Woofun AI reports that NVIDIA has initiated customer testing for its next-generation Vera Rubin server system, distributing test racks to clients including CoreWeave, Microsoft, OpenAI, Anthropic, and SpaceXAI. Each Rubin rack houses 72 GPUs and carries a unit price of approximately $7 million to $8 million, exceeding the current $5 million price point for Grace Blackwell 300 racks.

NVIDIA Senior Vice President of Hardware Engineering Andrew Bell stated that more than ten manufacturing partners will eventually achieve a production capacity of up to 1,000 Rubin racks per day. The company is expanding its role from a GPU supplier to a full-stack AI server infrastructure provider, offering CPUs, network switches, cables, storage, and cooling technology alongside its chips.