CoreWeave becomes first hyperscaler to deploy Nvidia’s GB300 NVL72

CoreWeave (NASDAQ:CRWV) has become the first hyperscaler to deploy Nvidia’s (NASDAQ:NVDA) GB300 NVL72 liquid-cooled, rack-scale AI system, the company announced today.

CoreWeave shares surged 6% by noon trading on Thursday. Nvidia had edged up 1.4%. Dell Technologies (NYSE:DELL), another partner involved in the system, was up 1.7%.

The GB300 NVL72 features 72 Nvidia Blackwell Ultra GPUs and 36 Arm-based (ARM) Nvidia Grace CPUs. The unit is housed within Dell’s integrated rack scale system. The system is 10 times faster in user responsiveness and has a five-time improvement in throughput per watt over the prior generation of Nvidia’s Hopper.

It is built for frontier-scale AI and allows users to train, optimize and deploy multi-trillion parameter models, CoreWeave said. It’s also engineered to handle test-time scaling inference, which is crucial for deploying advanced AI models.

Last month, Supermicro (SMCI) unveiled a range of enterprise AI solutions built on Nvidia’s Blackwell architecture to accelerate AI factory deployments in the European market. The company introduced more than 30 solutions designed for air or liquid-cooled NVIDIA HGX B200, liquid-cooled NVIDIA GB200 NVL72 and NVIDIA RTX PRO 6000 Blackwell Server Edition.

Core Scientific (CORZ), a long-time partner of CoreWeave, inched down about 1%. Last month, reports surfaced that CoreWeave was considering acquiring the company, prompting a stock surge.

CoreWeave launched its initial public offering in late March 2025. Its stock value has quadrupled since its IPO date.