Nvidia Puts Rubin AI Platform Into Full Production, Open-Sources Alpamayo for Autonomous Vehicles

Overview

• Jensen Huang unveiled Alpamayo as an open-source vehicle reasoning model, with its training data released so automakers can evaluate and retrain it.
• Mercedes‑Benz vehicles using Nvidia’s autonomous stack are slated to reach U.S. roads in the first quarter of 2026, followed by Europe in the second quarter and Asia later in the year.
• Nvidia said the Vera Rubin platform is in full production and expected to ship broadly in the second half of 2026, with CoreWeave named as an early adopter.
• The six‑chip Rubin system targets major efficiency gains, including up to 10x lower inference token cost, 4x fewer GPUs for MoE training, and roughly 5x the AI compute versus prior chips, enabled in part by a proprietary data format.
• New system components include a context memory storage layer for faster long conversations and next‑generation network switches using co‑packaged optics, as executives highlighted strong demand and readiness to supply, including high interest in H200s in China pending licensing.