How to Estimate AWS Costs Before You Deploy
Deploying to AWS without a cost estimate is like signing a lease without asking the rent. This page walks through how to use the AWS Pricing Calculator and — more importantly — which hidden costs most estimates miss entirely.
Why estimating matters
Cost estimation before deployment serves two purposes. First, it tells you whether the architecture you’re planning is financially viable at your expected scale. Second, it gives you a baseline — if your actual bill ends up significantly higher than the estimate, you know something is wrong and you can investigate why.
Teams that skip estimation often discover costs through surprise. A startup might design an architecture on paper, deploy it, and only look at the bill two months later when it’s already $3,000 higher than expected. By then, the over-spend has already happened.
The AWS Pricing Calculator at pricing.aws.amazon.com is the official starting point. It’s free to use, doesn’t require an AWS account, and lets you model costs for most AWS services.
Using the AWS Pricing Calculator
The calculator lets you build up an estimate service by service. Here’s the general flow:
- Go to
pricing.aws.amazon.com - Click Create estimate
- Select a region (pricing differs by region — model the region you’ll actually deploy in)
- Add each service you plan to use
- For each service, fill in the usage parameters
For EC2, you specify: instance type, OS, number of instances, utilisation percentage, and whether you want On-Demand or Reserved pricing. The calculator will show you the monthly cost.
For RDS, you specify: instance type, database engine, multi-AZ or single-AZ, storage amount, and backup storage.
For S3, you specify: storage volume, request counts (GET/PUT), and data transfer out.
The calculator saves your estimate with a shareable URL, which is useful for getting approval from finance or management.
Worked estimate: a web app with EC2, RDS, S3, and CloudFront
Let’s build an estimate for a straightforward web application: two EC2 instances running the application, one RDS PostgreSQL instance for the database, S3 for static assets, and CloudFront as the CDN.
EC2 (application servers)
- 2 × m5.large, Linux, On-Demand, running continuously
- $0.096/hr × 2 × 730 hr = $140/month
RDS (PostgreSQL)
- 1 × db.t3.medium, Single-AZ, 100GB SSD storage
- ~$0.068/hr × 730 + $0.115/GB × 100 = $61/month
S3 (static assets)
- 50GB stored, 100,000 GET requests/month, 10GB data transfer out
- 50 × $0.023 + 100k × $0.0004 + 10 × $0.09 = $2.99/month
CloudFront (CDN)
- 100GB data transfer out, 1M HTTPS requests
- ~$8.50 + $0.75 = $9.25/month
Estimated total: ~$213/month
But this estimate is missing several real costs that will show up on the actual bill.
Hidden costs beginners routinely miss
The estimate above looks reasonable, but it omits several line items that will inflate the real bill.
Data egress from EC2 The EC2 instances are sending responses to users. If those API responses total 500GB/month of outbound data from the instances to the internet, that’s $0.09/GB × 500 = $45/month in egress costs. Not included in the calculator unless you specifically add data transfer.
NAT Gateway If your EC2 instances are in private subnets (which is good security practice), they use a NAT Gateway to reach the internet for things like software updates and third-party API calls. NAT Gateway charges $0.045/hour for the gateway itself ($32.85/month) plus $0.045/GB of data processed. A gateway handling 200GB/month of outbound traffic adds another $9 — for a total of ~$42/month just for NAT. See the NAT Gateway page for more detail.
EBS volumes EC2 instances have EBS root volumes. Two m5.large instances with 30GB gp3 volumes each add $0.08/GB × 30 × 2 = $4.80/month. Small, but real.
CloudWatch CloudWatch collects EC2 metrics by default. If you enable detailed monitoring or ship application logs, CloudWatch Logs charges $0.50/GB ingested and $0.03/GB stored. A verbose application logging 50GB/month = $25/month in CloudWatch alone.
RDS backups RDS automated backups are free up to 100% of the database storage size. Beyond that, $0.095/GB/month applies.
Revised estimate: ~$213 + $45 + $42 + $5 + $25 = ~$330/month
That’s 55% higher than the initial estimate. This is not unusual.
