NVIDIA unveils Rubin at CES to cut AI inference costs
Technology
NVIDIA just dropped its new Rubin platform at CES, and it's all about making AI faster, cheaper, and easier to scale.
Rubin is designed for big AI projects: it cuts down on costs for running AI models (those "inference token costs") and needs fewer graphics cards than the previous Blackwell platform.
Inside, you'll find six powerful chips working together, including an energy-efficient Vera CPU with 88 custom Olympus cores.
AWS, Google Cloud to deploy Rubin
Rubin will roll out in different versions, like the NVIDIA Vera Rubin NVL72, and Amazon Web Services and Google Cloud are set to roll out to partners sometime in the second half of 2026.
This should make advanced AI tools more practical for large-scale AI deployment.