Jensen Huang’s Vision for Europe Runs on Plug-In Data Centers

At GTC in Berlin, Jensen Huang unveiled the DSX Pod, a prebuilt AI data center that ships assembled in a large metal container packed with Vera Rubin racks and, Nvidia says, can come online in 14 days rather than months or years. For a continent where big infrastructure projects move slowly, that is a direct play at the bottleneck.

Huang paired it with a full-stack humanoid robotics platform, a reference robot called Agile One due in the first quarter of next year, autonomous-vehicle partnerships with VW and Mercedes, and a prediction that every company will mix closed models with customized open ones.

We like the core idea. Europe's AI gap is less about talent than about power, permits and construction timelines, and a modular unit that arrives ready to run attacks the real constraint. Huang's line that Europe should produce AI, not just consume it, is a fair challenge, and faster deployment is one practical way to answer it.

But look at the shape of the offer. Compute, robots, cars, cell towers, open models: Nvidia presented itself as the base layer of all of it. A plug-in data center is speed, and it is also lock-in by the container load. European leaders who talk about sovereignty should notice that sovereignty built entirely on one American supplier's stack is a narrower thing than it sounds.

Then there is energy. Shrinking build time doesn't shrink power draw. A 14-day data center still needs grid capacity, cooling and siting decisions that communities get a say in. The Greenpeace protesters who unfurled a banner during the keynote were making a crude version of a real point, and Huang's argument that agents will keep pushing token consumption and compute demand higher only sharpens it.

Our read: the DSX Pod is a smart product for a real problem, and Europe should use it where it helps. I'm watching whether governments pair fast deployment with honest grid planning and some diversity of suppliers, or simply swap slow construction for fast dependence.

The March GTC in San Jose centered on inference and efficient compute; seven months later the Berlin keynote shifted to power, agents and physical AI. The roughly 90-minute talk at the Tempodrom drew a crowd of about 4,000 and featured Microduck robots from Hugging Face, which WIRED notes Nvidia recently acquired.