NVIDIA's Rubin unveiled, can cut AI costs up to 10x
Technology
NVIDIA just pulled the wraps off its Rubin platform at CES 2026, and it's a big leap for AI tech.
Rubin can cut the cost of running AI models by up to 10 times and requires four times fewer graphics cards to train mixture-of-experts (MoE) models compared to the older Blackwell platform.
The setup packs six chips into one powerhouse, including the new Vera CPU and Rubin GPU, with the Rubin GPU capable of delivering up to 50 petaflops of NVFP4 computational power.
Rubin coming to AWS Google Microsoft
Starting in the second half of 2026, you'll see Rubin rolling out on Amazon Web Services, Google Cloud, and Microsoft.
NVIDIA hopes this move will make advanced AI more accessible by lowering costs and making it easier for companies to build smarter tech without breaking the bank.