Fargate Spot cost savings: cutting container bills up to 70 percent

Quick answer

Fargate Spot runs ECS tasks on AWS spare capacity for up to about 70 percent less than regular Fargate's on-demand price, with the tradeoff that AWS can reclaim the capacity with a two-minute warning. It is ideal for stateless, fault-tolerant, and interruptible workloads: batch jobs, queue workers, CI runners, dev and test environments, and stateless web services behind a load balancer that can absorb task loss. A common pattern runs a baseline of tasks on regular Fargate for stability and the elastic portion on Fargate Spot, so you get resilience plus savings. The rule is: put interruptible container work on Fargate Spot and keep only what truly cannot tolerate interruption on regular Fargate.

Fargate makes running containers operationally simple, but its convenience price is higher than raw EC2. For workloads that can tolerate interruption, Fargate Spot recovers most of that premium by running tasks on spare capacity at a deep discount. It is the container equivalent of EC2 Spot, and for the right workloads it is one of the largest savings available on ECS.

How much Fargate Spot saves

Option	Relative price	Interruption	Best for
Regular Fargate	Baseline (on-demand)	None	Steady, interruption-sensitive tasks
Fargate Spot	Up to ~70% cheaper	2-minute warning	Fault-tolerant, stateless, batch
Compute Savings Plan on Fargate	Up to ~50% cheaper	None	Steady baseline, committed

Regular Fargate charges about $0.04048 per vCPU-hour and $0.004445 per GB-hour in us-east-1. A task with 1 vCPU and 2 GB costs about $0.0494 per hour, or roughly $36 per month running continuously. On Fargate Spot the same task can cost around $0.015 per hour, roughly $11 per month, a saving of about 70 percent for work that can survive the occasional interruption.

What you trade for the discount

AWS can reclaim Fargate Spot capacity when it needs it, giving the task a two-minute warning (a SIGTERM) before stopping it. So the workload must handle being stopped and restarted gracefully: no in-memory state that cannot be lost, idempotent processing, and orchestration that reschedules the task. For stateless services and retryable jobs this is a non-issue; for a stateful task holding a long-running session, it is a dealbreaker.

Workloads that fit

Batch and data-processing jobs that can restart, queue workers pulling from SQS that can reprocess a message, CI and build runners, dev and test environments, and stateless web or API services behind a load balancer that redistributes traffic when a task drops. These absorb interruption naturally, which is exactly the profile that makes Spot safe. The same logic governs EC2 Spot on the instance side.

The baseline-plus-Spot pattern

The safest way to use Fargate Spot is a capacity provider strategy that splits tasks between regular Fargate and Fargate Spot. Run a baseline of tasks on regular Fargate so the service always has some stable capacity, and place the elastic scaling portion on Fargate Spot for savings. If Spot capacity is reclaimed, the baseline keeps serving while replacements reschedule. This is configured through ECS capacity providers, covered in the capacity providers guide.

Putting it together on cost

Identify which of your ECS tasks are interruptible, move those to Fargate Spot, and keep a stable baseline on regular Fargate (ideally under a Compute Savings Plan for a further discount on the steady portion). For a service running many tasks, shifting even half of them to Spot can cut the container bill by a third or more. Weigh this against running on EC2 nodes, which can be cheaper still at scale. Price your Fargate and Fargate Spot mix against the resource catalog so the savings are a concrete number before you deploy.

FAQ

How much does Fargate Spot save?

Up to about 70 percent versus regular Fargate's on-demand price. A task with 1 vCPU and 2 GB that costs about $0.0494 per hour (roughly $36 per month) on regular Fargate can run around $0.015 per hour (roughly $11 per month) on Fargate Spot. The saving comes from running on AWS spare capacity, in exchange for the risk that AWS can reclaim the task with a two-minute warning.

What workloads are safe on Fargate Spot?

Stateless, fault-tolerant, and interruptible workloads: batch and data-processing jobs that can restart, queue workers that can reprocess a message, CI and build runners, dev and test environments, and stateless web or API services behind a load balancer that redistributes traffic when a task drops. Any task that can be stopped and rescheduled without losing critical state is a good fit for the discount.

What is the downside of Fargate Spot?

AWS can reclaim the capacity when it needs it, giving the task a two-minute warning (a SIGTERM) before stopping it. So the workload must handle interruption gracefully: no unsaved in-memory state, idempotent processing, and orchestration that reschedules the task. Stateful tasks holding long-running sessions or in-progress work that cannot be lost are not suitable for Fargate Spot.

Can I mix Fargate Spot with regular Fargate?

Yes, and it is the recommended pattern. An ECS capacity provider strategy can run a baseline of tasks on regular Fargate for stability and place the elastic scaling portion on Fargate Spot for savings. If Spot capacity is reclaimed, the baseline keeps the service running while replacement tasks reschedule. This gives you resilience plus most of the Spot discount on the variable portion of the workload.

How does C3X help with Fargate Spot cost?

C3X prices your ECS and Fargate configuration from Terraform against a live catalog, so you can compare a regular-Fargate deployment against a Fargate Spot or mixed strategy before you deploy. That surfaces the up-to-70-percent savings on interruptible tasks as a concrete monthly number in the pull request, so you put the right work on Spot instead of paying full Fargate rates for everything.

What to do next

Cut container cost with Fargate Spot where it is safe. C3X reads your Terraform and prices your resources against a live catalog. Start with the quickstart.