Nvidia confirms Vera Rubin is shipping to OpenAI, CoreWeave, and other major customers

Nvidia ↘ 1.56% confirmed on Monday that its next-generation Vera Rubin AI platform has reached full production, with systems now shipping to customers including OpenAI, CoreWeave ↘ 1.86%, Google Cloud, Microsoft Azure ↘ 2.24%, Meta ↘ 3.36%, and Dell ↘ 0.56%, according to a briefing at the company's headquarters first reported by Bloomberg.thenextweb

Ian Buck, Nvidia's vice president of accelerated computing, said production shipments are underway, with OpenAI planning to deploy Vera Rubin systems at scale during the third quarter. CoreWeave, one of the first cloud providers to receive the hardware, reported that its NVL72 racks are delivering ten times the token output of the prior Grace Blackwell generation.coreweave+1

A Generational Leap in AI Hardware

The Rubin GPU at the heart of the platform features 336 billion transistors, 288 GB of HBM4 memory delivering 22 TB/s of bandwidth, and 50 sparse petaflops of NVFP4 inference performance — five times the inference capability of the previous Blackwell architecture. The Vera Rubin NVL72, a full-rack system pairing 72 Rubin GPUs with 36 Vera CPUs, uses liquid cooling to eliminate internal cabling, allowing components to be packed more tightly together.tech-insider+3

Nvidia also used the briefing to compare its custom Vera CPU against AMD's Advanced Micro Devices, Inc. ↘ 2.29% Turin chip, claiming nearly double the performance on Python workloads — a benchmark chosen because Python dominates the AI inference software stack. Anthropic, Perplexity, SpaceX, and Oracle ↘ 4.61% are among the early recipients alongside OpenAI and CoreWeave.thenextweb

Supply Chain and Market Context

The platform's supply chain spans more than 350 factories across 30 countries, with over 150 ecosystem partners in Taiwan alone. Dell was the first systems vendor to ship Vera Rubin NVL72 PowerRack units to CoreWeave, touting up to 10 times lower cost per token than Blackwell for large-scale agentic AI inference.investor.nvidia+3

The production milestone follows months of preparation. Jensen Huang first declared Vera Rubin in full production at Computex in early June, and Nvidia formally unveiled the seven-chip platform at GTC 2026 in March. Monday's briefing provided the first customer-reported performance figures, though independent verification of the claims remains forthcoming.datacenterknowledge+2