Reserve capacity through your account team

This document explains how to obtain capacity by using future reservations in AI Hypercomputer, which are available for clustered GPUs. To learn about all the capacity options in AI Hypercomputer, see Capacity overview.

For a very high assurance that your workloads have the resources they need, request a future reservation from Google. You can reserve blocks of compute capacity and, optionally, a Hyperdisk pool in a single request for a defined duration, starting on a specific date and time that you choose. Based on your request, Google creates a draft future reservation request. After you review and submit this draft request, and Google Cloud approves it, Compute Engine creates an empty reservation and prepares your Hyperdisk pool. Then, at your chosen start time, Compute Engine provisions your requested capacity into the auto-created reservation and Hyperdisk pool all at once. You can then use the reserved resources to create Compute Engine instances until the reservation period ends.

Limitations

This section describes the limitations for future reservation requests, and for the auto-created reservations and Hyperdisk pools for a request.

Limitations for future reservation requests

After Google creates a draft future reservation request for you, the following limitations apply:

  • Pre-existing pools:
    • Hyperdisk Storage Pools: you can't reserve capacity for a pre-existing pool. Google automatically creates a new pool when it delivers your reservation.
    • Hyperdisk Exapools: you can increase the capacity of an existing Exapool by submitting subsequent future reservation requests. To add capacity to an existing Exapool, contact your account team.
  • Modifications: after you submit the request for review, and Google Cloud approves it, you can't modify the request details.
  • Cancellations: after the request is submitted, approved, and its state changes to PROVISIONING, you can no longer cancel or delete the request. You commit to pay for the requested capacity from the request's start time, regardless of usage.

Limitations for auto-created reservations and Hyperdisk pools

After Compute Engine creates a reservation and, if requested, a Hyperdisk pool to fulfill your reservation request, the following limitations apply:

Limitations for attached commitments

If your auto-created reservation includes an attached resource-based commitment, then the following limitations apply:

  • You can't renew, extend, merge, split, or upgrade that commitment.
  • If you use your reserved resources before the commitment becomes active, then you pay on-demand rates. To receive committed use discounts (CUDs), wait until the commitment becomes active before you use your reservation.

Before you begin

  • If you haven't already, then review the available VM and cluster configurations.

  • Select the tab for how you plan to use the samples on this page:

    Console

    When you use the Google Cloud console to access Google Cloud services and APIs, you don't need to set up authentication.

    gcloud

    In the Google Cloud console, activate Cloud Shell.

    Activate Cloud Shell

    At the bottom of the Google Cloud console, a Cloud Shell session starts and displays a command-line prompt. Cloud Shell is a shell environment with the Google Cloud CLI already installed and with values already set for your current project. It can take a few seconds for the session to initialize.

    REST

    To use the REST API samples on this page in a local development environment, you use the credentials you provide to the gcloud CLI.

      Install the Google Cloud CLI.

      If you're using an external identity provider (IdP), you must first sign in to the gcloud CLI with your federated identity.

    For more information, see Authenticate for using REST in the Google Cloud authentication documentation.

Required roles

To get the permissions that you need to create a future reservation request, ask your administrator to grant you the Compute Future Reservation User (roles/compute.futureReservationUser) IAM role on the project. For more information about granting roles, see Manage access to projects, folders, and organizations.

This predefined role contains the permissions required to create a future reservation request. To see the exact permissions that are required, expand the Required permissions section:

Required permissions

The following permissions are required to create a future reservation request:

  • To allow Compute Engine to auto-create reservations: compute.reservations.create on the project
  • To create a future reservation request: compute.futureReservations.create on the project
  • To specify an instance template: compute.instanceTemplates.useReadOnly on the instance template

You might also be able to get these permissions with custom roles or other predefined roles.

Quota

As part of the future reservation request process, Google manages quota for your reserved resources. You don't need to request quota. At the start time of your approved future reservation, Google automatically increases your quota if your project doesn't have enough for the reserved resources.

Overview

To reserve blocks of capacity, complete the following steps:

  1. Request capacity through your account team. Contact your account team to specify the type and number of resources that you want to reserve.
  2. Review and submit a draft reservation request. After Google creates a draft future reservation request, review it. If it looks correct, then submit the request for review. Google Cloud immediately approves it.

