NVIDIA's next-generation platform for agentic AI and reasoning at rack scale.
The newest NVIDIA generation. Rubin GPUs pair with Vera CPUs and sixth-generation NVLink, and are bought mainly as complete racks or eight-GPU systems rather than single cards.
Who it is for
Typical workloads
Previous: nvidia-blackwell-ultra
Next: —
An entire rack that behaves as a single enormous computer. It is bought by organisations training frontier models or running very large reasoning services, and needs liquid cooling and datacentre-grade power.
Memory
20736 GB
Deployment
Datacentre
From £3,250,000
ex VAT
Allocation only
A complete, supported AI computer in a single chassis. You get eight of the newest accelerators, the software stack and NVIDIA support without designing a server yourself.
Memory
2304 GB
Deployment
Datacentre
From £545,000
ex VAT
Allocation only
The GPU module that sits inside a purpose-built AI server. You buy it as part of a complete server build rather than fitting it to an existing machine.
Memory
2304 GB
Deployment
Rack server
From £465,000
ex VAT
Allocation only