Nvidia Launches Next-Generation Rubin AI Compute Platform at CES 2026 (www.servethehome.com)

Nvidia has unveiled its next-generation AI compute platform, Rubin, at CES 2026, marking a significant step in the evolution of its hardware offerings. The Rubin platform, which includes a new GPU architecture and associated chips, promises to enhance AI performance dramatically, with up to 5x improved AI inference and 3.5x improved AI training performance compared to the previous generation, Blackwell. This advancement is underpinned by an updated transformer engine that supports compression, aiming to solidify Nvidia's leadership in the rapidly evolving AI landscape. Key components of the Rubin platform feature a dual-die GPU fabbed on TSMC's 3nm process, boasting 336 billion transistors and 50 PFLOPS for inference. The new NVLink Switch chips will double inter-GPU bandwidth to facilitate larger-scale AI models and systems. Additionally, the introduction of the Vera CPU, with 88 cores and significant memory enhancements, signifies Nvidia’s focus on integrated system design for high-performance applications. With advanced networking solutions such as the ConnectX-9 NIC and Spectrum-6 Ethernet switches featuring co-packaged optics, Nvidia's Rubin platform is tailored for robust performance, efficiency, and reliability in AI workloads, positioning it as a pivotal player in the future of AI/ML technology.