NVIDIA Launches Rubin AI Platform

(RTTNews) - Rubin, NVIDIA's first extreme-codesigned, six-chip AI platform, is now in full production, marking a major leap in accelerated computing. Introduced by CEO Jensen Huang at CES 2026 in Las Vegas, Rubin integrates GPUs, CPUs, networking, and storage in a unified design to eliminate bottlenecks and deliver AI tokens at one-tenth the cost.

Huang explained that extreme codesign — building all components together — is essential to scale AI to gigascale, enabling faster model training and dramatically reducing costs. The Rubin platform also debuts an AI-native storage system that boosts long-context inference with fivefold improvements in speed, efficiency, and cost-effectiveness.

Alongside Rubin, Huang unveiled Alpamayo, a new family of open reasoning models for autonomous driving. Trained on NVIDIA supercomputers, Alpamayo joins NVIDIA's portfolio of open models spanning healthcare, climate science, robotics, embodied intelligence, and multimodal AI. The first passenger car featuring Alpamayo — the new Mercedes-Benz CLA — will debut in the U.S. this year.

Huang also highlighted DGX Spark, a desktop supercomputer enabling personal AI agents, and showcased NVIDIA's Cosmos foundation model for simulation and physical reasoning. Together, these innovations reflect NVIDIA's mission to bring AI from supercomputers into everyday devices and the physical world.

"Our vision is that someday, every single car, every single truck will be autonomous," Huang said, underscoring NVIDIA's ambition to drive the next era of AI.