Nvidia launched its next-generation Rubin platform on Monday, introducing a suite of six new chips designed to power advanced artificial intelligence systems and slash the cost of training and inference, as AI adoption accelerates globally.
The Rubin platform, named after pioneering astronomer Vera Rubin, is aimed at meeting surging demand for compute power in AI development. It integrates six key components — the Nvidia Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU and Spectrum-6 Ethernet Switch — through what the company describes as “extreme codesign,” allowing it to optimize performance and efficiency across AI workloads.
“Rubin arrives at exactly the right moment, as AI computing demand for both training and inference is going through the roof,” said Nvidia CEO Jensen Huang. “With our annual cadence of delivering a new generation of AI supercomputers — and extreme codesign across six new chips — Rubin takes a giant leap toward the next frontier of AI.”
Among the Rubin platform’s core innovations are the latest generation of Nvidia NVLink interconnect technology, a new Transformer Engine, Confidential Computing and a RAS (reliability, availability, and serviceability) Engine, all aimed at powering large-scale “agentic AI” and mixture-of-experts (MoE) model inference. According to Nvidia, Rubin enables MoE model training using one-fourth the GPUs compared to its predecessor, while inference token costs are reduced by a factor of seven compared to the Blackwell platform.
The hardware will be available in rack-scale and system configurations: the Nvidia Vera Rubin NVL72 and the HGX Rubin NVL8, respectively.
Industry support for the platform spans a wide range of major AI labs, cloud providers, and hardware partners. Early adopters include Amazon Web Services, Microsoft, Meta, OpenAI, Google, Dell Technologies, Oracle, Lenovo, Cisco, HPE, Supermicro, Anthropic, Cohere, Mistral AI and Elon Musk’s xAI.
“Intelligence scales with compute,” said OpenAI CEO Sam Altman. “When we add more compute, models get more capable, solve harder problems and make a bigger impact for people. The Nvidia Rubin platform helps us keep scaling this progress so advanced intelligence benefits everyone.”