Nvidia CEO Denies Vera Rubin Platform Delays

Jensen Huang confirms Nvidia's Vera Rubin AI platform is in production, refuting reports of roadmap delays for next-gen hardware.

Nvidia CEO Jensen Huang has denied reports of delays regarding the company's next-generation Vera Rubin AI platform, stating that the hardware is already in production. This follows Nvidia's previous confirmation of production in January and sampling in February. The platform features Vera CPUs, Rubin GPUs, and Vera Rubin NVL72 systems. Despite Huang's statements and a spokesperson's claim that the roadmap remains intact, SemiAnalysis reported a delay for the Kyber NVL144 rack-scale solution from 2027 to 2028 due to PCB midplane manufacturing challenges. The Kyber architecture connects 144 Rubin Ultra GPUs via copper-based NVLink 7 fabric. Reports also suggest the cancellation of the NVL72x2 design and a quad-compute-chiplet Rubin Ultra, alongside postponements for the NVL576 configuration. Meanwhile, competitors continue development, with AMD's Mega Pod featuring Verano CPUs and Google's TPU 8t reaching 9,600 chip packages per domain.