Nvidia Unveils Vera Rubin AI Platform at CES

Nvidia Corp. unveiled its next-generation Vera Rubin AI computing platform on Monday at the Consumer Electronics Show in Las Vegas, positioning the technology as a major advancement in artificial intelligence hardware amid surging global demand.

Chief Executive Officer Jensen Huang announced that the Rubin platform, comprising six interconnected chips, is already in full production and slated for customer deployment later this year. The flagship component integrates one Vera central processing unit with two Rubin graphics processing units into a single superchip, designed to enhance efficiency for complex AI workloads such as training large language models and running inference tasks. Nvidia executives stated that the platform delivers 3.5 times the performance of its predecessor, the Blackwell architecture, in model training and five times the speed in inference operations, while requiring roughly one-fourth the number of chips for equivalent results.

Huang highlighted the platform's innovations during his keynote address, noting that it achieves these gains through a proprietary data format that Nvidia hopes the industry will adopt. "This is how we were able to deliver such a...