AWS Batch cost: free service, you pay for the compute it runs

Quick answer

AWS Batch is free; you pay only for the compute your jobs run on, whether EC2, Fargate, or Spot. So the bill is set by the compute environment you choose and how efficiently jobs pack onto it. Run fault-tolerant jobs on Spot (up to ~90% off), right-size job resource requirements, and let Batch scale to zero between jobs.

AWS Batch schedules and runs batch computing jobs, and like several other AWS orchestration services it charges nothing for itself. Your Batch bill is entirely the compute the jobs consume, so the compute environment configuration is where all the cost decisions live.

You pay for the compute environment

Compute environment	Bills for	Best for
EC2 On-Demand	Instance hours	Steady, interruption-sensitive
EC2 Spot	Spot instance hours (up to ~90% off)	Fault-tolerant batch
Fargate	vCPU + memory per job	Simplicity, spiky small jobs
Fargate Spot	Discounted Fargate	Fault-tolerant, no instance management

Batch can run jobs on managed EC2 (on-demand or Spot), or on Fargate (or Fargate Spot). It scales the compute environment up as jobs queue and down, to zero, when the queue is empty, so you do not pay for idle capacity between jobs.

Spot is the big lever

Most batch work is fault-tolerant: if an instance is reclaimed, the job reschedules. That makes it ideal for Spot, for up to about 90% off compared to on-demand. Running Batch compute environments on Spot is the single biggest saving for suitable workloads.

Right-size and pack efficiently

Set each job's vCPU and memory requirements to what it actually needs so Batch packs jobs efficiently onto instances rather than wasting capacity. Choose Fargate for simplicity and spiky small jobs, and managed EC2 with Spot for large or cost-sensitive batch. Because Batch scales to zero, the levers are the compute environment choice, Spot, and job right-sizing, the same principles as Fargate versus EC2.

FAQ

How is AWS Batch priced?

AWS Batch itself is free; you pay only for the compute your jobs run on, whether managed EC2, Fargate, or Spot, plus any storage and data transfer. The compute environment choice and how efficiently jobs pack determine the bill.

How do I reduce AWS Batch cost?

Run fault-tolerant jobs on Spot compute environments (up to about 90% off), right-size each job's vCPU and memory so Batch packs jobs efficiently, choose Fargate for spiky small jobs and EC2 with Spot for large batch, and rely on Batch scaling to zero between jobs.

Can AWS Batch use Spot instances?

Yes, and it is the biggest saving. Most batch work is fault-tolerant, so a reclaimed Spot instance just reschedules the job. Running Batch compute environments on EC2 Spot or Fargate Spot cuts compute cost by up to about 90% for suitable workloads.

Does AWS Batch charge when no jobs run?

No, if configured to scale to zero. Batch scales the compute environment up as jobs queue and down to zero when the queue is empty, so you do not pay for idle capacity between jobs. Only running jobs incur compute cost.

Batch on Fargate or EC2 for cost?

Fargate is simpler and bills per job's vCPU and memory, good for spiky small jobs with no instance management. Managed EC2, especially with Spot, is usually cheaper per unit of compute for large or steady batch. Choose based on job size, volume, and whether you want to manage instances.

Does C3X estimate AWS Batch cost?

C3X prices the underlying compute (EC2 instances or Fargate) that a Batch compute environment uses, from your Terraform. Job runtime and Spot usage depend on the workload, which you provide as usage assumptions, since Batch itself is free.

What to do next

Price the compute behind your batch jobs before you run them. C3X reads your Terraform and prices your resources against a live catalog. Start with the quickstart.