Merz and Huang Forge Berlin AI Pact

On 19 October 2026, Kraftwerk Berlin hosted a fireside conversation between Friedrich Merz and Huang, the chief executive of Nvidia, during the two-day Invest in Germany Summit. The Deutschland portal reported almost 200 business leaders in the room. Both sides used the evening to sign a joint declaration and to anchor cooperation in the Industry 4.0 Platform, with the stated goal of carrying AI into German factories and offices. The summit recap on the Nvidia blog confirms the details.

Huang traced the relationship back to 2000, when a partnership with the German firm ELSA carried Nvidia graphics into professional workstations and, from there, to carmakers. Audi followed in 2005, moving the collaboration inside the vehicle, and the Mercedes link later extended to self-driving work on the flagship sedan, whose chief executive handed Merz a scale replica on stage. Heise recalls that ELSA went bankrupt in 2002, which tempers the nostalgia. The point of the history was simple: Germany helped shape the company long before the current AI boom.

Huang then sketched the works of tomorrow: each firm will run two plants, one physical and one virtual, with the virtual replica faithful enough to stand in for the real thing. Engineers will spend much of their time inside that replica, designing vehicles, testing production lines, and training robotic helpers before anything moves on the real floor. Vehicles will carry their own onboard intelligence, and software will run the plants that assemble them. The image is deliberately grand, yet the practical claim is narrow: design and manufacturing collapse into one continuous loop.

Merz directed his warmest appeal at the Mittelstand, urging smaller and mid-sized firms to welcome AI tools and to put company data to work. The plea matters because firm-level data is, in his telling, the decisive edge of the country. CNBC summit coverage put that appeal at the center of the evening. Huang answered with a sovereignty argument: national data holds the culture, principles, and know-how of a country, so it should be refined at home into domestic intelligence capacity rather than shipped abroad and bought back.

The coming surge, Huang argued, belongs to embodied AI: robots, production lines, and vehicles that act in the material world. Physical production and craft knowledge resist replacement by pure software, he said, because no all-round system can match a specialist in electromechanical engineering. Such tools therefore augment human capability instead of replacing it. For an aging workforce, that message landed well: let experienced staff teach the tools, and keep the resulting know-how in Germany.

Huang placed the bet in a longer history. Electrification, he noted, produced three frontrunners: Germany, America, and Japan. The computer age then shifted heavy production toward China, which grew comfortable in both manufacturing and computing. Now comes a reset, because computers increasingly program themselves once people show them the desired outcome. Past software advantages therefore matter less, while atoms still matter as much as electrons: digital tools will not erase physical production.

Money followed the rhetoric. Huang said Nvidia put almost $1 billion into Germany over the past year and added four names this week: Agile Robots, Avelios Medical, Mirelo, and Reflex Aerospace. Tech.eu listed the four newcomers alongside the earlier $300 million round for Black Forest Labs. The company also recommitted to a roughly $10 billion European pledge and pointed to its Berlin office, which now employs about 1,000 people and contributed several core breakthroughs.

The industrial story extends beyond startups. Siemens is turning its plants toward the self-steering factory, using operational data with tools such as Nemotron and Cosmos. Earlier German hopes such as Gaia-X and Catena-X showed how hard shared data spaces are to build. Heise adds a cool-headed reminder here: grand announcements need delivery dates and open interfaces before family-run suppliers can rely on them.

Regulation closed the evening on a sober note. Huang offered a three-part answer: invent protection technology first, since there is nothing to govern without it; spread proven routines across companies, especially young ones; and govern applications rather than the underlying techniques. Merz welcomed the balance and pointed to a Berlin win in Brussels: industrial equipment stays outside the bloc-wide AI statute, with the Digital Omnibus meant to simplify the rest. Politico reported behind the scenes that the equipment exemption survived the talks. Merz added that Europe alone is not enough and called for global safety rules.

The evening ended with promises rather than signatures. Merz invited Huang back for a yearly review of progress, hoping to greet a satisfied investor in his office. Huang praised the opening of retirement savings to venture funding and cheered the statistic that half of the firms founded in Germany during the previous 24 months work on AI. His closing line drew laughter: may his German investments strike gold.