At the Consumer Electronics Show, Nvidia CEO Jensen Huang introduced the company’s latest Rubin computing architecture, positioning it as a significant advancement in AI hardware. Named after astronomer Vera Florence Cooper Rubin, the Rubin architecture features six distinct chips that work together, with the Rubin GPU at its core. This new architecture aims to address the growing computational demands of AI applications, representing a notable step forward in AI computing capabilities.
The Rubin architecture is a testament to Nvidia’s commitment to hardware innovation. Set to replace the Blackwell architecture, the Rubin chips have already secured partnerships with major cloud providers like Anthropic, OpenAI, and Amazon Web Services. Additionally, these chips will power prominent supercomputers such as HPE’s Blue Lion and the upcoming Doudna supercomputer at Lawrence Berkeley National Lab.
With a focus on enhancing storage and interconnection efficiency, the Rubin architecture integrates improvements in the Bluefield and NVLink systems, alongside the introduction of the new Vera CPU tailored for agentic reasoning tasks. Nvidia’s senior director of AI infrastructure solutions, Dion Harris, highlighted the architecture’s ability to meet the evolving memory demands of modern AI systems, particularly in cache-intensive workflows.
Source: TechCrunch