Particle.news
Download on the App Store

NVIDIA Puts Rubin AI Platform Into Production as Qualcomm Unveils New Robotics Chips at CES

The twin rollouts point to AI infrastructure extending from cloud supercomputers to low‑power robots and vehicles.

Overview

  • NVIDIA introduced Vera Rubin, an extreme‑codesigned six‑chip platform now in full production, promising roughly one‑tenth the inference token cost versus its prior platform and major performance gains over Blackwell.
  • Rubin integrates new GPUs and Vera CPUs with NVLink 6, Spectrum‑X networking, ConnectX‑9 SuperNICs and BlueField‑4 DPUs, plus an AI‑native context‑memory storage tier that delivers up to 5x higher token throughput and power efficiency for long‑context inference.
  • NVIDIA said partner rollouts of Rubin systems begin in the second half of 2026, naming Microsoft Azure and CoreWeave among the first cloud providers planning deployments.
  • For automotive and robotics, NVIDIA released Alpamayo open VLA models with 1,700 hours of driving data and the AlpaSim simulator, and said the first passenger car using Alpamayo on NVIDIA DRIVE will be the new Mercedes‑Benz CLA.
  • Qualcomm announced new Dragonwing Q‑7790 and Q‑8750 IoT processors and an end‑to‑end robotics architecture powered by its Dragonwing IQ10 Series to bring efficient, scalable AI to personal service robots, industrial AMRs and full‑size humanoids.