Request TPU capacity in All Capacity mode

Preview

This feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

Contact your Google Cloud account team to request an All Capacity mode reservation.

Request TPU capacity in All Capacity mode with optional project sharing

Contact your account team with the required intake information. Your account team will create a future reservation in draft state for you to review through the Google Cloud console.

You can request a reservation that can be shared across other Google Cloud projects in your organization. You can also request a reservation that can be shared with another Google Cloud service like Vertex AI. Indicate your reservation share preference as part of the intake information you give your account team:

Property	Description
Project number	The number (not ID) of the project where your account team creates the request and Compute Engine provisions the capacity.
TPU generation	Trillium or Ironwood
Zone	The zone where you want to reserve capacity.
Number of chips	The total number of TPU chips to reserve.
Start time	The start time of the reservation period
End time	The end time of the reservation period
Reservation name	The name of the reservation that Compute Engine creates to deliver your reserved capacity
Reservation automatic deletion	Whether you want Compute Engine to automatically delete the auto-created reservation 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.
Share type	Whether only your project can use the auto-created reservation (LOCAL), or other projects can use the reservation (SPECIFIC_PROJECTS). You can't change this setting after you submit the request. To share reserved capacity with other projects in your organization, do the following: If you haven't already, verify the project where Google creates the request is allowed to create shared reservations. Provide the IDs 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 you must purchase and attach a resource-based commitment to your reserved resources. You can purchase a commitment with a 1-year or 3-year plan. If you share the reserved capacity with other projects, those projects get discounts only if they use the same Cloud Billing account as the project where you reserve capacity. For details, see Enable CUD sharing for resource-based commitments.

The following properties are preset for your All Capacity mode reservation:

Property	Description
Maintenance scheduling type	How Compute Engine schedules host maintenance for the VMs in your cluster. TPU Cluster Director only supports the GROUPED option, which means maintenance is synchronized across all TPU VMs in a sub-block of your reservation.
Reservation operational mode	How Google Cloud manages the VM recovery process after a host error or faulty host report. TPU Cluster Director only supports the `ALL_CAPACITY` option, which means you must self-manage the maintenance and recovery process for a VM. However, you have full access to all machines in your reservation, as well as more granular control over your VM topology and maintenance.
Deployment type	Capacity deployment type. TPU Cluster Director only supports `DENSE`.

Confirm your capacity reservation draft

You must confirm the draft reservation your account team creates for you in order to receive the TPU Cluster Director capacity in your Google Cloud project. Follow the instructions from Review and submit a draft reservation request to confirm the reservation.

Check the status of a capacity request

You can determine the status of your TPU Cluster Director capacity request. If the request is approved, the capacity is allocated at the requested start time for the requested duration. Billing starts at the reservation start time.

For more details, see View future reservations.

To view the details of a future reservation request (a reservation named "exfr1"), run the following command in the Google Cloud CLI:

gcloud compute future-reservations describe exfr1 \
    --zone=us-central1-a

The output is similar to the following:

creationTimestamp: '2025-04-01'
kind: compute#FutureReservation
name: 'exfr1'
   aggregateReservations:
     vmFamily:VM_FAMILY_CLOUD_TPU_LITE_DEVICE_CT6L
status:
  lockTime: 2024-03-01T17:00:00Z
  procurementStatus: APPROVED
timeWindow:
  endTime: 2025-06-01T17:00:00Z
  startTime: 2024-05-01T17:00:00Z
...

Expand an existing capacity request

You can receive additional capacity through an existing reservation. When additional capacity is added, it inherits all of the reservation properties such as reservation sharing policies. You can request a CUD with the additional capacity. The new CUD will point to the same reservation. Your account team will complete the reservation expansion process for you.

View reservation details from the Google Cloud console

Once your future reservation is fulfilled, you can find it on the Google Cloud console Reservation page and view its details. A TPU reservation with a deployment type Dense indicates that it's a TPU Cluster Director reservation.