Skip to content
CoreWeave logo

CoreWeave

By CoreWeave

The essential cloud platform purpose-built for accelerating AI workloads with NVIDIA GPUs.

Paid plans only
Coverage from Wall Street Journal CNBC Bloomberg
2 reviews tracked 675 press mentions

The Bottom Line

Price
Paid plans only. Compare plans

Key facts

  • Specialized cloud infrastructure optimized for AI workloads.
  • Provides high-performance GPU compute, storage, and networking in a Kubernetes-native environment.
  • Features CoreWeave ARENA for pre-production AI workload validation and performance testing.

Pros

  • Accelerates AI development cycles and time-to-market
  • Achieves high cluster goodput (up to 96%) for efficient training and inference
  • Delivers real-time reliability with fewer interruptions (up to 50% fewer per day)
  • Provides cutting-edge bare-metal infrastructure
  • Offers 24/7 support from dedicated engineering teams

Cons

  • Specific pricing details are not publicly available and likely require direct consultation.
  • Primarily focused on AI workloads, which might not be suitable for general-purpose cloud computing needs.

What is CoreWeave?

CoreWeave provides a specialized cloud infrastructure designed for high-performance AI workloads. It offers GPU compute, flexible storage, and high-performance networking within a Kubernetes-native environment. The platform aims to accelerate AI development cycles, enabling faster inference spin-up times and quicker time-to-market for AI solutions. It is built on bleeding-edge bare-metal infrastructure with automated provisioning and supports leading workload orchestration frameworks. CoreWeave is ideal for AI labs, platforms, and enterprises that require robust, scalable, and reliable infrastructure for training and deploying complex AI models. It emphasizes maximizing 'goodput' and minimizing interruptions, ensuring high cluster utilization and real-time issue resolution. The platform includes managed software services, cluster health management, and a comprehensive suite of tools for observability, security, and machine learning, all backed by 24/7 engineering support. CoreWeave ARENA is a key component, serving as a production AI lab where teams can test and validate AI workloads at scale. This allows for assessing performance, scaling, and cost in a live-like environment before committing to full production, helping identify potential issues and optimize deployments.

Available on: Web

Ratings on G2

CoreWeave: 5.0 from 2 reviews on G2, last checked March 19, 2026.

Rating read on G2. Learn more

Preview

Key Features

  • Kubernetes-native GPU compute environment
  • Purpose-built storage solutions for AI
  • High-performance networking for cluster scale-out
  • Managed software services
  • Integrated cluster health management and performance monitoring
  • Robust platform with observability, security, and ML tools
  • Early access to NVIDIA GPUs
  • Automated provisioning

Pricing Plans

Pricing checked Nov 9, 2026. Source: coreweave.com/pricing, read Nov 9, 2026

CoreWeave plans and prices, checked November 2026
Plan Price Details
NVIDIA GB300 NVL72

Contact us

On-demand hourly price requires contacting sales. North America spot: N/A. Europe spot: N/A.

  • 4 GPUs
  • 279 GB VRAM
  • 144 vCPUs
+2 more
  • 960 GB system RAM
  • 61.44 TB local storage
NVIDIA GB200 NVL72

$42.00 / hour

On-demand price. Spot: N/A in North America and Europe. Inference single GPU: $10.50/hour.

  • 4 GPUs
  • 186 GB VRAM
  • 144 vCPUs
+2 more
  • 960 GB system RAM
  • 30.72 TB local storage
NVIDIA HGX B300

Contact us

On-demand hourly price requires contacting sales. Spot: $35.84/hour in North America, $36.70/hour in Europe.

  • 8 GPUs
  • 270 GB VRAM
  • 192 vCPUs
+2 more
  • 4,096 GB system RAM
  • 61.44 TB local storage
NVIDIA HGX B200

$68.80 / hour

On-demand price. Spot: $34.11/hour in North America, $34.87/hour in Europe. Inference single GPU: $8.60/hour.

  • 8 GPUs
  • 180 GB VRAM
  • 128 vCPUs
+2 more
  • 2,048 GB system RAM
  • 61.44 TB local storage
NVIDIA RTX PRO 6000 Blackwell Server Edition (High Memory)

$20.00 / hour

On-demand price. Spot: $11.09/hour in North America, $11.01/hour in Europe. Inference single GPU: $2.50/hour.

  • 8 GPUs
  • 96 GB VRAM
  • 128 vCPUs
+2 more
  • 1,024 GB system RAM
  • 7.68 TB local storage
NVIDIA RTX PRO 6000 Blackwell Server Edition (Standard Memory)

Contact us

On-demand hourly price requires contacting sales. Spot: $9.56/hour in North America and Europe.

