Nvidia Unleashes Next-Generation Rubin AI Platform, Offering Quintupled Performance
At CES 2026, AI chip titan Nvidia made a significant announcement. CEO Jensen Huang unveiled the next-generation Rubin AI platform, a powerhouse of six new chips designed to provide unparalleled AI computing power. The platform comprises the Rubin GPU, Vera CPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Ethernet Switch.
The Rubin GPU, the platform’s crown jewel, is built on TSMC’s 3nm process with 336 billion transistors. This GPU promises a monumental leap in AI performance, delivering five times the inference performance of the current Blackwell architecture, backed by 50 petaflops of FP4 compute power. Moreover, the new platform slashes the training requirements to merely a quarter of the GPUs needed for mixture-of-experts models and reduces inference token costs by up to 10x.
The Vera Rubin NVL72 rack-scale system amalgamates 72 GPUs with 3.6 exaflops of AI compute in a single rack. Major cloud providers, including AWS, Google Cloud, Microsoft, and Oracle Cloud Infrastructure, are set to be among the first to deploy Vera Rubin-based instances. Nvidia confirmed that the platform is in full production, with availability slated for the second half of 2026.
This announcement cements Nvidia’s supremacy in the AI infrastructure market. As of January 2026, the company boasts a market valuation of approximately $4.6 trillion.
Source: Globe Newswire