Archive · 22 Jul 2026
Nvidia's Vera Rubin AI superchip 10x more efficient than Blackwell
Nvidia has introduced a new AI superchip called Vera Rubin, which is 10 times more efficient than its current system, Grace Blackwell.
This new platform combines GPUs and CPUs to handle large AI workloads.
Nvidia is also selling the Vera CPU separately, aiming to compete with AMD and Intel.
The Vera Rubin system uses liquid cooling, reduces cabling, and offers significantly higher memory bandwidth.
OpenAI is already testing the new platform.
Nvidia's Vera Rubin NVL72 platform is 10 times more efficient than Grace Blackwell.
The platform combines 72 Rubin GPUs with 36 Vera CPUs for large AI workloads.
Nvidia is also selling the Vera CPU separately, expanding into the CPU market.
Vera Rubin uses 100% liquid cooling and reduces cabling for easier installation.
The platform offers nearly three times the memory bandwidth of Grace Blackwell.
- Who
- Nvidia
- What
- Launch of Vera Rubin AI superchip
- When
- During a technical briefing at Nvidia's headquarters
- Where
- Santa Clara, California
- Why
- To power the next generation of artificial intelligence and expand into complete AI computing systems
Nvidia's Strategic Shift
Competition in AI Chip Market
Nvidia's Vision
Nvidia's Strategic Shift
Nvidia aims to transition from being a leading GPU maker to a supplier of complete AI computing systems.
Competition in AI Chip Market
Nvidia's expansion into CPUs challenges traditional market leaders like AMD and Intel.
- Company
- Nvidia
- Platform
- Vera Rubin NVL72
- Efficiency
- 10 times more AI tokens per watt than Grace Blackwell
- Memory Bandwidth
- Nearly three times that of Grace Blackwell
- Cooling
- 100% liquid cooling
- Early Customer
- OpenAI