Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.coreweave.com/llms.txt

Use this file to discover all available pages before exploring further.

Capacity Finder is a tool in the Cloud Console that returns ranked Availability Zones for a Spot capacity request. Use it before you create a Spot Node Pool to find a Zone where your request is likely to fit. Spot Node Pools draw from unreserved CoreWeave capacity, so no guarantee exists that a given Zone has room for the instances you want.
  • You can compare placement availability across Zones for the same instance type.
  • You can shape your request by Node count, so a Zone that fits a small request but not a large one is scored accordingly.
  • You can launch a Spot Node Pool directly from a recommended Zone, pre-filled with the Zone and instance type, in a cluster you already operate there.

Limitations

  • Capacity Finder only evaluates General Access Zones. Single-tenant or reserved Zones aren’t included in results.
  • Capacity Finder does not consider Superchip-powered instance types (for example, GH200, GB200, GB300), which are not supported by Spot Node Pools.
  • Capacity Finder uses cached capacity samples that CoreWeave refreshes on a fixed cadence. Results reflect the most recent refresh, not live inventory.
  • Capacity Finder is informational. Results do not guarantee a reservation. Capacity can shift between the query and provisioning, and Spot Node Pools remain preemptible after they are created.

Use Capacity Finder in the Cloud Console

To run a search, provide an instance type and Node count. Capacity Finder returns ranked Zones based on the available capacity for that request.
  1. In the Cloud Console, open the Compute page and select the Capacity finder tab.
  2. Enter at least one row. Each row pairs an instance type with a Node count. You can fill in one row or both, but at least one row must have an instance type selected and a Node count greater than zero:
    • GPU instance type with Node count.
    • CPU instance type with Node count.
  3. Click Find capacity. Capacity Finder displays Zone cards with availability labels for your request.
The Compute page with the Capacity finder tab and Find capacity button highlighted. The form shows GPU and CPU instance type dropdowns and Node count fields. Capacity Finder returns Zone cards organized into one or two sections:
  • Recommended. Zones where you already operate at least one cluster and capacity is Likely available for your request.
  • Other capacity zones. Every other General Access Zone that returned a placement result for your request. If no Zones are recommended, this section is called All capacity zones instead. If no Zones at all return a placement result, you see a callout that reads “No zones match your request. Try a different instance type or Node count.”
Each Zone card shows:
  • The Zone name.
  • A GPU or CPU tag with the availability label for that part of your request (or both tags, if both rows were filled in).
  • The clusters you operate in that Zone, or Cluster setup required if you have none.
  • An action button: Provision capacity if you have a cluster in the Zone, or Create cluster otherwise.
Capacity Finder results with the Recommended and Other capacity zones sections highlighted. The Recommended section shows US-EAST-04A with GPU: Likely available and a Provision capacity button.

How Capacity Finder ranks results

Each search evaluates every General Access Zone for the instance type and Node count you enter. Capacity Finder queries placement scores for each Zone and ranks Zone cards by those scores, highest first. Within the results view:
  • Recommended Zones appear first, sorted by availability (highest placement score first among Zones where you already have a cluster and capacity is Likely available).
  • Other capacity zones (or All capacity zones when none are recommended) list every other Zone that returned a result for your request, also sorted by placement score.

Availability labels

Each availability tag on a Zone card carries one of five labels derived from a placement score. Use this table to interpret what each label means for your provisioning decision.
LabelMeaning
Likely availableComfortable or abundant capacity for the requested Node count. The request is most likely to fit.
Possibly availableCapacity is close to the requested Node count. Full or partial fulfillment are possible.
Limited availabilityCapacity is well below the requested Node count. The request is unlikely to fit at the requested size.
Not availableThe Zone has zero allocatable Nodes of the requested instance type.
SKU unavailableThis instance type is not available in this Zone. Check a different zone if you need this specific SKU.

Provision a Spot Node Pool from a result

The action on a Zone card depends on whether you already operate a cluster in that Zone:
  • You have a cluster in the Zone. Click Provision capacity. The Provision capacity wizard opens with the Zone, instance type, and Node count pre-filled from your search. Confirm or adjust the Spot Node Pool configuration, then submit the form to create the Spot Node Pool in the cluster.
  • You have no cluster in the Zone. Create a cluster in the Zone first, then provision the Spot Node Pool either from the Node pools tab or by returning to the Capacity finder tab.
A Node Pool created through Capacity Finder follows the standard Spot Node Pool lifecycle, including preemption behavior. After provisioning, the Node Pool appears in your cluster alongside any other Node Pools you have created. The Provision capacity wizard showing a two-step form. Step 1 has a cluster selector pre-filled with the sunk-latest cluster. Step 2 shows the Node Pool name, cluster, zone, instance type, and Node count fields, with the Provision capacity submit button highlighted.
Last modified on May 27, 2026