Request capacity through your account team

Contact your account team and provide the following information for Google to create a draft future reservation request:

  • Project number: the number of the project where your account team creates the request and Compute Engine provisions the capacity.
  • Machine type: the machine type to reserve. You can specify one of the following:
    • A4X Max (a4x-maxgpu-4g-metal)
    • A4X (a4x-highgpu-4g)
    • A4 (a4-highgpu-8g)
    • A3 Ultra (a3-ultragpu-8g)
    • A3 Mega (a3-megagpu-8g)
    • A3 High (a3-highgpu-8g)
  • Zone: the zone where you want to reserve capacity. To review the available regions and zones for a GPU machine type, see Accelerator availability.
  • Deployment type: specify a dense deployment type. This configuration allocates your accelerators physically close to each other to minimize network hops and optimize for the lowest latency. Google recommends this configuration for clustered GPU workloads to reduce network latency.
  • Total count: the total number of compute instances to reserve. You can reserve only multiples of two compute instances. Block sizes and compute instance count per block vary based on machine type and availability. Your account team can provide more details for your request.
  • Start time: the start time of the reservation period. You can use the reserved capacity at that time. If you request a Hyperdisk pool, then the start time must be on or before the start time for the compute instances. Format a start time as an RFC 3339 timestamp: 
    YYYY-MM-DDTHH:MM:SSOFFSET

    Replace the following:

    • YYYY-MM-DD: a date formatted as a four-digit year, two-digit month, and a two-digit day of the month, separated by hyphens (-).
    • HH:MM:SS: a time formatted as a two-digit hour by using a 24-hour time, two-digit minutes, and two-digit seconds, separated by colons (:).
    • OFFSET: the time zone formatted as an offset of Coordinated Universal Time (UTC). For example, to use the Pacific Standard Time (PST), specify -08:00. To use no offset, specify Z.
  • End time: the end time of the reservation period. The end time of the reservation for compute capacity and Hyperdisk pool capacity must match. You must format an end time as an RFC 3339 timestamp. At that time, Compute Engine does the following:
    • Deletes the auto-created reservation and any associated empty Hyperdisk pools.
    • Based on the termination action that you specify for your compute instances, Compute Engine stops or deletes any compute instances that you created by using the auto-created reservation.
  • Reservation name: the name of the reservation that Compute Engine creates to deliver your reserved capacity. Compute Engine can create only specifically targeted reservations.
  • Reservation automatic deletion: whether you want Compute Engine to automatically delete the auto-created reservation and any associated Hyperdisk pool at the end of the reservation period. If you want to manually delete the reservation, then you must contact your account team to delete the reservation.
  • Maintenance scheduling type: how Compute Engine schedules host maintenance for the compute instances in your cluster. Specify one of the following options:
    • To synchronize maintenance across all A4X Max, A4X, A4, or A3 Ultra instances in your reserved blocks, use GROUPED.
    • To use different maintenance schedules for your A4 or A3 Ultra instances, use INDEPENDENT.
  • Reservation operational mode: whether you or Google Cloud manages the compute instance recovery process after a host error or faulty host report. The mode that you use depends on the machine type that you reserve:
    • All capacity mode (ALL_CAPACITY). This mode is for A4X Max and A4X machine types only. You must manually manage the maintenance and recovery process for a compute instance. However, you have full access to all machines in your reservation, as well as more granular control over your compute instance topology and maintenance.
    • Managed mode (HIGHLY_AVAILABLE_CAPACITY). This mode is for A4, A3 Ultra, A3 Mega, and A3 High machine types only. You let Google Cloud manage the maintenance and recovery process for a compute instance by automatically replacing faulty machines. This approach increases the availability of your compute instances and helps you quickly restart them on new hosts. However, the topology of your compute instances might change after Compute Engine replaces a machine.
  • Share type: whether only your project can use the auto-created reservation (LOCAL), or other projects can use the reservation (SPECIFIC_PROJECTS). To share reserved capacity with other projects in your organization, do the following:
    1. If you haven't already, then verify that the project where Google creates the request is allowed to create shared reservations.
    2. Provide the numbers of the projects to share the reserved capacity with. You can specify up to 100 projects in your organization.

  • Commitment name: if your reservation period is one year or longer, then Google Cloud creates a resource-based commitment for those resources and attaches the commitment to your reservation. If your reservation period is less than 3 years, then you receive committed use discounts (CUDs) that apply to the 1-year plan for that commitment. If your reservation period is 3 years or more, then you receive CUDs that apply to the 3-year plan.

    If you share the reserved capacity with other projects, then those projects get CUDs only if they use the same Cloud Billing account as the project where you reserve capacity. For details, see Share resource-based CUDs across projects.

