Home / Technology / Nvidia Unveils Vera Rubin: AI's Next Supercomputer
Nvidia Unveils Vera Rubin: AI's Next Supercomputer
17 Mar
Summary
- Nvidia launched Vera Rubin, a new AI platform built from seven chips.
- It promises 10x more inference throughput per watt than Blackwell.
- Key customers include OpenAI, Meta, Anthropic, and major cloud providers.
Nvidia has unveiled Vera Rubin, a comprehensive new AI computing platform that is now in full production. This platform integrates seven breakthrough chips, designed to accelerate artificial intelligence workloads and represents a significant advancement over existing systems.
This generational leap promises up to 10 times greater inference throughput per watt and a tenfold reduction in cost per token compared to Nvidia's recent Blackwell systems. The platform has garnered support from leading AI companies including Anthropic, OpenAI, and Meta, as well as all major cloud providers.
CEO Jensen Huang highlighted Vera Rubin as the beginning of the greatest infrastructure buildout in history, specifically designed to power autonomous AI agents. This shift from traditional chatbots to agentic AI necessitates a fundamental redesign of computing infrastructure, balancing compute, memory, storage, and networking.
Nvidia is also expanding its software and open-source model offerings to support this new hardware. Initiatives like the Agent Toolkit and the Nemotron Coalition aim to foster innovation and solidify Nvidia's position as a foundational provider for the AI industry.
The platform's reach extends beyond data centers, with applications in healthcare, autonomous vehicles, and even space computing, showcasing its versatility and broad impact across various sectors.
