What it does
AWS Fargate is a serverless compute engine for containerized workloads managed through Amazon ECS or Amazon EKS. A team packages an application as a container, specifies CPU and memory, configures networking and IAM policies, then launches a task or pod without managing the host fleet. AWS positions it for web applications, APIs, microservices, and data processing. Fargate is the compute layer, not a standalone container orchestrator, so a deployment still uses ECS or EKS to define and manage workloads.
Specify a task or pod, not a virtual-machine fleet
A typical Fargate deployment starts with a container image and a task definition in ECS, or a pod specification in EKS. The operator selects CPU and memory, configures networking and IAM access, then asks the orchestrator to launch the workload. Fargate supplies isolated compute for the task; AWS documentation says teams no longer need to provision or scale EC2 clusters for those containers. This makes it useful for services whose capacity changes with demand, as well as scheduled or parallel data-processing jobs.
The main architectural choice is which orchestrator owns the workload. ECS offers AWS's own task and service model; EKS uses Kubernetes. Fargate does not replace either scheduler, and the AWS product page presents it as a compute engine used with those services. For a small HTTP service, for example, a team can define an ECS service around its image and let Fargate supply task capacity. That example is illustrative, not a deployment performed for this listing.
Billing is based on requested vCPU, memory, operating system, CPU architecture, and storage while an image is downloaded and the task or pod runs, rounded to the nearest second. The pricing page notes that Windows containers and ARM architecture are currently supported only with ECS. This model removes host administration but does not remove work around IAM, networking, image management, or choosing suitable task resources. No AWS account or task was created for this entry.
Strengths
- Removes the need to provision and scale EC2 clusters for container tasks managed by ECS or EKS.
- Per-second billing is based on selected compute and storage resources, with no upfront capacity purchase.
- AWS documents use cases spanning web applications, APIs, microservices, and parallel data processing.
Watch out for
- A workload must still be defined and managed through Amazon ECS or Amazon EKS; Fargate is not an orchestrator by itself.
- Cost varies with vCPU, memory, operating system, architecture, storage, and task runtime.
- The pricing page limits Windows containers and ARM CPU architecture to ECS rather than EKS Fargate.
Best for
- AWS teams deploying container services without operating a fleet of worker virtual machines.
- ECS or EKS workloads whose compute demand changes over time or arrives as parallel jobs.
- Containerized APIs and microservices already designed for AWS networking and IAM.
Sources
- Serverless Compute Engine – AWS Fargate – AWS (Oct 25, 2026)
- Architect for AWS Fargate for Amazon ECS - Amazon Elastic Container Service (Oct 25, 2026)
- AWS Fargate Pricing (Oct 25, 2026)