    The commitment has the same start and end dates as your reservation, but the activation times might differ. If you specify the reservation start time as 12:00 AM US and Canadian Pacific Time, then the commitment becomes active at 12:00 AM on your chosen start date. If you specify a different start time, then the commitment becomes active at 12:00 AM on the following day of your chosen start date.

  • Hyperdisk pool: whether you want to reserve storage capacity along with your compute instances as a Hyperdisk pool. If you submitted a reservation request for compute capacity without storage, then you can submit a subsequent request. To submit a request for a Hyperdisk pool, provide the following details:

    • Hyperdisk pool name: a name for the Hyperdisk pool. The name must be unique within your project.

    • Hyperdisk pool type: choose whether you want to create a Storage Pool or an Exapool, and whether you want to create only Hyperdisk Balanced or Hyperdisk Throughput volumes in your pool. For guidance, see How to choose a Hyperdisk pool type and Choose a Hyperdisk type. After you make your choices, specify one of the following values:

      • hyperdisk-balanced: creates a Hyperdisk Balanced Storage Pool. This pool supports boot disks and non-boot disks. You must also specify the provisioned capacity in GiB, provisioned IOPS, and provisioned throughput in MiB/s.
      • hyperdisk-throughput: creates a Hyperdisk Throughput Storage Pool. This pool supports non-boot disks only. You must also specify the provisioned capacity in GiB and provisioned throughput in MiB/s.
      • hdb-exapool: creates a Hyperdisk Balanced Exapool for workloads that require 500 TiB or more of storage. This pool supports boot disks and non-boot disks. You must specify the provisioned capacity in GiB and your workload optimization type (read-optimized, write-optimized, or capacity-optimized).
      • hdt-exapool: creates a Hyperdisk Throughput Exapool for workloads that require 500 TiB or more of storage. This pool supports non-boot disks only. You must specify the provisioned capacity in GiB and your workload optimization type (read-optimized, write-optimized, or capacity-optimized).

After Google creates the draft future reservation request, your account team contacts you.

Review and submit a draft reservation request

After you give the resource details to your account team, Google creates a draft future reservation request. You can review the draft request and, if correct, submit it for review. You must submit the request before the request start time.

To review and submit a draft future reservation request, select one of the following options:

Console

  1. In the Google Cloud console, go to the Reservations page.

    Go to Reservations

  2. Click the Future reservations tab. The Future Reservations table lists each future reservation request in your project, and each table column describes a property.
  3. In the Name column, click the name of the draft request that Google created for you. A page that gives the details of the future reservation request opens.
  4. In the Basic information and resource details sections, verify that the request details, such as Dates and Share type, are correct. Also, if you requested a commitment, verify that it's specified. If any of these details are incorrect, then contact your account team.

  5. If everything looks accurate, then submit your request:

    1. Click Edit draft. A page to modify the draft request appears.
    2. Click Create. The Reservation page appears. Within a few minutes, Google Cloud approves your request, and then Compute Engine creates an empty reservation. If you requested storage resources, then Compute Engine creates a new Storage Pool or Exapool, or increases capacity for an existing Exapool.

gcloud

  1. To view a list of future reservation requests in your project, use the gcloud compute future-reservations list command with the --filter flag set to PROCUREMENT_STATUS=DRAFTING: 
    gcloud compute future-reservations list --filter=PROCUREMENT_STATUS=DRAFTING
  2. In the command output, look for the reservation request that has the name that you provided to your account team.
  3. To view the details of the draft request, use the gcloud compute future-reservations describe command: 
    gcloud compute future-reservations describe FUTURE_RESERVATION_NAME \
    --zone=ZONE

    Replace the following:

    • FUTURE_RESERVATION_NAME: the name of the draft future reservation request.
    • ZONE: the zone where Google created the request.

    The output is similar to the following:

