NVIDIA unveils Rubin: Next-gen AI supercomputer drops at CES 2026
NVIDIA just introduced Rubin, its latest AI supercomputing platform, at CES 2026 in Las Vegas.
Built for seriously demanding workloads, Rubin brings together six advanced components—including the new Rubin GPU and Vera CPU.
It's already in production and set to hit the market later this year.
What makes Rubin special?
Rubin's NVL72 rack packs a massive punch, with the GPUs boasting 50 petaflops of performance with ultra-fast memory, while the Vera CPUs offer 88 custom cores.
Plus, it's the first rack-scale platform with third-gen Confidential Computing across all major parts—so security gets an upgrade too.
Why does it matter?
If you're into AI or tech trends, here's the cool part: Rubin can cut operational costs for data centers by slashing inference token costs up to tenfold and reducing hardware needs for big models by four times compared to Blackwell.
For companies building next-level AI tools—or anyone curious about where AI is headed—Rubin is a big leap forward in both speed and efficiency.