CoreWeave: Rent NVIDIA GPUs by the hour
CoreWeave is a cloud provider that rents NVIDIA GPU and CPU compute, storage and networking for AI training, inference and VFX rendering workloads.
What is CoreWeave?
CoreWeave is a cloud provider built around NVIDIA GPU capacity. Its product list covers GPU compute, CPU compute, storage, networking, managed services and bare metal servers, alongside platform tools such as the Fleet LifeCycle Controller, Node LifeCycle Controller, Tensorizer and observability. Workloads it names include AI model training, AI inference and VFX and rendering. The pricing page lists per-hour on-demand and spot rates for GPU instances across North America and Europe, from the single-GPU NVIDIA GH200 at $6.50 per hour up to eight-GPU configurations such as the NVIDIA HGX B200 at $68.80 per hour. Some models, including the NVIDIA GB300 NVL72 and the standard-memory RTX PRO 6000 Blackwell, are listed as contact sales. CPU instances are also priced per hour.
Focus areas: AI Training.
Who is CoreWeave for?
- AI teams training and serving large models
- Foundation model providers needing large GPU clusters
- VFX and rendering studios running GPU-heavy jobs
How much does CoreWeave cost?
Pricing model: paid. Always check the vendor’s own pricing page before you buy.
| Plan | Price | What you get |
|---|---|---|
| NVIDIA GH200 | $6.50/hr | 1 GPU, 96 GB VRAM, 72 vCPUs, 480 GB RAM |
| NVIDIA L40 | $10.00/hr | 8 GPUs, 48 GB VRAM each, 128 vCPUs, 1,024 GB RAM |
| NVIDIA L40S | $18.00/hr | 8 GPUs, 48 GB VRAM each, 128 vCPUs, 1,024 GB RAM |
| NVIDIA RTX PRO 6000 Blackwell Server Edition (High Memory) | $20.00/hr | 8 GPUs, 96 GB VRAM each, 128 vCPUs, 1,024 GB RAM |
| NVIDIA A100 | $21.60/hr | 8 GPUs, 80 GB VRAM each, 128 vCPUs, 2,048 GB RAM |
| NVIDIA GB200 NVL72 | $42.00/hr | 4 GPUs, 186 GB VRAM, 144 vCPUs, 960 GB RAM |
Is CoreWeave worth it?
A GPU-focused cloud with published hourly rates for many NVIDIA instance types, though several top-end models require a sales conversation
Where CoreWeave is strong
- Published per-hour on-demand and spot pricing for many NVIDIA GPU models
- Broad infrastructure stack beyond GPUs, including CPU, storage and networking
- Support model with direct access to cross-trained technical staff
Where it falls short
- Several GPU models, including GB300 NVL72, show contact sales instead of a listed price
- No free tier or trial is mentioned on the pricing page
How do you get started with CoreWeave?
- 01Review the GPU and CPU instance list on the pricing page for your region
- 02Compare on-demand and spot rates for the instance type you need
- 03Contact sales for models listed as contact sales, or sign in to provision an instance
CoreWeave: frequently asked questions
What kinds of workloads does CoreWeave target?
Its solutions cover AI model training, AI inference, and VFX and rendering, running on GPU and CPU compute with storage and networking services.
How is CoreWeave priced?
GPU and CPU instances are listed with per-hour on-demand and spot prices on the pricing page, split by North America and Europe regions. Some models are listed as contact sales.
What is the cheapest listed GPU instance?
The NVIDIA GH200 at $6.50 per hour on-demand is the lowest GPU rate shown on the pricing page, with 1 GPU, 96 GB VRAM, 72 vCPUs and 480 GB system RAM.
Does CoreWeave offer a free tier or trial?
The pricing page shows per-hour rates and contact sales options, and does not mention a free tier or free trial.
CoreWeave alternatives
The closest tools in our directory, matched on category, focus areas and audience.
| Tool | Best for | Free plan | Starting price | Focus areas |
|---|---|---|---|---|
| CoreWeave | Teams running large AI training, inference or rendering jobs | No | — | AI Training |
| Sesterce Cloud | European GPU cloud platform delivering high-performance, sovereign, and energy-efficient infrastructure for AI training, inference, and HPC. | Unknown | — | AI Training |
| Crusoe | Teams running large-scale AI training and inference on managed GPU cloud | No | — | — |
| Autoheal | Enterprise engineering teams automating delivery work | No | — | — |
| Traversal | Enterprise SRE and platform teams | No | — | — |
| Alfred AI | Teams needing one correlated view of API health, risk and usage | Yes | $233/mo | — |
| Pulumi | Platform teams automating provisioning, governance and cost checks | Yes | $40/mo | — |
| HeyCloud | DevOps and developer teams working primarily in AWS | Yes | — | — |
| Middleware | Engineering teams monitoring cloud-native apps and infrastructure | Yes | — | — |
| Cloudgeni | Teams bringing cloud resources under code with reviewable fixes | Unknown | — | — |
| SmartBird | Teams needing dedicated AI compute without managing hardware | Unknown | — | — |