    autoCreatedReservationsDeleteTime: '2026-02-10T19:20:00Z'
creationTimestamp: '2025-11-27T11:14:58.305-08:00'
deploymentType: DENSE
id: '7979651787097007552'
kind: compute#futureReservation
name: example-draft-request
planningStatus: DRAFT
reservationName: example-reservation
schedulingType: INDEPENDENT
selfLink: https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b/futureReservations/example-draft-request
selfLinkWithId: https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b/futureReservations/7979651787097007552
specificReservationRequired: true
specificSkuProperties:
  instanceProperties:
    guestAccelerators:
    -   acceleratorCount: 8
      acceleratorType: nvidia-h200-141gb
    localSsds:
    -   diskSizeGb: '375'
      interface: NVME
    ...
  machineType: a3-ultragpu-8g
totalCount: '2'
storagePoolProperties:
  requestedStoragePoolProvisionedCapacity:
    poolProvisionedCapacityGb: '102400'
    poolProvisionedIops: '100000'
    poolProvisionedThroughput: '1024'
  storagePoolType: 'hyperdisk-balanced'
status:
  autoCreatedReservations:
  -   https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b/reservations/example-reservation
  fulfilledCount: '2'
  lockTime: '2026-01-27T19:15:00Z'
  procurementStatus: DRAFTING
timeWindow:
  endTime: '2026-02-10T19:20:00Z'
  startTime: '2026-01-27T19:20:00Z'
zone: https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b
  4. In the output, verify that the request details, such as the reservation period, share type, and compute instance properties are correct. Also, if you requested a commitment, verify that it's specified in the output. If the details are incorrect, then contact your account team.
  5. To submit the draft request for review, use the gcloud compute future-reservations update command with the --planning-status flag set to SUBMITTED: 
    gcloud compute future-reservations update FUTURE_RESERVATION_NAME \
    --planning-status=SUBMITTED \
    --zone=ZONE

    Within a few minutes, Google Cloud approves your request, and then Compute Engine creates an empty reservation. If you requested storage resources, then Compute Engine creates a new Storage Pool or Exapool, or increases capacity for an existing Exapool.

REST

  1. To view a list of future reservation requests in your project, make a GET request to the futureReservations.list method. In the request URL, include the filter query parameter and set it to status.procurementStatus=DRAFTING: 
    GET https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/futureReservations?filter=status.procurementStatus=DRAFTING

    Replace the following:

    • PROJECT_ID: the ID of the project where Google created the draft future reservation request.
    • ZONE: the zone where the request exists.
  2. In the request output, look for the reservation request that has the name that you provided to your account team.
  3. To view the details of the draft request, make a GET request to the beta futureReservations.get method: 
    GET https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/futureReservations/FUTURE_RESERVATION_NAME

    Replace FUTURE_RESERVATION_NAME with the name of the draft future reservation request.

    The output is similar to the following:

    {
  "specificSkuProperties": {
    "instanceProperties": {
      "machineType": "a3-ultragpu-8g",
      "guestAccelerators": [
        {
          "acceleratorType": "nvidia-h200-141gb",
          "acceleratorCount": 8
        }
      ],
      "localSsds": [
        {
          "diskSizeGb": "375",
          "interface": "NVME"
        },
        ...
      ]
    },
    "totalCount": "2"
  },
  "storagePoolProperties": {
    "requestedStoragePoolProvisionedCapacity": {
      "poolProvisionedCapacityGb": "102400",
      "poolProvisionedIops": "100000",
      "poolProvisionedThroughput": "1024"
    },
    "storagePoolType": "hyperdisk-balanced"
  },
  "kind": "compute#futureReservation",
  "id": "7979651787097007552",
  "creationTimestamp": "2025-11-27T11:14:58.305-08:00",
  "selfLink": "https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b/futureReservations/example-draft-request",
  "selfLinkWithId": "https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b/futureReservations/7979651787097007552",
  "zone": "https://www.googleapis.com/compute/v1/projects/example-project/zones/europe-west1-b",
  "name": "example-draft-request",
  "timeWindow": {
    "startTime": "2026-01-27T19:20:00Z",
    "endTime": "2026-02-10T19:20:00Z"
  },
  "status": {
    "procurementStatus": "DRAFTING",
    "lockTime": "2026-01-27T19:15:00Z"
  },
  "planningStatus": "DRAFT",
  "specificReservationRequired": true,
  "reservationName": "example-reservation",
  "deploymentType": "DENSE",
  "schedulingType": "INDEPENDENT",
  "autoCreatedReservationsDeleteTime": "2026-02-10T19:20:00Z"
}
  4. In the output, verify that the request details, such as the reservation period, share type, and compute instance properties are correct. Also, if you requested a commitment, verify that it's specified in the output. If the details are incorrect, then contact your account team.
  5. To submit the draft request for review, make a PATCH request to the beta futureReservations.update method. In the request URL, include the updateMask query parameter and set it to planningStatus: 
    PATCH https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/futureReservations/FUTURE_RESERVATION_NAME?updateMask=planningStatus

{
  "name": "FUTURE_RESERVATION_NAME",
  "planningStatus": "SUBMITTED"
}

    Within a few minutes, Google Cloud approves your request, and then Compute Engine creates an empty reservation. If you requested storage resources, then Compute Engine creates a new Storage Pool or Exapool, or increases capacity for an existing Exapool.

What's next