Comment on page
Data Center Regions
Learn about CoreWeave's data center regions
CoreWeave Cloud is served from three geographically diverse regions in the United States. Located directly adjacent to large metropolitan population centers, CoreWeave's data centers provide low latency access to accelerated compute to over 51 million people.
Our data centers are broken up into three geographical buckets, each of which service the regions for which they're named:
Each data center features redundant, 200Gbps+ public Internet connectivity from Tier 1 global carriers, and are connected to each other with 400Gbps+ of dark fiber transport, allowing for easy, free data transfers within CoreWeave Cloud.
To access and run the speedtests linked below in your terminal, either click the link to download them, or use
curlon the link's URL.
curl http://http.speedtest.lga1.coreweave.com/1G --output /dev/null
CoreWeave's data center regions are:
Located at the mouth of the Lincoln Tunnel,
LGA1serves the Eastern US, and provides ultra-low latency to the NYC metropolitan area, covering over 20 million people. This ISO 27001, SOC2 compliant data center is packed with NVIDIA GPU accelerated Cloud Instances, with near-unlimited compute and storage infrastructure to go with it.
Located just outside downtown Chicago,
ORD1serves the Central US. This ISO 27001, SOC2, HIPAA compliant data center is built for the most demanding workloads, including distributed training using NVIDIA A100 NVLINK accelerators connected with Infiniband GPUDirect RDMA.
Las Vegas, NV
Located in fabulous Las Vegas,
LAS1serves the Western US. Powered by 100% renewable energy, LAS1's ISO 27001, SOC2 and HIPAA compliant infrastructure brings scaled NVIDIA GPU accelerated compute to the US West, serving both Las Vegas and the Los Angeles basin with low latency connectivity.
To schedule a Virtual Server in a specific region, we've added an easy-to-use region selector on the CoreWeave Cloud Virtual Server UI when creating a new Virtual Server, which allows you to easily select the desired data center.
Data region selectors on CoreWeave Cloud UI
To schedule resources in a specific region using the Kubernetes API, simply add an affinity such as the one shown below to your deployment or other deployable resources:
- key: topology.kubernetes.io/region
The tables below describe the types of compute resources available in each data center.
At this time, the only CPU nodes capable of 100G bandwidth networking are the
intel-xeon-icelakenodes, and the only GPU nodes capable of 100G networking are A100 HGX NVLINK (
A100_NVLINK) and H100 HGX NVLINK (
PCIe nodes are not equipped for 100G networking.
The table below does not list current availability, only which types of resources are hosted in each region.
H100 HGX and A100 HGX distributed training nodes are only available via CoreWeave Bare Metal Kubernetes. These node types are not offered as Virtual Servers to eliminiate any virtualization overhead.
The following table provides the storage class names, correspondent to each storage type, per data center.
Each of our data center regions provides different levels of security compliance. The table below describes the compliance and certification levels for each data center region.