part of that effort, he announced planned investments in four German startups, naming Agile Robots, Avelios Medical, Mirelo and Reflex Aerospace.

Agile Robots, based in Munich, builds AI-powered robots including robotic arms and its Agile ONE humanoid, aimed at uses spanning manufacturing, logistics and healthcare. Avelios Medical, also in Munich, develops an AI-native hospital information system described as a core system of record for clinical and administrative workflows, integrating AI into how hospitals run so doctors and nurses can spend more time on patient care. Mirelo, a German startup focused on media production, creates AI models to generate sound effects, speech and music intended to match video visuals. Reflex Aerospace, located in Berlin, designs and builds high-performance satellites, with the company stating that they are ready for launch in about 12 months instead of years.

The meeting occurred alongside industrial indicates from major German manufacturers. The NVIDIA blog also noted that Siemens announced a partnership with NVIDIA aimed at advancing AI-driven, autonomous factories. Siemens showcased new AI solutions at its Erlangen facility, including a virtual production ramp-up approach meant to reduce engineering effort by up to 85%, and agent-based production planning intended to enable more adaptive manufacturing and cost savings of up to 20%, according to what was presented at the event.

NVIDIA also referenced European high-performance computing research in the run-up to the meeting. The company said its researchers used JUPITER, described as Europe’s largest supercomputer at the Jülich Supercomputing Centre, to run a record simulation of a fault-tolerant quantum computer. NVIDIA said the work involved a 104-qubit simulation, characterized as the largest and most detailed of its kind to date.

What remains unclear is the timeline and scale of the Germany-NVIDIA joint initiative beyond the existence of the declaration of intent and the broad goal of accelerating physical AI model development. The announcement does not specify budget figures for the joint program, the performance targets, or which industrial use cases will be prioritized first, and it does not disclose how procurement, governance, or data-sharing arrangements will be handled. Still, by combining a government-backed declaration with targeted startup investments and ongoing industrial partnerships, NVIDIA is betting that Germany’s industrial AI push will be both policy-driven and commercially grounded.