CoreWeave

CoreWeave

AI teams, AI labs, and enterprises that need scalable GPU infrastructure for training, fine-tuning, and inference of large AI models.

Livingston, NJ, USA
coreweave.com
Headquarters
Livingston, NJ, USA
Type
Private company
Pricing
Usage-based
Visit Website

Overview

CoreWeave is an AI-native cloud provider that offers GPU infrastructure and related services for AI training, inference, and other compute-intensive workloads. The company positions itself as "The Essential Cloud for AI," providing a fully integrated platform of technology, tools, and teams to support complex AI workloads. CoreWeave's offerings include GPU compute (NVIDIA Vera Rubin, Blackwell, Hopper, Ada Lovelace), CPU compute, bare metal servers, networking, data storage, managed Kubernetes, runtime acceleration (serverless inference, dedicated inference), and model/agent development tools. The company also provides Mission Control for observability and fleet management, and CoreWeave ARENA for validating AI workloads before production. CoreWeave is headquartered in Livingston, NJ, USA. The company has raised significant funding, including a $1.1 billion Series C round led by Coatue in May 2024, reportedly valuing the company at $19 billion. CoreWeave has been recognized as a Visionary in the Gartner Magic Quadrant for Cloud AI Infrastructure and has earned SemiAnalysis' Platinum ClusterMAX rating. It serves a range of customers, including AI labs like Mistral AI, OpenAI, and IBM, as well as companies in media, entertainment, and other industries.

What it does

CoreWeave provides cloud infrastructure services for AI and high-performance computing. It offers on-demand and spot GPU instances with various NVIDIA GPU models (e.g., GB200, H100, H200) for training and inference workloads. The platform includes data storage solutions (AI object storage, distributed file storage), managed Kubernetes, and runtime acceleration tools like serverless inference and dedicated inference. CoreWeave also provides model development tools (training, fine-tuning, sandbox) and an agent development environment. The company emphasizes performance validation through MLPerf benchmarking and offers capacity plans for guaranteed capacity with flexible pricing.

Key features

  • GPU compute with NVIDIA Vera Rubin, Blackwell, Hopper, and Ada Lovelace GPUs
  • On-demand and spot GPU instances with per-hour pricing
  • AI object storage with hot, warm, and cold tiers
  • Managed Kubernetes for container orchestration
  • Serverless and dedicated inference for low-latency AI inference
  • Model training and fine-tuning tools
  • Mission Control for observability, security, and fleet lifecycle management
  • CoreWeave ARENA for validating AI workloads before production

Use cases

  • Large-scale AI model training
  • Reinforcement learning for AI agents
  • Low-latency AI inference for production applications
  • Agentic AI development and deployment
  • VFX rendering and media production
  • High-performance computing (HPC) workloads

Pricing

Usage-basedFrom $2.23

Pricing is per hour for GPU instances. Example: NVIDIA HGX H100/H200 starts at $2.23/hr. On-demand pricing for GB200 is $42.00/hr. Storage pricing: AI Object Storage Hot $0.06/GB/mo, Warm $0.03/GB/mo, Cold $0.015/GB/mo. Some high-end GPUs require contacting sales.

View current pricing

Pricing is gathered from the company's public pages and may change. Check the vendor's site before buying.

Pros and cons

Strengths

  • Purpose-built for AI with a focus on performance and scale
  • Recognized as a Visionary in Gartner Magic Quadrant for Cloud AI Infrastructure
  • Earned SemiAnalysis' Platinum ClusterMAX rating, the only AI cloud to do so twice
  • Strong performance in MLPerf benchmarking for training and inference
  • Flexible pricing with on-demand, spot, and capacity plans
  • Trusted by major AI labs like OpenAI, Mistral AI, and IBM

Limitations

  • Pricing for some high-end GPUs (e.g., GB300, B300) requires contacting sales
  • No free tier or trial mentioned on the website
  • Primarily focused on AI workloads, may not suit general-purpose cloud computing needs
  • Enterprise-focused, may not be ideal for small teams or individuals

What sets it apart

  • AI-native cloud platform purpose-built for AI workloads
  • Record-setting MLPerf benchmarking results
  • Platinum ClusterMAX rating from SemiAnalysis
  • Zero egress data migration with no fees and no lock-in
  • Capacity plans that can cut AI cloud TCO by up to 47%

Ecosystem

Notable customers

OpenAIMistral AIIBMPerplexityReelData

Frequently asked questions

What GPU options does CoreWeave offer?

CoreWeave offers NVIDIA GPUs including Vera Rubin, Blackwell (GB200, GB300, B300), Hopper (H100, H200), and Ada Lovelace. These are available for on-demand, spot, and capacity plan usage.

How is CoreWeave priced?

CoreWeave uses per-hour pricing for GPU instances. For example, NVIDIA HGX H100/H200 starts at $2.23/hr, and GB200 is $42.00/hr on-demand. Storage is priced per GB per month. Some high-end GPUs require contacting sales for pricing.

Does CoreWeave offer a free tier or trial?

No free tier or trial is mentioned on the website. CoreWeave offers on-demand, spot, and capacity plans, but no free usage.

What is CoreWeave ARENA?

CoreWeave ARENA is a service that lets you run real AI workloads on purpose-built infrastructure to validate performance, scaling, and cost before committing to production. It provides evidence-backed recommendations.

Who are CoreWeave's notable customers?

CoreWeave's customers include OpenAI, Mistral AI, IBM, Perplexity, and ReelData, among others.

Does CoreWeave support data migration?

Yes, CoreWeave offers Zero Egress Migration (0 EM) with no fees, no lock-in, and expert-led transfers to high-performance AI object storage.

Sources

This profile was compiled from the company's own pages and public web research.

Last researched August 11, 2026.

Listed on:claude

Manage this listing

Own this company? Update your listing information.

Update Listing
← Back to all companies