Luisa Crawford
Feb 05, 2025 01:57
CoreWeave has launched NVIDIA GB200 NVL72-based instances, marking the first general availability of NVIDIA Blackwell in the cloud, offering unprecedented AI performance and scalability.
In a significant advancement for cloud computing and artificial intelligence, CoreWeave has announced the general availability of NVIDIA GB200 NVL72-based instances, making it the first cloud service provider to offer the NVIDIA Blackwell platform to a broad audience. This launch is expected to facilitate the deployment of advanced AI models at unprecedented scales, according to a recent announcement by CoreWeave.
NVIDIA GB200 NVL72 on CoreWeave
The NVIDIA GB200 NVL72 platform is a cutting-edge, liquid-cooled, rack-scale solution featuring a 72-GPU NVLink domain. This configuration allows the GPUs to function as a single, massive unit, significantly enhancing computational power. The platform integrates several technological breakthroughs, such as the fifth-generation NVLink, which provides 130TB/s of GPU bandwidth, and the second-generation Transformer Engine, which supports FP4 for faster AI performance without sacrificing accuracy.
CoreWeave’s cloud services are specifically optimized for Blackwell, with features like the CoreWeave Kubernetes Service, which ensures efficient workload scheduling, and Slurm on Kubernetes (SUNK) for intelligent workload distribution. Additionally, CoreWeave’s Observability Platform offers real-time insights into NVLink performance and GPU utilization.
Full-Stack Accelerated Computing for AI
NVIDIA’s full-stack AI platform combines advanced software with Blackwell-powered infrastructure, providing enterprises with the tools needed to develop scalable AI models and agents. Key components include NVIDIA Blueprints for customizable workflows, NVIDIA NIM for secure AI model deployment, and NVIDIA NeMo for model training and customization. These elements are part of the NVIDIA AI Enterprise software platform, enabling scalable AI solutions on CoreWeave’s infrastructure.
Next-Generation AI in the Cloud
The introduction of NVIDIA GB200 NVL72-based instances on CoreWeave underscores the companies’ commitment to delivering cutting-edge computing solutions. This collaboration provides enterprises with the necessary resources to drive the next generation of AI reasoning models. Enterprises can now access these high-performance instances via CoreWeave Kubernetes Service in the US-WEST-01 region.
For those interested in leveraging these powerful cloud-based solutions, further information and provisioning options can be explored through CoreWeave’s official channels.
For more detailed information, please refer to the official NVIDIA blog.
Image source: Shutterstock
