What is CoreWeave used for?

CoreWeave is a GPU cloud infrastructure provider built for AI, offering NVIDIA GPU clusters, bare metal compute, networking and AI-tuned storage for training and inference. It is a good fit for training foundation models at cluster scale, serving models in production and academic research on GPU infrastructure.

How much does CoreWeave cost?

CoreWeave is a paid product with no free plan. Billed per instance-hour on demand, with lower spot rates. An 8-GPU HGX H100 node costs $49.24 per hour and an HGX B200 node $68.80; some newer systems such as B300 on-demand are quoted by sales.

Is CoreWeave open source?

No. CoreWeave is proprietary (closed-source) software and can't be self-hosted. Open-source alternatives to CoreWeave include beta9 and Backend.AI.

What are some alternatives to CoreWeave?

CoreWeave competes with Lambda, Crusoe and Nebius. For open-source options, see Enlisted's ranked list of open-source CoreWeave alternatives.

CoreWeave: features, pricing and alternatives

About CoreWeave

CoreWeave is a cloud platform purpose-built for AI workloads. Rather than offering a general-purpose cloud with GPUs added, it focuses on providing large GPU clusters, high-bandwidth networking and storage for training and serving models at scale.

Its infrastructure offerings include GPU compute across NVIDIA Hopper, Blackwell and Vera Rubin generations, CPU compute for supporting workloads, bare metal access without virtualization overhead, a high-bandwidth, low-latency networking fabric, AI object storage and dedicated storage clusters built on VAST. A zero egress migration program lets customers move data in without egress fees, and CoreWeave Forge is a newer product aimed at connecting an AI improvement loop.

CoreWeave is a proprietary commercial service with a pricing page. The company also runs programs for academic researchers and startups, operates a partner network, publishes a trust center and works with engineers directly through a direct-to-expert option.

Key features

Large-scale NVIDIA GPU clusters
Bare metal compute
High-bandwidth AI networking
AI object storage and dedicated storage
Zero egress migration program
Partner network and trust center

Good fit for

→Training foundation models at cluster scale
→Serving models in production
→Academic research on GPU infrastructure

Tags: gpu-cloud; ai-infrastructure; ai-training; ai-inference; nvidia; bare-metal; cloud; storage

CoreWeave: questions and answers

What is CoreWeave used for?: CoreWeave is a GPU cloud infrastructure provider built for AI, offering NVIDIA GPU clusters, bare metal compute, networking and AI-tuned storage for training and inference. It is a good fit for training foundation models at cluster scale, serving models in production and academic research on GPU infrastructure.
How much does CoreWeave cost?: CoreWeave is a paid product with no free plan. Billed per instance-hour on demand, with lower spot rates. An 8-GPU HGX H100 node costs $49.24 per hour and an HGX B200 node $68.80; some newer systems such as B300 on-demand are quoted by sales.
Is CoreWeave open source?: No. CoreWeave is proprietary (closed-source) software and can't be self-hosted. Open-source alternatives to CoreWeave include beta9 and Backend.AI.
What are some alternatives to CoreWeave?: CoreWeave competes with Lambda, Crusoe and Nebius. For open-source options, see Enlisted's ranked list of open-source CoreWeave alternatives.