NVIDIA Launches Rubin Platform: A Quantum Leap in AI Computing Performance
NVIDIA’s CEO Jensen Huang inaugurated CES 2026 in Las Vegas by unveiling the company’s next-generation Rubin AI platform, now in full production. Named after the pioneering astronomer Vera Rubin, this platform signifies NVIDIA’s first extreme-codesigned, six-chip AI platform, set to revolutionize the future of artificial intelligence.
The Rubin platform boasts the Vera CPU with 88 custom Olympus cores and Rubin GPUs outfitted with HBM4 memory. This powerful combination pledges a staggering 5x enhancement in AI inference performance and a 10x reduction in token costs compared to the preceding Blackwell architecture. Furthermore, the system necessitates 4x fewer GPUs for training mixture-of-experts models, significantly slashing infrastructure costs for AI companies.
“Rubin arrives at exactly the right moment, as AI computing demand for both training and inference is skyrocketing,” Huang affirmed. Major cloud providers like Microsoft, AWS, and Google Cloud have pledged to deploy Rubin-based systems in the latter half of 2026. Microsoft plans to incorporate the platform into its next-generation Azure AI data centers and Fairwater AI superfactory sites.
The announcement also spotlighted Alpamayo, an open reasoning model family for autonomous vehicles, and broadened partnerships with Siemens for industrial AI applications. AI labs including OpenAI, Anthropic, Meta, and xAI are already strategizing to exploit the Rubin platform for training larger, more competent models.
Source: https://blogs.nvidia.com/blog/2026-ces-special-presentation/