Nvidia held its GTC conference in Berlin from 20 to 22 October, with chief executive Jensen Huang giving the keynote on 21 October. The company used the event to sell Europe on building its own AI infrastructure, and the announcements leaned towards factories, robots and cars rather than new silicon.
AI factories in a box. The headline launch was DSX POD, which Nvidia calls “a reference design for complete NVIDIA Vera Rubin AI factories built by partners inside modular containers”. Partners assemble, cable and test the systems before shipping, so the site only needs power, cooling and networking ready. Nvidia says this cuts deployment “from six months to as little as 14 days” at a prepared site. Partners include Dell Technologies, HPE, Lenovo, Foxconn, Schneider Electric and Vertiv, with deliveries planned “worldwide in early 2027”. (Nvidia)
A reference humanoid. Nvidia named the Agile ONE SN from Agile Robots as a reference humanoid for its Isaac robotics platform. It runs Nvidia’s GR00T robot models on a Jetson Thor computer and is scheduled for early 2027. Nvidia also made Isaac Lab 3.0, its open-source robot training framework, generally available. (Nvidia)
Self-driving in Europe. Volvo Cars is adopting Nvidia’s Hyperion platform and Alpamayo driving software, following last week’s deals. Bolt picked Alpamayo for a European robotaxi programme starting with future Lucid vehicles, and plans to start collecting driving data in Madrid in 2027, subject to regulatory approval. Mercedes-Benz extended its partnership for Level 3 and Level 4 driving, where the car rather than the driver handles some or all of the driving in set conditions. (Nvidia)
Supply. Nvidia also announced deals with suppliers including Amkor, IBIDEN, Samsung Electro-Mechanics and KYEC. It plans a 2.5x increase in substrate capacity (substrates are the layer that connects a chip to the board) over three years, and expects its chip testing capacity to double from 2026 to 2027. (Nvidia)
Separately, Google Cloud said its A5X instances built on Nvidia’s Vera Rubin NVL72 systems will reach general availability next month. (Nvidia blog)
Why it matters: Nvidia announced no new chip in Berlin. It focused on getting existing Vera Rubin systems built and running faster, and on Europe’s push to own its AI infrastructure. The release dates to watch are early 2027, for both the container factories and the humanoid.