Newsletter Subscribe
Enter your email address below and subscribe to our newsletter
[forminator_form id="25163"]

coreweave+1news.bmsnvidianews.nvidia+1NVIDIA Corporation's Vera Rubin NVL72 rack-scale AI platform is moving rapidly toward broad commercial availability, with CoreWeave publishing the first measured silicon performance results and Bristol Myers Squibb Bristol-Myers Squibb Company announcing plans to deploy the architecture for drug discovery.
CoreWeave on Sunday published benchmark data showing that the Vera Rubin NVL72 generated 10 times the token throughput per megawatt compared to the previous-generation NVIDIA GB200 NVL72 when running DeepSeek's R1 reasoning model at matched interactivity targets. The test used state-of-the-art inference optimizations including expert parallelism, NVFP4 precision, multi-token prediction, and disaggregated prefill and decode. At full scale, a Vera Rubin NVL72 rack produced roughly 800,000 tokens per second at the same wattage where a Grace Blackwell system managed 80,000, according to reporting by WCCFTech.wccftech+1
On the same day, Bristol Myers Squibb announced it will deploy a second NVIDIA DGX SuperPOD built on eight DGX Vera Rubin NVL72 systems, which it described as the most powerful AI computing infrastructure in the life sciences industry. The company said the platform will help scientists evaluate larger chemical spaces, run more complex molecular predictions, and compress research timelines. According to NVIDIA's blog, BMS could become the first biopharmaceutical company globally to operate the Vera Rubin architecture.pulse2+2
CoreWeave completed the industry's first bring-up and validation of the Vera Rubin NVL72 at its Livingston, New Jersey facility on June 1, using Dell Technologies' PowerEdge XE9812 server platform. The rack features 72 Rubin GPUs and 36 Vera CPUs connected by a 260 TB/s NVLink 6 fabric.mlq+2
NVIDIA CEO Jensen Huang first confirmed the platform was in full production during his CES 2026 keynote in January, with the company later reiterating at Computex that production shipments would begin in fall 2026. Cloud partners including Google Cloud Alphabet Inc. , Microsoft Azure, and Oracle Cloud Infrastructure are expected to offer Vera Rubin instances in the second half of this year.hashrateindex+4
The platform integrates seven co-designed chips spanning compute, networking, storage, and security, and is designed for agentic AI workloads where reasoning models generate far more tokens per query than conventional inference.coreweave+1