During CES, Nvidia unveiled the Rubin platform, a new generation of AI infrastructure comprising six chips. The platform is designed to deliver AI supercomputers at a lower cost and accelerate the adoption of artificial intelligence. The first Rubin systems will be available in the second half of 2026 via AWS, Microsoft, and Google.
The Rubin platform combines six chips into a single integrated system: the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Ethernet Switch. “Rubin arrives at exactly the right moment, as AI computing demand for both training and inference is going through the roof,” said Nvidia CEO Jensen Huang. The platform is named after American astronomer Vera Florence Cooper Rubin.
The new architecture promises significant performance improvements. According to Nvidia, Rubin delivers up to 10 times lower cost per token than the Blackwell platform when inferencing mixture-of-experts (MoE) models. In addition, MoE models can be trained with 4 times fewer GPUs compared to its predecessor.
Five key innovations
The platform introduces five technological breakthroughs. The sixth-generation NVLink offers 3.6TB/s bandwidth per GPU, while the entire Vera Rubin NVL72 rack delivers 260TB/s. That’s more bandwidth than the entire internet, according to Nvidia. The new Vera CPU is specifically designed for agentic reasoning and features 88 custom Olympus cores with full Armv9.2 compatibility.
The Rubin GPU includes a third-generation Transformer Engine with hardware-accelerated adaptive compression and delivers 50 petaflops of NVFP4 compute for AI inferencing. The Vera Rubin NVL72 rack system is the first to offer Nvidia Confidential Computing at rack scale, keeping data secure across CPU, GPU, and NVLink domains. The second-generation RAS Engine provides real-time health checks and proactive maintenance, with a modular design enabling up to 18 times faster assembly than Blackwell.
Broad ecosystem support
Rubin will see widespread adoption. Cloud providers such as AWS, Google Cloud, Microsoft, and Oracle Cloud Infrastructure are among the first to roll out Vera Rubin-based instances in 2026. Microsoft is integrating Nvidia Vera Rubin NVL72 rack systems into next-generation AI data centers, including future Fairwater AI factories. In addition, Dell Technologies, HPE, Lenovo, and Supermicro are supplying servers based on Rubin products.
OpenAI, Anthropic, Meta, and xAI have committed to embracing the Rubin platform to train larger models and run long-context, multimodal systems at lower latency and cost. Nvidia is in full production, and Rubin-based products will be available through partners in the second half of 2026.