Skip to content

Add support for AWS EC2 G7 and G7e instance families #9854

Description

@kappa8219

What would you like to be added

Add the new AWS EC2 G7 and G7e GPU instance families to the cluster-autoscaler static instance type list.

These instances became generally available on June 18, 2026:

G7 Family (NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, Intel Xeon 6)

Instance Type vCPUs Memory (GiB) GPUs Network (Gibps)
g7.2xlarge 8 32 1 60
g7.4xlarge 16 64 1 60
g7.8xlarge 32 128 2 60
g7.12xlarge 48 192 4 60
g7.24xlarge 96 384 4 60
g7.48xlarge 192 768 8 700

G7e Family (NVIDIA GB202, Intel Xeon Emerald Rapids)

Instance Type vCPUs Memory (GiB) GPUs GPU Memory (GiB) Network (Gibps)
g7e.2xlarge 8 64 1 96 50
g7e.4xlarge 16 128 1 96 50
g7e.8xlarge 32 256 1 96 50
g7e.12xlarge 48 512 2 192 50
g7e.24xlarge 96 1024 4 384 50
g7e.48xlarge 192 2048 8 768 50

Note: G7e family was already added in #9097. Only G7 (non-e) is missing.

Why is this needed

Without these instance types in the autoscaler static list, clusters using G7/G7e instances in Auto Scaling Groups get errors:

updating Auto Scaling Group: operation error Auto Scaling: UpdateAutoScalingGroup, StatusCode: 400, api error ValidationError: The specified instance type g7.2xlarge is not valid

Region availability

Currently available in:

  • US East (Ohio) - us-east-2
  • US West (Oregon) - us-west-2

Reference

Similar to #8509 which added r8i and r8i-flex support.

Source: https://instances.vantage.sh/aws/ec2/g7.2xlarge

Metadata

Metadata

Assignees

No one assigned

    Labels

    area/cluster-autoscalerIssues or PRs related to the Cluster Autoscaler componentneeds-triageIndicates an issue or PR lacks a `triage/foo` label and requires one.

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions