Vera Rubin is in full production. We just kicked off the next...
NVIDIA@nvidia
29 views
Jan 06, 2026
~1 min read
Advertisement
1
Vera Rubin is in full production.
We just kicked off the next generation of AI infrastructure with the NVIDIA Rubin platform, bringing together six new chips to deliver one AI supercomputer built for AI at scale.
Here are the top 5 things to know 🧵
We just kicked off the next generation of AI infrastructure with the NVIDIA Rubin platform, bringing together six new chips to deliver one AI supercomputer built for AI at scale.
Here are the top 5 things to know 🧵
0:27
2
1. Vera Rubin is six new chips, one AI supercomputer
It brings together the NVIDIA Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4, and Spectrum-6 Ethernet.
They are co-designed to operate as one platform.
It brings together the NVIDIA Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4, and Spectrum-6 Ethernet.
They are co-designed to operate as one platform.
3
2. The performance and efficiency gains are massive.
Compared with NVIDIA Blackwell, Rubin delivers up to 10x lower token cost and trains mixture-of-experts models with 4x fewer GPUs, enabled by extreme co-design, with all chips built by NVIDIA. And these cost savings will make SOTA AI easier to scale and deploy.
Compared with NVIDIA Blackwell, Rubin delivers up to 10x lower token cost and trains mixture-of-experts models with 4x fewer GPUs, enabled by extreme co-design, with all chips built by NVIDIA. And these cost savings will make SOTA AI easier to scale and deploy.
0:05
4
3. Rubin is built for rack-scale systems.
NVIDIA Vera Rubin NVL72 combines 72 Rubin GPUs and 36 Vera CPUs with NVLink 6, so the rack functions as a single compute unit. NVIDIA Vera Rubin pod combines 1,152 Vera Rubin GPUs in 16 NVL72 racks.
This approach is designed for reliability, efficiency, and nonstop operation at scale.
NVIDIA Vera Rubin NVL72 combines 72 Rubin GPUs and 36 Vera CPUs with NVLink 6, so the rack functions as a single compute unit. NVIDIA Vera Rubin pod combines 1,152 Vera Rubin GPUs in 16 NVL72 racks.
This approach is designed for reliability, efficiency, and nonstop operation at scale.
5
4. NVIDIA Context Memory Storage Platform: Extending context memory to enable next-level agentic AI.
A New POD-Level Context Memory Extension Platform, powered by BlueField-4 and Spectrum-X, delivering 5X higher Tokens per Second and 6X Better Power Efficiency vs Traditional Enterprise Storage.
A New POD-Level Context Memory Extension Platform, powered by BlueField-4 and Spectrum-X, delivering 5X higher Tokens per Second and 6X Better Power Efficiency vs Traditional Enterprise Storage.
6
5. Networking CPO
NVIDIA Spectrum-6 Ethernet is the next generation of Ethernet for AI networking, built to scale Rubin-based AI factories with higher efficiency and greater resilience, and enabled by 200G SerDes communication circuitry, co-packaged optics and AI-optimized fabrics.
NVIDIA Spectrum-6 Ethernet is the next generation of Ethernet for AI networking, built to scale Rubin-based AI factories with higher efficiency and greater resilience, and enabled by 200G SerDes communication circuitry, co-packaged optics and AI-optimized fabrics.
7
NVIDIA Rubin is designed to make advanced AI more efficient, more scalable, and more practical to deploy in the real world.
Learn more from NVIDIA CEO Jensen Huang’s CES keynote and the full announcement: nvda.ws/4syWcYb
Learn more from NVIDIA CEO Jensen Huang’s CES keynote and the full announcement: nvda.ws/4syWcYb