The data transfer cost trap
Data transfer is where AWS cost estimates most commonly fall short. Here’s the rule:
- Ingress to AWS (data coming in from the internet): Free
- Egress from AWS to the internet (data going out): $0.09/GB for the first 10TB/month from US regions
- Between AWS services in the same region: mostly free
- Between AWS regions: $0.02/GB (approximate)
- Between AZs in the same region: $0.01/GB each way
The between-AZ cost surprises teams running Kubernetes across multiple AZs. If pods in AZ-a communicate with pods in AZ-b, each GB of traffic costs $0.02 round-trip. A busy microservices application can generate hundreds of GB of cross-AZ traffic per month.
CloudFront helps with egress costs. Data transferred from S3 to CloudFront is free. CloudFront to internet is cheaper ($0.0085–$0.085/GB depending on volume and region) than S3 directly to internet ($0.09/GB). If you’re serving large files or high traffic from S3, routing through CloudFront usually saves money beyond the CDN performance benefit.
Using existing bills as a baseline
If you already have a running workload, your current bill is the most accurate input for future estimates. The AWS Cost Explorer (covered in Cost Explorer reports) lets you see your historical costs by service.
The practical approach for scaling estimates: if your current architecture costs $X/month at Y users/requests, model what happens at 2Y or 5Y. Identify which costs scale linearly (egress, requests, storage) versus which are fixed (the RDS instance, the NAT Gateway hourly charge).
For a new workload with no history, look for similar case studies in the AWS documentation or ask in AWS community forums. AWS also publishes reference architectures with cost breakdowns.
Total cost of ownership considerations
AWS cost estimates typically only cover the cloud bill. A true Total Cost of Ownership (TCO) includes:
- Engineering time to build and maintain the infrastructure
- Tooling costs (monitoring, deployment, security scanning)
- Support plan costs (Developer, Business, Enterprise — $29/month up to 10%+ of monthly bill)
- Training costs for the team
- Cost of incidents caused by infrastructure failures
AWS offers a TCO Calculator for comparing on-premises vs cloud costs, which is mainly used to justify migration decisions. For most cloud-native teams, the relevant question is which cloud architecture is cheapest at the scale you’re planning, not on-premises vs cloud.
Note: Build your cost estimate before finalising architecture decisions. Sometimes a small architectural change — like using S3 + CloudFront instead of streaming from EC2, or using SQS to batch Lambda invocations — can cut costs significantly. It’s much cheaper to change the design on paper than after it’s deployed.
Common mistakes
- Only estimating compute and storage — Forgetting data transfer, NAT Gateway, CloudWatch, and API request costs can make an estimate 50–100% too low.
- Assuming 100% utilisation — The Pricing Calculator assumes continuous use. A dev environment that runs 8 hours/day costs about one third of the 24-hour estimate. Always adjust utilisation percentages to match reality.
- Not modelling growth — An estimate is accurate for the current scale. Model what happens at 2× and 5× traffic to understand whether the architecture remains viable as you grow.
- Ignoring support plan costs — The Business support plan costs 10% of your monthly bill (minimum $100). At $5,000/month spend, that’s $500/month extra. Factor it in.
- Not sharing estimates with finance — Engineering estimates are often optimistic. Getting finance to review the estimate creates accountability and catches assumptions that look unrealistic to non-engineers.
Summary
- Use the AWS Pricing Calculator at pricing.aws.amazon.com to build estimates before deploying
- Common hidden costs: data egress, NAT Gateway per-GB, CloudWatch Logs, EBS volumes, cross-AZ transfer
- Data ingress to AWS is free; egress to the internet costs ~$0.09/GB from US regions
- CloudFront reduces egress costs compared to serving directly from S3 or EC2
- Build a 15–20% buffer into estimates for unknowns
- Use existing bills as a scaling baseline for workloads already running
- A real estimate for a web app is often 40–60% higher than the first-pass calculator output
Frequently asked questions
Is the AWS Pricing Calculator free to use?
Yes. The AWS Pricing Calculator at pricing.aws.amazon.com is a free tool — no AWS account required. You can build, save, and share estimates without signing in.
Why is my actual AWS bill higher than my estimate?
The most common reasons are data egress costs (often underestimated), NAT Gateway per-GB charges, CloudWatch log storage, and API Gateway request counts growing beyond projections. Estimates also assume efficient resource use — idle time and over-provisioning add cost.
How accurate is the AWS Pricing Calculator?
The calculator is accurate for the rates it uses, but estimates are only as good as the inputs. If you underestimate traffic, storage growth, or API call volume, your estimate will be too low. Build in a buffer of 15–20% for unknowns.