CoreWeave
AI-focused cloud that rents large NVIDIA GPU clusters for model training and inference.
Updated · AppBreakup editorial · Web
Is CoreWeave free?
No, it's a paid product
- HGX B200 on-demand price:
- $68.80 per hour (8 GPUs)
Checked on CoreWeave's site Nov 8, 2026
1 alternative with a free optionSwitch from CoreWeave to
All 7- Verified facts
- 5 of 5
- Alternatives ranked
- 7
- Approved reviews
- 0
- Visits sent from here
- 0
About CoreWeave
CoreWeave is a cloud provider built around NVIDIA GPUs for AI labs and enterprises. Rather than single hobbyist instances, it focuses on large clusters of the newest hardware (Hopper, Blackwell and GB200/GB300 NVL72 systems) linked by high-speed networking, managed through its own Kubernetes service and offered as bare metal without virtualisation overhead. Around the compute it sells AI object storage, fleet monitoring tools and, since acquiring Weights & Biases, an MLOps layer called Forge for experiment tracking, fine-tuning and serverless inference billed per token. It is used by foundation-model companies and large enterprises that need thousands of GPUs, usually on reserved contracts, though on-demand and spot instances are listed with public hourly prices. It is a poor fit for someone who just wants one cheap GPU for an afternoon. Access is through a web console, Kubernetes and APIs.
Best for: AI labs and enterprises that need large, reserved clusters of the newest NVIDIA GPUs.
CoreWeave pricing (checked Nov 8, 2026)
Each line is quoted from CoreWeave's own site, with the page it came from.
| Plan or term | What it says | Checked | Source |
|---|---|---|---|
| HGX B200 on-demand price | $68.80 per hour (8 GPUs) | Nov 8, 2026 | coreweave.com |
Pros and cons
What works
- Early access to the newest NVIDIA systems at cluster scale
- Managed Kubernetes and bare-metal options built for AI
- Public on-demand and spot hourly prices
- Experiment tracking and inference tools from Weights & Biases
Why people leave
- Instances are sold as 8-GPU nodes, so the entry cost is high
- Best capacity usually requires a sales conversation and a contract
- Overkill for hobbyists or single-GPU experiments
CoreWeave vs Lambda, RunPod and Vast.ai
From each app's listing and the prices verified on its own site. A dash means we haven't verified it yet.
Top CoreWeave alternatives
See all 7- 1
Lambda68.8Another AI-only GPU cloud with clusters, but also self-serve single-GPU instances.
- 2
RunPod64.8Pay-by-the-second pods and serverless endpoints for smaller teams without contracts.
- 3
Vast.ai61.8Marketplace pricing on a huge range of GPUs when cost matters more than scale.
- 4
Paperspace55.9On-demand GPU machines and notebooks from DigitalOcean, easier to start with than CoreWeave's Kubernetes-first setup.
- 5
Together AI52.6Rent dedicated GPU clusters or skip infrastructure and call hosted open models through its inference API.
Why people left CoreWeave
In reviewers' own words, from reviews of the apps they moved to. These are their opinions, not ours.
No switching stories yet. If you moved away from CoreWeave, review the app you switched to and tell us why.
Reviews
No approved reviews yet. How we verify reviewers
Write a review of CoreWeave
CoreWeave: common questions
What is CoreWeave?
CoreWeave: AI-focused cloud that rents large NVIDIA GPU clusters for model training and inference. Best for AI labs and enterprises that need large, reserved clusters of the newest NVIDIA GPUs. Available for Web.
Is CoreWeave free?
No, CoreWeave is a paid product. HGX B200 on-demand price: $68.80 per hour (8 GPUs) (checked on CoreWeave's own site).
How much does CoreWeave cost?
CoreWeave is a paid product. HGX B200 on-demand price: $68.80 per hour (8 GPUs). Checked on CoreWeave's own site, November 2026.
Is CoreWeave open source?
No, CoreWeave is not open source.
What platforms does CoreWeave run on?
CoreWeave is available for Web.
What are the best alternatives to CoreWeave?
The highest-ranked CoreWeave alternatives right now are Lambda, RunPod and Vast.ai. See the full ranked list for pricing, platforms and verified facts.
CoreWeave vs Lambda: what's the difference?
Both CoreWeave and Lambda are paid products. Verified prices: CoreWeave $68.80 per hour (8 GPUs), Lambda $0.69 per GPU/hour (Quadro RTX 6000), from 16 Nov 2026. Both run on Web.