CoreWeave is a Kubernetes-native GPU cloud, founded in 2017 and based in Livingston, New Jersey. It is built for teams running large training and inference jobs, not for grabbing a single consumer card for an afternoon. The catalogue we scrape covers datacenter hardware: H100 SXM, H100 NVL, H200 SXM, A100, L40, L40S, and the RTX 6000 Ada server edition.
Why CoreWeave stands out
The platform is recorded in our data as a full stack rather than a bare rental page. Kubernetes, a programmatic API, Docker, Jupyter, multi-GPU nodes, and persistent storage are all set. Spot and reserved capacity are both listed, and the provider is marked SOC 2 compliant. Payment methods on file are credit card and invoice.
That is the profile of a cloud you put under an existing MLOps setup. The live price table on this page is the thing to compare. Our billing-unit field for CoreWeave is still unknown, so confirm on their site whether a row is one GPU or a whole node before you treat it as an hourly rate.
Pros
- Datacenter GPUs in the scraped catalogue, including H100 SXM, H200 SXM, and A100
- Kubernetes, API, Docker, Jupyter, multi-GPU, and persistent storage all recorded
- Spot and reserved capacity both listed
- SOC 2 marked compliant, with credit card and invoice payments
- Founded in 2017, so this is not a new marketplace
Cons
- Sits at the premium end of this directory. It is the wrong place to hunt for the cheapest 4090
- Billing unit is not recorded in our data, so a row may not be a single GPU per hour
- Less interesting if you only need one consumer GPU and a web console
Getting started
- Open CoreWeave and confirm whether the price you care about is per GPU or per node.
- Create an account. Credit card and invoice are the payment methods we have on file.
- Pick a GPU from the table on this page and match the region and node size on their console.
- Deploy through Kubernetes or the API if that is how the rest of your stack already runs.
Best for: Teams that want a Kubernetes-native cloud with current NVIDIA datacenter GPUs and can pay premium rates for that platform, after checking the billing unit on the quote.