Nvidia Says Vera Rubin Is 10 Times Faster in AI Than Grace Blackwell
Nvidia has announced that its new Vera Rubin platform is entering mass production. This platform is made to support agentic AI around the world.
According to Nvidia, the market no longer wants only single chips. It now needs rack-scale supercomputers that can handle complex tasks with many steps.
Unlike older AI models that give simple answers, agentic AI can carry out tasks on its own. Nvidia CEO Jensen Huang explained that one prompt can start a process with thousands of reasoning steps, including data searches, tool use, and decision-making.
Nvidia’s Vera Rubin Promises a Major Jump Over Grace Blackwell
To handle this huge workload, Vera Rubin offers much stronger performance. Nvidia says it gives 10 times more agent processing performance than the previous Grace Blackwell generation.
The platform works as one large supercomputer made from five special racks. It includes the new Rubin GPUs, the new Vera CPU, Nvidia Groq 3 LPX inference accelerators, BlueField-4 STX storage, and Spectrum-6 SPX Ethernet connectivity.
The new Vera CPU is also said to be twice as efficient and 50% faster at processing data than traditional data center CPUs.
Vera Rubin has strong support from the industry. More than 150 technology partners in Taiwan are involved, covering 350 factories in 30 countries.
Big companies such as OpenAI, Anthropic, Meta, and Microsoft have already shown interest in using Vera Rubin to grow their AI systems and lower the cost of processing each token.
Vera Rubin also brings new power-saving tools, including NVIDIA Spectrum-X Ethernet Photonics and the NVIDIA DSX system.
Spectrum-X Ethernet Photonics is a new switching technology based on CPO. Nvidia says it offers five times better power efficiency, five times higher AI uptime, and 1.3 times faster setup than networks using traditional transceivers.
DSX is made to help build and run AI factories with better energy use and lower token costs. It brings together reference designs, simulation, infrastructure software, facilities, and other technologies.
As AI begins to work through billions of digital agents on its own, Nvidia wants Vera Rubin to become one of the main platforms powering this next stage of technology. Production shipments of Vera Rubin will begin this fall.


