  • 8 GPUs
  • 96 GB VRAM
  • 128 vCPUs
+2 more
  • 512 GB system RAM
  • 7.68 TB local storage
NVIDIA GH200

$6.50 / hour

On-demand price. Spot: N/A in North America and Europe. Inference single GPU: $6.50/hour.

  • 96 GB VRAM
  • 72 vCPUs
  • 480 GB system RAM
+1 more
  • 7.68 TB local storage
NVIDIA HGX H100

$49.24 / hour

On-demand price. Spot: $19.71/hour in North America, $19.51/hour in Europe. Inference single GPU: $6.16/hour.

  • 8 GPUs
  • 80 GB VRAM
  • 128 vCPUs
+2 more
  • 2,048 GB system RAM
  • 61.44 TB local storage
NVIDIA L40

$10.00 / hour

On-demand price. Spot: $6.27/hour in North America, N/A in Europe. Inference single GPU: $1.25/hour.

  • 8 GPUs
  • 48 GB VRAM
  • 128 vCPUs
+2 more
  • 1,024 GB system RAM
  • 7.68 TB local storage
NVIDIA HGX H200

$50.44 / hour

On-demand price. Spot: $20.93/hour in North America, $20.64/hour in Europe. Inference single GPU: $6.31/hour.

  • 8 GPUs
  • 141 GB VRAM
  • 128 vCPUs
+2 more
  • 2,048 GB system RAM
  • 61.44 TB local storage
NVIDIA L40S

$18.00 / hour

On-demand price. Spot: $7.88/hour in North America, N/A in Europe. Inference single GPU: $2.25/hour.

  • 8 GPUs
  • 48 GB VRAM
  • 128 vCPUs
+2 more
  • 1,024 GB system RAM
  • 7.68 TB local storage
NVIDIA A100

$21.60 / hour

On-demand price. Spot: $9.65/hour in North America, $9.51/hour in Europe. Inference single GPU: $2.70/hour.

  • 8 GPUs
  • 80 GB VRAM
  • 128 vCPUs
+2 more
  • 2,048 GB system RAM
  • 7.68 TB local storage
AMD Genoa (9274F), High Performance

$6.42 / hour

On-demand price. Spot: $3.50/hour in North America, $3.41/hour in Europe.

  • 96 vCPUs
  • 768 GB system RAM
  • 7.68 TB local storage
AMD Genoa (9454), General Purpose

$7.78 / hour

On-demand price. Spot: $3.32/hour in North America, $3.23/hour in Europe.

  • 192 vCPUs
  • 1,536 GB system RAM
  • 7.68 TB local storage
AMD Genoa (9454), General Purpose (High Storage)

$8.86 / hour

On-demand price. Spot: $4.36/hour in North America, $4.27/hour in Europe.

  • 192 vCPUs
  • 1,536 GB system RAM
  • 61.44 TB local storage
AMD Genoa (9654), High Core

$7.54 / hour

On-demand price. Spot: $4.23/hour in North America, $4.14/hour in Europe.

  • 384 vCPUs
  • 1,536 GB system RAM
  • 61.44 TB local storage
AMD Turin (9655P), General Purpose (High Memory)

$8.18 / hour

On-demand price. Spot: $4.62/hour in North America, $4.63/hour in Europe.

  • 192 vCPUs
  • 1,536 GB system RAM
  • 7.68 TB local storage
AMD Turin (9655P), General Purpose (Standard Memory)

Contact us

On-demand hourly price requires contacting sales. Spot: $3.94/hour in North America, $4.04/hour in Europe.

  • 192 vCPUs
  • 768 GB system RAM
  • 7.68 TB local storage
AMD Turin (9655P), General Purpose (High Storage)

$9.31 / hour

On-demand price. Spot: $5.25/hour in North America, $5.26/hour in Europe.

  • 192 vCPUs
  • 1,536 GB system RAM
  • 61.44 TB local storage
Intel Emerald Rapids (8562Y+), General Purpose (High Memory)

$5.31 / hour

On-demand price. Spot: $3.27/hour in North America, $3.31/hour in Europe.

  • 64 vCPUs
  • 512 GB system RAM
  • 7.68 TB local storage
Intel Emerald Rapids (8562Y+), General Purpose (Standard Memory)

Contact us

On-demand hourly price requires contacting sales. North America spot: $2.86/hour. Europe pricing is not shown in the provided page text.

