Nvidia leans on Israeli R&D in new Rubin AI platform
Key networking and processing components developed in Israel.
Nvidia on Monday unveiled Rubin, a new platform for training and running artificial intelligence models that the company says can sharply reduce the computing power required for both training and inference.
The platform integrates six chips into what Nvidia describes as a tightly coupled AI “supercomputer.” At its core is a new AI processor, Rubin, accompanied by a central processing unit (CPU) known as Vera, a data processing unit (DPU), and three communications chips. Several of these components were developed at Nvidia’s research and development center in Israel or with significant involvement from the company’s Israeli engineering teams.
“Rubin arrives at exactly the right moment, as AI computing demand for both training and inference is going through the roof,” said Jensen Huang, founder and CEO of Nvidia. “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.”
The platform is named after Vera Rubin, the Jewish-American astronomer whose work on galaxy dynamics played a central role in the discovery of dark matter. In addition to the Rubin AI chip and the Vera CPU, the system includes Nvidia’s BlueField-4 DPU and three networking components, the ConnectX-9 SuperNIC, NVLink 6 Switch, and Spectrum-6 Ethernet Switch, all developed in Israel.
According to Nvidia, Rubin delivers substantial performance gains compared with its previous-generation Blackwell platform. The company says the system can train Mixture of Experts models using four times fewer AI chips, while reducing inference costs by as much as 90%.
Sam Altman, CEO of OpenAI: “Intelligence scales with compute. 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.”
Dario Amodei, co-founder and CEO of Anthropic: “The efficiency gains in the NVIDIA Rubin platform represent the kind of infrastructure progress that enables longer memory, better reasoning and more reliable outputs. Our collaboration with NVIDIA helps power our safety research and our frontier models.”
Nvidia said systems based on the Rubin platform are expected to become available to enterprise customers in the second half of 2026.
