CoreWeave launches NVIDIA GB200 NVL72-based instances for AI reasoning models

Investing.com -- CoreWeave, a cloud service provider, has launched NVIDIA (NASDAQ: NASDAQ:NVDA) GB200 NVL72-based instances, marking the first general availability of the NVIDIA Blackwell platform. The launch aims to meet the increasing demand for massive compute and optimized software to scale AI reasoning models and agents.

The NVIDIA GB200 NVL72 is a liquid-cooled, rack-scale solution with a 72-GPU NVLink domain. This configuration allows the 72 GPUs to function as a single, large GPU. The NVIDIA Blackwell platform features numerous technological breakthroughs that enhance inference token generation, thereby improving performance while lowering service costs. For instance, the fifth-generation NVLink offers 130TB/s of GPU bandwidth in a single 72-GPU NVLink domain, while the second-generation Transformer Engine enables FP4 for quicker AI performance while maintaining high accuracy.

The new instances are integrated with rack-scale NVIDIA NVLink across 72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUs. They can scale up to 110,000 GPUs with NVIDIA Quantum-2 InfiniBand networking, providing the necessary scale and performance to build and deploy the next generation of AI reasoning models and agents.

CoreWeave’s GB200 NVL72 instances also feature NVIDIA Quantum-2 InfiniBand networking, delivering 400Gb/s bandwidth per GPU for clusters up to 110,000 GPUs. NVIDIA BlueField-3 DPUs provide accelerated multi-tenant cloud networking, high-performance data access, and GPU compute elasticity for these instances.

CoreWeave’s portfolio of managed cloud services is tailored for Blackwell. The CoreWeave Kubernetes Service optimizes workload orchestration by revealing NVLink domain IDs, ensuring efficient scheduling within the same rack. Additionally, CoreWeave’s Observability Platform offers real-time insights into NVLink performance, GPU utilization, and temperatures.

NVIDIA’s full-stack AI platform, which pairs advanced software with Blackwell-powered infrastructure, assists enterprises in building fast, accurate, and scalable AI agents. This platform includes NVIDIA Blueprints, NIM, and NeMo, tools that help developers create, deploy, and fine-tune AI models for modern enterprise use cases.

The general availability of NVIDIA GB200 NVL72-based instances on CoreWeave is the latest outcome of the companies’ collaboration, aimed at delivering cutting-edge accelerated computing solutions to the cloud. With these instances, enterprises can now access the scale and performance needed to power the next wave of AI reasoning models and agents.

Customers can begin provisioning GB200 NVL72-based instances through the CoreWeave Kubernetes Service in the US-WEST-01 region using the gb200-4x instance ID.

This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.