Cloud spending has a way of sneaking up on small teams. Industry surveys have repeatedly found that a large share of organizations overshoot their cloud budgets, and serverless container services are a classic culprit because the meter runs quietly in the background. AWS Fargate is one of those services: no servers to patch, no clusters to babysit, just containers that run when you tell them to. The catch is that “no servers” does not mean “no bill,” and understanding where the charges come from is the difference between a $10 monthly line item and a $200 surprise.
Quick Answer: How much does AWS Fargate cost for a small project or business?
A small project typically lands between $10 and $40 per month in raw Fargate compute, depending on how much vCPU and memory each task uses and whether it runs around the clock. You pay per second for the CPU and memory your containers consume, with a one-minute minimum, and you can check exact rates on the AWS Fargate pricing page and model your own numbers in the AWS pricing calculator.
That figure covers compute only. Load balancers, NAT gateways, log storage, and container image storage are billed separately and can easily double a tiny bill. Fargate Spot offers discounted capacity for workloads that can tolerate interruptions, which changes the math for batch jobs.
Before committing, compare per-vCPU and per-GB rates in your region, the free tier allotment for new accounts, Spot discounts, storage and networking charges, and whether a cheaper compute option like EC2 or Lightsail fits better.
The Two Numbers That Drive Every Fargate Bill
Fargate pricing is refreshingly simple on the surface. You are charged for two things: vCPU-hours and GB-hours of memory, billed per second with a one-minute minimum. Multiply the resources you request by the time the task runs, and you have your compute cost.
In the us-east-1 region, on-demand Linux rates have historically sat around $0.04048 per vCPU-hour and $0.004445 per GB-hour. Those exact figures shift over time and vary by region and operating system, so treat them as a planning reference and confirm current rates on the AWS Fargate pricing page before you build a budget.
Here is what that looks like in practice for common small-project configurations, assuming 730 hours in a month:
| Task size | vCPU | Memory | Approx. monthly compute |
|---|---|---|---|
| Tiny API or cron worker | 0.25 | 0.5 GB | ~$9–10 |
| Small web app | 0.5 | 1 GB | ~$18–20 |
| Standard app container | 1 | 2 GB | ~$35–40 |
| Heavier service | 2 | 4 GB | ~$70–80 |
The pattern is linear, which makes forecasting easy. Double the resources, double the bill. Run the task for half the month, halve the bill. That predictability is one of Fargate’s underrated strengths for small teams that need to explain costs to a founder or a client.
Why Your Real Bill Is Higher Than the Compute Estimate
The compute number is the headline, not the whole story. In practice, the extras are where small teams get caught out, and it happens often enough that it is worth calling out plainly.
Load balancers. If your containers sit behind an Application Load Balancer, expect roughly $16–20 per month as a baseline before traffic charges. For a $10 Fargate task, the load balancer can be the bigger line item.
NAT routing. Containers in private subnets that need outbound internet access route through NAT, which has historically cost around $32 per month plus data processing. This single component can triple a small bill, and it is the most common budgeting mistake teams make.
Logging and storage. CloudWatch Logs charges for ingestion and retention. Container images in ECR charge for storage. Neither is huge at small scale, but both grow quietly.
Data transfer. Outbound traffic to the internet is billed per GB. Most small projects stay well within reasonable limits, but media-heavy apps should check the numbers.
A useful habit: before launching anything, sketch the full architecture and price every component, including the containers. The AWS pricing calculator lets you add each service and see the combined monthly figure, which is far more honest than pricing Fargate in isolation. The same discipline applies to any subscription stack a small business runs; comparing line items for tools like 1Password uses the same rigor as reviewing cloud infrastructure, and it pays off.
Fargate Spot, Savings Plans, and the Free Tier
Three levers can pull a Fargate bill down, and small teams often ignore all of them.
Fargate Spot sells spare capacity at a steep discount, historically up to around 70% off on-demand rates. The trade-off is interruption: AWS can reclaim Spot capacity with a short warning, so it suits batch jobs, background workers, and anything that can checkpoint and retry. It does not suit a customer-facing API that must stay up.
Compute Savings Plans trade a one- or three-year commitment for a lower hourly rate. If you know a workload will run continuously for a year, the discount is real money. If you are still experimenting, skip it; a commitment on a project you abandon is worse than paying full price.
The AWS Free Tier has historically included a monthly allotment of Fargate vCPU-hours and GB-hours for new accounts. The amounts have varied year to year, so check the AWS Free Tier page for current terms. For a very small project, the free allotment can cover the first month or two entirely, which makes Fargate a low-risk way to test an idea.
