Nvidia’s Huang urges Europe to build AI capacity with modular data centers

Nvidia CEO Jensen Huang urged Europe to build more AI computing capacity at the company’s GTC keynote in Berlin on October 21, 2026. His practical pitch was a modular data-center design: deliver computing systems already integrated inside containers, rather than leave that work until they reach the site.

Moving integration ahead of delivery

Nvidia calls the design DSX POD. Its keynote coverage describes a reference design for complete AI computing facilities using the company’s Vera Rubin platform. Partners build the systems inside modular containers, then integrate and validate them before delivery. It is a design for partners to implement, not simply a shipping box for individual chips.

Nvidia says this can shorten onsite deployment from six months to 14 days, once site utilities are ready. That condition limits what the comparison means: the claim concerns deployment at a prepared site, not the entire schedule for creating a new data center. The figures are Nvidia’s, rather than independently verified construction results.

Huang framed the approach as a supply-chain change, saying it would alter how people install and build data centers.

Europe’s computing shortfall

Europe has about 5% of the world’s AI computing capacity, compared with 75% in the United States, according to the Associated Press. That gap supplies the immediate context for Huang’s appeal. European developers need access to the large amounts of computing power used to train AI applications.

“The big bottleneck is compute,” Jeannette zu Fürstenberg, president and managing director of investment firm General Catalyst, said ahead of the speech. Her assessment puts the constraint beneath the applications themselves: developers need the physical computing resources to train them, not only ideas for what those applications should do.

Huang made infrastructure the starting point of his argument. He described AI as a chain running from energy through computers and models to applications. “Long before the AI models emerge, the infrastructure has to exist,” he said, comparing that prerequisite with networking for the internet and the grid for electricity.

Producing AI, not only consuming it

His economic case went beyond buying more capacity. “To be part of this AI economy, you don’t want to just be consuming the AI; you want to be producing the AI,” Huang said. He argued that Europe could build across that chain, rather than participate only at the application end.

Nvidia already has a substantial commercial network behind that argument. The company says it works with more than 12,000 partners across Europe. Huang opened by naming European businesses working with Nvidia on AI, including L’Oréal, Revolut, Spotify, Merck, Vodafone, Zara, Porsche, Nestlé and HSBC.

Those examples span consumer goods, finance, telecommunications and manufacturing. Huang’s message was that the customer base extends well beyond companies building language models. He predicted that every company would eventually become an AI company—a statement of his expected direction for adoption, not a claim that Europe has already reached it.