Nvidia's latest AI supercomputers, powered by the Grace Blackwell Ultra Superchip, have been deployed for the first time at CoreWeave, a leading AI cloud service provider. These supercomputers are built on Dell's Integrated Racks equipped with 72 Nvidia Blackwell Ultra GPUs, 36 Arm-based 72-core Grace CPUs, and 36 BlueField DPUs per rack.
Each rack delivers 1.1 ExaFLOPS of dense FP4 inference and 0.36 ExaFLOPS of FP8 training performance, with 20 TB of HBM3E and 40 TB of total RAM. Due to the extreme power consumption of up to 1,400W per GPU, the systems are liquid-cooled.
The rack-scale systems come assembled and tested, designed for fast installation and deployment for production use. They are densely configured to maximize training and inference performance. This deployment is a result of close work between Dell and CoreWeave, combining hardware, software, and services into integrated offerings that can be rapidly deployed.
The partnership between Nvidia, Dell, and CoreWeave highlights the growing demand for high-performance AI computing solutions. As AI continues to drive innovation and growth, the need for powerful and efficient computing infrastructure will only continue to increase. With this deployment, CoreWeave is well-positioned to support the development of advanced AI applications and services.
