Jensen Huang’s pitch for Europe’s role in artificial intelligence centered on producing AI, not simply using it. At Nvidia’s global conference in Berlin, he described a future built around data centers, software agents and robots, with Nvidia technology at the core.
A data center designed to arrive ready
The main infrastructure announcement was the DSX Pod, a modular “AI factory” that Nvidia describes as plug-and-play. The equipment comes assembled inside a large metal container and includes Nvidia’s latest Vera Rubin compute racks.
Huang said the system can be brought online in 14 days, compared with the months or years that conventional construction can take. The shorter setup could matter in Europe, where infrastructure projects are described as slow. Nvidia also says faster deployment would reduce the time computing equipment sits idle before a data center begins operating.
The proposal puts speed and packaged infrastructure at the center of Huang’s case for expanding AI capacity. Instead of waiting for a facility to be built piece by piece, organizations could receive a prepared unit and move toward operation more quickly.
Robots and industry are part of the plan
Huang called humanoid robotics the “next great industry” and argued that Europe’s manufacturing heritage could put it in a strong position. The vision extends beyond humanoid machines: he said telecom operators are adding Nvidia GPUs to cell towers, and announced partnerships with European carmakers VW and Mercedes on autonomous vehicles.
Nvidia also introduced a full-stack robotics platform intended to standardize configuration. The company plans to release Agile One, a preconfigured reference robot, in the first quarter of next year. The stated aim is to give developers a starting point for building physical AI systems.
Together, the platform and reference robot suggest a push to make robotics development easier to begin. Huang’s remarks framed physical AI as a broad industrial opportunity, with applications reaching beyond robot makers to manufacturing, telecommunications and cars.
More software agents could mean more computing
Huang expects companies to use both closed models and open models customized to their own data. At the conference, he presented Nvidia’s open-source models as capable of frontier performance after customization with an organization’s proprietary information.
He also argued that AI agents will bring more people into software tools. Because agents can reason, iterate and use tools, Huang said, they increase token consumption and demand for computing. That demand, in turn, supports the case for more Nvidia GPUs.
This links the conference’s software predictions to its infrastructure pitch: if more agents run more tasks, they need computing capacity. The modular data center is one answer Nvidia is proposing for making that capacity available faster.
A vision built around Nvidia technology
The 90-minute presentation moved from products to wider predictions, presenting Nvidia as a technology provider across industries. Small Microduck robots made by recent Nvidia acquisition Hugging Face appeared at the end, carrying out Huang’s instructions.
The event also included a moment of protest. Two Greenpeace protesters unfurled a banner reading “Clean up your dirty business,” and Huang told security to “Go easy” as they were escorted out.
Huang’s message to the Berlin audience was that Europe could help produce AI through computing infrastructure, customized models and physical systems. The DSX Pod, robotics platform and planned Agile One robot offer concrete parts of that pitch, while the larger promise depends on the industries and developers that choose to build on them.