I Tried Out the Newly Available Option to Specify ARM64 When Creating a Service in ECS Express Mode
This page has been translated by machine translation. View original
Introduction
With AWS CLI v2.36.44, released on September 11, 2026, you can now specify ARM64 as the CPU architecture when creating an ECS Express Mode service.
You can now also select the CPU architecture on the service creation screen in the Management Console.

Verification with the CLI
I used AWS CLI v2.36.46 for the CLI operations in this article.
$ aws --version
aws-cli/2.36.46 Python/3.14.6 Linux/...
The --cpu-architecture option was added in v2.36.44, so v2.36.44 or later is required if you want to try it with the CLI. The CLI operations below were performed in the ap-northeast-1 region with the default cluster.
Creating a Service with ARM64
I used public.ecr.aws/nginx/nginx:stable-alpine as the container image. It was a multi-architecture image that supports both amd64 and arm64.
I created the service without preparing a task definition. Network-related parameters (such as the VPC and subnet specification) differ by environment, so I specified them by referring to the CLI reference.
aws ecs create-express-gateway-service \
--service-name express-arm64-explicit \
--primary-container "image=public.ecr.aws/nginx/nginx:stable-alpine,containerPort=80" \
--cpu-architecture ARM64 \
--region ap-northeast-1
When I checked the created service, cpuArchitecture was set as specified.
{
"cpu": "1024",
"memory": "2048",
"cpuArchitecture": "ARM64",
"primaryContainer": {
"image": "public.ecr.aws/nginx/nginx:stable-alpine"
}
}
When I checked the launched Fargate task with describe-tasks, the ecs.cpu-architecture attribute was arm64.
"attributes": [
{
"name": "ecs.cpu-architecture",
"value": "arm64"
}
]
From the response headers of the service endpoint, I confirmed that nginx was responding normally.
HTTP/2 200
server: nginx/1.30.4
Custom Task Definition
To use ARM64 when adding a service, a custom task definition is required.
In my verification, I specified the ARN of the task execution role in executionRoleArn. Because I set awslogs-create-group to true, the log group was created automatically on first launch.
{
"family": "express-arm64-custom-td",
"networkMode": "awsvpc",
"requiresCompatibilities": ["FARGATE"],
"cpu": "256",
"memory": "512",
"runtimePlatform": {
"cpuArchitecture": "ARM64",
"operatingSystemFamily": "LINUX"
},
"executionRoleArn": "arn:aws:iam::123456789012:role/ecsTaskExecutionRole",
"containerDefinitions": [
{
"name": "Main",
"image": "public.ecr.aws/nginx/nginx:stable-alpine",
"essential": true,
"portMappings": [
{
"containerPort": 80,
"protocol": "tcp",
"name": "http"
}
],
"logConfiguration": {
"logDriver": "awslogs",
"options": {
"awslogs-group": "/ecs/express-arm64-custom-td",
"awslogs-region": "ap-northeast-1",
"awslogs-stream-prefix": "main",
"awslogs-create-group": "true"
}
}
}
]
}
Conclusion
Previously, when newly creating ECS Express Mode, the initial service was launched on x86_64, so to use ARM64 you had to create a service with a custom task definition specifying ARM64 and replace the initial service, which was a hassle. With this update, you can use ARM64 from the time the initial service is created.
If your development environment is ARM (such as a Mac) and you want to avoid cross-builds, or if you are newly building ECS Express Mode for workloads you want to standardize on the ARM architecture for cost reduction, give this update a try.