  • 64 vCPUs
  • 256 GB system RAM
  • 7.68 TB local storage
Intel Ice Lake (6342), General Purpose

$3.36 / hour

On-demand price. North America spot: $2.29/hour. Europe pricing is not shown in the provided page text.

  • 96 vCPUs
  • 384 GB system RAM
  • 19.2 TB local storage

Is CoreWeave worth the price?

Good value

CoreWeave's GPU pricing is competitive for high-end AI workloads, with the NVIDIA HGX H200 at $50.44/hour and HGX H100 at $49.24/hour offering strong value versus major cloud providers.

However, several tiers like the GB300 NVL72 require contacting sales, which adds friction, and spot pricing is often unavailable or limited to specific regions. This platform is best for AI teams needing dedicated NVIDIA GPU capacity without long-term commitments.

Hidden Costs & Gotchas

GB300 NVL72 and HGX B300 require contacting sales for pricing

Europe spot pricing not shown for Intel Ice Lake and Emerald Rapids standard memory

Inference single GPU pricing only available for select tiers

Full CoreWeave pricing: every plan, hidden costs and alternatives

Email me when CoreWeave changes its price

We email you when CoreWeave changes its pricing. We read its pricing page every week. One email per change, unsubscribe any time.

Have an account? Sign in and it is on in one click.

Reviews

Improve Your Thinking Patterns Using ChatGPT cover
$99 Free with your review

Write the first CoreWeave review, get a free AI guide

Share your experience and we will send you Improve Your Thinking Patterns Using ChatGPT, free.

Write the first review

Best CoreWeave Alternatives

Top alternatives based on features, pricing, and user needs.

CoreWeave by the numbers

CoreWeave compared with 44 GPU Cloud tools on Toolradar, as of Nov 2026
Metric CoreWeave Peer median Position (1 is best)
Free plan No 17 of 44 have one n/a
Listed as an alternative to 14 tools 1 #7 of 44 (joint)
Compare-page visits, last 30 days 43 visits 43 #9 of 17 tools with 20+ compare-page visits

Listed as an alternative to: #8 on 2026-10-16 to #7 on 2026-11-11 (rank among 44 tools, 1 is best).

Each position is counted among the tools that have data for that row, as noted under it. All GPU Cloud data · How we compute this. Advertising and placement fees are not inputs to any position. Visit counts include visits from every source.

Cite this: Toolradar, “CoreWeave by the numbers”, Nov 2026.

Most buyers shortlist 2 or 3 tools before committing. Pull a side-by-side comparison or browse the full alternatives shortlist below.

Explore More

CoreWeave FAQ

How does CoreWeave ARENA help validate AI workloads before production deployment?

CoreWeave ARENA functions as a production-scale AI lab that allows teams to test real workloads under conditions that mirror live production environments. This enables assessment of performance, scaling behavior, and cost implications, helping to identify potential issues and optimize models and pipelines before they are committed to full production.

What kind of infrastructure does CoreWeave utilize to achieve high performance for AI?

CoreWeave leverages bleeding-edge bare-metal infrastructure, providing direct access to NVIDIA GPUs within a Kubernetes-native environment. This is complemented by purpose-built storage solutions and high-performance networking designed for optimal cluster scale-out and connectivity, ensuring maximum reliability and efficiency for demanding AI tasks.

How does CoreWeave ensure high cluster utilization and minimize interruptions for AI training and inference?

CoreWeave achieves up to 96% cluster goodput and significantly fewer interruptions through resilient infrastructure, rigorous node lifecycle management, and deep observability. This is further supported by an integrated suite of services for cluster health management and 24/7 support from dedicated engineering teams, allowing for near real-time issue resolution.

Can CoreWeave accommodate various AI workload orchestration frameworks?

Yes, CoreWeave's Kubernetes-native developer experience is designed to support leading workload orchestration frameworks. This flexibility allows users to integrate their existing tools and processes for managing and deploying AI models effectively.

What specific benefits did companies like OpenAI, Mistral AI, and IBM experience by using CoreWeave?

OpenAI found CoreWeave delivered reliable performance and expertise for rapidly bringing game-changing AI models to the world. Mistral AI achieved 2.5x faster training speeds for their open-source reasoning models, cutting training time in half. IBM accelerated their Granite™ models by up to 80%, benefiting from CoreWeave's scale, speed, and reliability for their AI infrastructure needs.

Beyond compute, what other services does CoreWeave provide to manage AI infrastructure?

In addition to GPU compute, CoreWeave offers flexible, purpose-built storage and high-performance networking. It also includes managed software services, integrated cluster health management, performance monitoring, and a robust platform that encompasses observability, security, and various ML tools to lift the management burden from users.

Guides & Articles