AWS EC2 Cost Calculator

The instance is the small part of the bill

An m5.large running continuously costs $70.08 a month — 730 hours at roughly $0.096 an hour — which is about ₹6,027 at ₹86 to the dollar, or $840.96 a year.

That figure is the floor, not the bill. A production EC2 instance also carries EBS storage, snapshots, data transfer out, an Elastic IP when detached, load balancer hours, and CloudWatch beyond the free metrics. It is entirely normal for the compute line to be half of what the workload actually costs.

How EC2 compute is priced

Monthly cost = hourly rate × hours per month

730 hours is the convention AWS itself uses for monthly estimates: 8,760 hours a year divided by twelve. It is not 24 × 30 = 720, and using 720 understates a continuously running instance by a little over 1%.

Billing is per second for Linux instances with a 60-second minimum, so a machine running 18 hours a day genuinely costs 18/24 of the full-time rate. That makes shutting down non-production instances outside working hours one of the few savings available with no commitment and no architectural change.

Where the discounts actually are

On-demand is the most expensive way to run EC2 and the rate this calculator uses. The alternatives, in rough order of saving:

Savings Plans — commit to a dollar amount per hour for one or three years. Compute Savings Plans apply across instance families, regions, and even Fargate and Lambda, typically saving up to 66%. EC2 Instance Savings Plans are restricted to a family in a region and save slightly more. For most workloads the flexibility of the compute plan is worth the small difference.

Reserved Instances — the older mechanism, tied to instance type and region, with comparable discounts. Standard RIs save more; Convertible RIs allow changing family later for a smaller discount.

Spot Instances — spare capacity at up to 90% off, reclaimed with a two-minute warning. Excellent for batch processing, CI runners, rendering and anything checkpointed. Unsuitable for a stateful single-node database.

Graviton instances — AWS's ARM processors (the `g` suffix, such as m7g.large) typically deliver 20% better price-performance than the x86 equivalent. For anything that compiles for ARM — most modern runtimes do — this is a straightforward saving with no commitment.

Reading your estimate

The rate here is an on-demand US-region approximation. Real pricing varies by region — the Mumbai region (ap-south-1) generally runs somewhat above us-east-1 — and by operating system, since Windows and commercial Linux distributions carry licence costs on top.

The INR figure depends on the exchange rate you enter. AWS bills Indian customers in USD or INR depending on the account's registration, and where billing is in USD your effective cost moves with the rupee. A 5% currency move on a large annual commitment is a real budget variance.

GST applies at 18% on AWS services for Indian customers, which the estimate excludes.

Common mistakes

Sizing from peak load. Most instances run far below their provisioned capacity. Check CloudWatch utilisation before renewing a size.

Forgetting EBS. A gp3 volume costs about $0.08 per GB-month and continues charging when the instance is stopped — a stopped instance is not a free instance.

Leaving Elastic IPs allocated. Unattached public IPv4 addresses are billed hourly, and since 2024 all public IPv4 addresses carry a charge.

Ignoring egress. Data transfer out to the internet is priced separately and is frequently the line item that surprises.

Committing before measuring. A three-year Savings Plan on an oversized instance locks in the waste. Right-size first, commit second.