We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Paid plans start at $3.27/mo.
CoreWeave Cloud has no listed free plan or free trial. The pricing page lists on-demand and spot compute rates per instance-hour, with prices varying by hardware, region and availability. Spot starts at $3.27 per hour for Intel Emerald Rapids (8562Y+), General Purpose (High Memory); some on-demand rates require contacting sales. Inference GPU rates are listed separately for North America. Budget by workload runtime and confirm storage, networking and other charges with CoreWeave before deployment.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| NVIDIA GB300 NVL72 | — | Quote | Large-scale AI training |
| NVIDIA GB200 NVL72 | — | Quote | Large-scale AI training |
| NVIDIA HGX B300 | — | Quote | Large-scale AI training |
| NVIDIA HGX B200 | — | Quote | Large-scale AI training |
| NVIDIA RTX PRO 6000 Blackwell Server Edition (High Memory) | — | Quote | GPU workloads needing high memory |
| NVIDIA RTX PRO 6000 Blackwell Server Edition (Standard Memory) | — | Quote | GPU workloads needing standard memory |
| NVIDIA GH200 | — | Quote | Single-GPU workloads |
| NVIDIA HGX H100 | — | Quote | AI training and inference |
| NVIDIA L40 | — | Quote | GPU rendering and inference |
| NVIDIA HGX H200 | — | Quote | AI training and inference |
| NVIDIA L40S | — | Quote | GPU rendering and inference |
| NVIDIA A100 | — | Quote | AI training and inference |
| AMD Genoa (9274F), High Performance | — | Quote | High-performance CPU workloads |
| AMD Genoa (9454), General Purpose | — | Quote | General-purpose CPU workloads |
| AMD Genoa (9454), General Purpose (High Storage) | — | Quote | CPU workloads needing more storage |
| AMD Genoa (9654), High Core | — | Quote | High-core CPU workloads |
| AMD Turin (9655P), General Purpose (High Memory) | — | Quote | CPU workloads needing high memory |
| AMD Turin (9655P), General Purpose (Standard Memory) | — | Quote | General-purpose CPU workloads |
| AMD Turin (9655P), General Purpose (High Storage) | — | Quote | CPU workloads needing more storage |
| Intel Emerald Rapids (8562Y+), General Purpose (High Memory) | — | Quote | CPU workloads needing high memory |
| Intel Emerald Rapids (8562Y+), General Purpose (Standard Memory) | — | Quote | General-purpose CPU workloads |
Add-ons and overages
- Inference GPU rates are listed for North America: GB200 $10.50/hr; HGX B200 $8.60/hr; RTX PRO 6000 High Memory $2.50/hr; GH200 $6.50/hr; H100 $6.16/hr; L40 $1.25/hr; H200 $6.31/hr; L40S $2.25/hr; A100 $2.70/hr. Contact sales for other listed inference rates.
02What the plan grid doesn't tell you
Storage and networking
Compute rates include listed local storage, but confirm the price of object storage, dedicated VAST clusters, networking and any related services.
Spot availability
Spot rates are lower for many configurations, but some have no spot rate listed. Confirm availability and interruption behavior before relying on spot capacity.
Sales-quoted instances
Some on-demand rates require contacting sales. Request a written estimate for the configuration, region and runtime you expect to use.
Inference pricing
North America inference GPU rates are listed separately from instance rates. Contact sales for other listed inference rates and confirm which service applies.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Single-GPU workloads | NVIDIA GH200 | Lists one GPU with 96 GB VRAM and a $6.50 on-demand rate per instance-hour. |
| AI training and inference | NVIDIA HGX H100 | Lists eight GPUs with 80 GB VRAM; on-demand is $49.24 per instance-hour, with spot rates also listed. |
| CPU workloads on a budget | Intel Emerald Rapids (8562Y+), General Purpose (High Memory) | The listed spot rate starts at $3.27 per instance-hour; on-demand is $5.31. |
04Is CoreWeave Cloud worth the money?
CoreWeave Cloud is worth pricing when a workload needs its listed GPU configurations, cluster infrastructure or managed AI services. Published hourly prices let teams compare specific compute options, but the absence of a free tier makes experimentation a paid activity, and some on-demand rates require sales contact. Spot pricing may reduce the compute bill, though teams should confirm availability and interruption behavior. Self-hosting is not an alternative for the platform itself. Get a complete estimate for compute, storage, networking and inference before committing.
CoreWeave Cloud pricing FAQ
What is the cheapest listed CoreWeave Cloud rate?
The lowest listed rate is $3.27 per instance-hour for spot NVIDIA Emerald Rapids (8562Y+), General Purpose (High Memory). Its on-demand rate is $5.31 per instance-hour. Spot prices and availability can differ from on-demand, so confirm the rate and capacity for your region.
Does CoreWeave Cloud charge monthly or annually?
The pricing page lists hourly on-demand and spot rates per instance and does not list monthly or annual billing. Estimate costs from expected instance-hours, then ask CoreWeave whether your intended workload has other billing terms or service charges.
How much does CoreWeave Cloud inference cost?
North America inference GPU rates include H100 at $6.16 per hour, A100 at $2.70, GH200 at $6.50 and L40 at $1.25. Other listed inference rates may require contacting sales. Confirm the applicable service, GPU and region before estimating production costs.
Are CoreWeave Cloud spot instances always available?
The pricing page shows spot rates for many configurations, but not all: some are listed as unavailable or have no spot price shown. Ask CoreWeave about current capacity and interruption behavior for your hardware and region before making spot instances a production dependency.