NVIDIA Unveils Rubin Platform: 10x Cheaper Inference, Major Labs Onboard

In January 2026, NVIDIA presented the Rubin platform, featuring CPU Vera, GPU Rubin, NVLink 6, superNIC ConnectX-9, DPU BlueField-4, and Spectrum-6 switch. The platform promises 10x inference cost reduction and 4x fewer GPUs for mixture-of-experts models.

Key Highlights

• Components: Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4, Spectrum-6
• Efficiency: 10x reduction in inference costs
• GPU Savings: 4x fewer GPUs needed for MoE models
• Customers: OpenAI, Anthropic, and Meta plan to adopt
• Focus: Next-generation AI training and inference

Industry Adoption

Major AI laboratories including OpenAI, Anthropic, and Meta are planning to use the new Rubin chips, signaling a new era in AI hardware capabilities and efficiency. The platform addresses the growing demand for more cost-effective AI inference.