Fargate Versus the Alternatives for a Small Business
Fargate is rarely the cheapest option in absolute terms. It is often the cheapest option once you count engineering time, and that distinction matters enormously for a small team.
| Option | Typical monthly cost | Operational effort | Best for |
|---|---|---|---|
| Fargate (0.25 vCPU) | ~$9–10 compute, plus extras | Very low | Small APIs, cron jobs, prototypes |
| EC2 t4g.small | ~$12–15 | Medium (patching, scaling) | Steady workloads with predictable load |
| Lightsail container | ~$7–15 flat | Low | Simple apps with modest traffic |
| Lambda | Pennies at low volume | Low | Event-driven, bursty workloads |
The honest framing: Fargate costs more per unit of compute than raw EC2, but it removes the patching, capacity planning, and cluster management that eats evenings. For a two-person team shipping a product, that trade is usually worth it. For a hobby project running one small container, Lightsail or a single EC2 instance will often be cheaper.
It helps to think in the same terms you would when choosing any business tool. Just as a small business weighs the monthly fee for accounting software against the hours it saves, Fargate should be judged on total cost including time, not on the per-hour rate alone. The same logic applies when comparing website builders for a small business: the sticker price is only part of the decision. (How much does a ps5 cost)
Practical Tips That Actually Reduce the Bill
After watching teams run small workloads on Fargate, a few habits consistently separate the tidy bills from the messy ones.
Right-size the task definition. The single biggest lever. Many teams default to 1 vCPU and 2 GB when 0.25 vCPU and 0.5 GB would do. Start small, watch CPU and memory utilization in CloudWatch, and scale up only when metrics demand it. Over-provisioning is the most common and most expensive mistake.
Avoid NAT for small projects. If your containers only need to serve traffic, place them in public subnets with a security group and skip NAT entirely. That one decision can save $32 or more per month.
Use ARM (Graviton) where possible. ARM-based Fargate tasks have historically carried lower rates than x86, and most modern container images support the architecture. Rebuild for ARM and the same workload costs less.
Turn off what you are not using. A staging environment that runs 24/7 for occasional testing is pure waste. Schedule it to stop overnight, or run it on Fargate Spot.
Set a billing alarm. A simple CloudWatch billing alarm at a threshold you care about costs almost nothing and catches runaway tasks before they become a story you tell at dinner.
Check your region. Rates differ by region, and data residency requirements sometimes force a more expensive one. Know the trade-off before you commit.
If you are also budgeting for the hardware and software around your project, the same careful approach helps: whether it is figuring out how much a PS5 costs for the office or how much Minecraft costs for a team-building session, small line items add up fast.
A Realistic Monthly Budget for a Small Project
Putting the pieces together, here is a plausible stack for a small business running a containerized API with modest traffic:
- Fargate task at 0.5 vCPU and 1 GB, running continuously: roughly $18–20
- Application Load Balancer: roughly $16–20
- CloudWatch Logs with modest retention: roughly $2–5
- ECR image storage: under $1
- Data transfer: usually under $2 at small scale
That lands around $40–50 per month for a production-ready setup, or closer to $10–20 if you skip the load balancer and expose the service differently. A development environment that runs only during working hours can cost a fraction of that again.
The key insight for a small business is that Fargate’s compute is cheap; the surrounding infrastructure is what determines the final number. Budget the architecture, not the service.
Conclusion
AWS Fargate is not the cheapest way to run a container, but for a small project or business it is often the most predictable and least labor-intensive. Compute costs for a typical small workload land between $10 and $40 per month, and the extras, especially load balancers and NAT, are what push real bills toward $50. Right-size your tasks, skip NAT when you can, consider ARM and Spot, and set a billing alarm. Do those four things and Fargate becomes one of the few cloud services a small team can actually forecast with confidence.
Frequently Asked Questions
Is Fargate cheaper than running EC2 for a small project?
Not on raw compute, no. A small EC2 instance costs less per hour than equivalent Fargate capacity. Fargate wins when you count the engineering time saved on patching, scaling, and cluster management. For a team with limited hours, that trade usually favors Fargate.
Does Fargate charge me when no tasks are running?
No. Fargate bills only for the vCPU and memory consumed by running tasks, per second with a one-minute minimum. An idle service with zero tasks costs nothing in compute, though associated resources like load balancers and log storage still bill.
What is the smallest Fargate task size available?
Fargate supports task sizes as small as 0.25 vCPU with 0.5 GB of memory, which is enough for lightweight APIs and scheduled jobs. That configuration is the cheapest always-on option and typically runs around $9–10 per month in compute.
Can I use Fargate Spot for a production workload?
Only if the workload tolerates interruption. Fargate Spot can reclaim capacity with a short warning, so it suits batch processing, background queues, and jobs that checkpoint. Customer-facing services that must stay available should stay on on-demand capacity.
Are there hidden costs with Fargate?
The compute rate is transparent, but the surrounding architecture is not automatically included. Load balancers, NAT gateways, CloudWatch Logs, ECR storage, and data transfer are billed separately and frequently exceed the compute cost for small deployments.
How do I estimate my Fargate bill before deploying?
Use the AWS pricing calculator to model each component of your architecture, including the containers. Multiply your task’s vCPU and memory by the regional hourly rates, add the supporting services, and compare the total against alternatives like Lightsail or EC2 before committing.
