Updated 1 Oct 2026 • 7 mins read
Amazon Web Server Price (2026)
Amazon web server pricing depends on instance size, operating system, region, and pricing plan, and the compute rate is only part of the bill. This guide gives monthly EC2 prices by size in US dollars, explains the 2025 change to the free tier, covers Savings Plans, Reserved Instances, and Spot, lists the hidden costs, and compares Lightsail.
Ask how much an Amazon web server costs and you will get answers ranging from free to five figures a month, and every one of them is true for someone. The reason is that AWS does not sell servers at a price; it rents compute by the second at a rate set by size, operating system, region, and how you agree to pay, and then bills the disk, the IP address, and the network traffic separately. That model is powerful once you understand it and baffling until you do.
This guide is the plain version. It gives real monthly prices for common Amazon EC2 server sizes in US East, in US dollars, explains the free tier that quietly changed in 2025 and that most guides still describe wrongly, covers the four ways to pay and what each saves, lists the charges that appear outside the compute line, and compares EC2 with Lightsail, AWS's fixed-price alternative for simple servers. All rates are from the official EC2 On-Demand pricing page as of September 2026; verify yours in the AWS Pricing Calculator before budgeting.
The short answer: An Amazon web server costs from about $7.59 a month for the smallest general-purpose EC2 instance (t3.micro, 2 vCPU, 1 GB RAM) in US East on Linux, to about $30 for a t3.medium, $70 for an m6i.large, and $140 for an m6i.xlarge, all running continuously at On-Demand rates. Add the boot disk, a public IP address, and data transfer, and a realistic small web server lands around $30 to $45 a month; Windows adds a license charge. Amazon Lightsail offers simpler fixed-price servers from $3.50 a month (IPv6-only) or $5 with a public IPv4 address. Savings Plans and Reserved Instances cut compute by up to 72 percent, Spot by up to 90 percent for interruptible work. New AWS accounts since July 2025 receive $100 to $200 in credits rather than a 12-month free server.
Key Takeaways
- EC2 servers in US East run from about $7.59 a month (t3.micro) to $140 (m6i.xlarge) and beyond, at On-Demand Linux rates, before storage, IP, and traffic.
- The compute rate is typically 60 to 75 percent of a real server bill; the boot disk, public IPv4 address ($3.65 a month), and data transfer make up the rest.
- The AWS free tier changed on July 15, 2025: new accounts get $100 to $200 in credits over six months, not 12 months of a free micro server.
- Savings Plans and Reserved Instances save up to 72 percent, Spot up to 90 percent; Graviton (ARM) instances cut cost per unit of performance by roughly 20 percent.
- For a simple website, Lightsail's fixed bundles from $3.50 to $5 a month are cheaper and simpler than EC2; EC2 wins once you need scaling, flexibility, or commitments.
How Amazon Prices a Server
AWS bills an EC2 instance per second, with a one-minute minimum, at an hourly rate determined by four things: the instance size and family (how many vCPUs, how much memory, and whether it is general-purpose, compute-, or memory-optimized), the operating system (Linux is the base rate; Windows and some commercial Linux distributions add a license charge), the region (US regions are generally cheapest), and the pricing plan (On-Demand, Savings Plan, Reserved Instance, or Spot). That rate covers compute only. The server's disk, its public IP address, its outbound network traffic, and any snapshots or load balancers are separate lines, which is why the invoice is always higher than the instance price. Our guide to understanding AWS pricing across EC2, S3, EBS, and RDS covers each of those services' billing dimensions.
Amazon Server Prices by Size (US East, 2026)
The table shows On-Demand rates for common sizes running 730 hours a month, always-on, on Linux and Windows. It covers the sizes most web servers actually use; the full catalog runs to hundreds of types, which we tabulate in our EC2 instance types cost reference. Figures exclude disk, IP, and traffic.
| Instance | vCPU / RAM | Family | Linux / hour | Linux / month | Windows / month |
|---|---|---|---|---|---|
| t3.micro | 2 / 1 GB | Burstable | $0.0104 | ~$7.59 | ~$14 |
| t3.small | 2 / 2 GB | Burstable | $0.0208 | ~$15 | ~$22 |
| t3.medium | 2 / 4 GB | Burstable | $0.0416 | ~$30 | ~$44 |
| t3.large | 2 / 8 GB | Burstable | $0.0832 | ~$61 | ~$88 |
| t4g.medium (Graviton) | 2 / 4 GB | Burstable, ARM | $0.0336 | ~$25 | n/a |
| m6i.large | 2 / 8 GB | General purpose | $0.096 | ~$70 | ~$137 |
| m6i.xlarge | 4 / 16 GB | General purpose | $0.192 | ~$140 | ~$275 |
| c6i.large | 2 / 4 GB | Compute optimized | $0.085 | ~$62 | ~$129 |
| r6i.large | 2 / 16 GB | Memory optimized | $0.126 | ~$92 | ~$159 |
The Free Tier: What Changed in 2025
Most guides to Amazon server pricing still say new accounts get a free micro server for a year. That has not been true since July 15, 2025. Accounts created on or after that date choose between a Free plan and a Paid plan, and both start with $100 in credits, rising to $200 if you complete onboarding tasks, usable across more than 200 services. The Free plan lasts six months or until the credits run out, after which the account closes unless you upgrade. Accounts created before that date remain on the legacy model: 750 hours a month of a t2.micro, t3.micro, or t4g.micro instance for 12 months, plus the always-free allowances. The official AWS Free Tier page describes both. The practical point is that a new account can run a t3.micro for roughly ten to twelve months on $100 of credit if it uses nothing else, or for far less time if the credits also fund storage, a database, and traffic. Budget accordingly.
The Four Ways to Pay, and What Each Saves
On-Demand is the ceiling. Three other plans reduce the compute rate, and they suit different workloads.
| Plan | Discount | Commitment | Best for |
|---|---|---|---|
| On-Demand | None (baseline) | None | New, variable, or short-lived servers |
| Savings Plans (1 or 3 yr) | Up to ~72% | Hourly dollar amount; flexible across families and regions | Steady compute that may change shape |
| Reserved Instances (1 or 3 yr) | Up to ~72% | Specific instance family and region | Servers that will not move or change |
| Spot Instances | Up to ~90% | None; two-minute interruption notice | Batch, CI, dev/test, fault-tolerant work |
The rule that keeps these from backfiring: commit only after right-sizing. A three-year Savings Plan on a server twice the size it needs locks in the waste at a discount. Cover the always-on baseline with a commitment, leave burst on On-Demand, and run interruptible work on Spot. Our guides to Reserved Instances versus Savings Plans versus Spot and how commitment discounts work go into the mechanics.
The Costs Outside the Instance Price
This is where a $30 server becomes a $45 invoice. Every EC2 server carries several separately billed components.
| Component | Typical US East cost | Notes |
|---|---|---|
| Boot volume (gp3 EBS, 30 GB) | ~$2.40 / month | Billed on provisioned size, used or not |
| Public IPv4 address | $3.65 / month | Every public IP, in use or idle, since Feb 2024 |
| Data transfer out | $0.09 / GB after 100 GB free | Traffic to the internet; inbound is free |
| EBS snapshots | ~$0.05 / GB-month | Backups accumulate unless retention is set |
| Application Load Balancer | ~$16 / month plus usage | Production HA; uses 2 to 3 public IPs of its own |
| NAT Gateway | ~$33 / month plus $0.045 / GB | Private-subnet outbound access |
| Windows Server license | Roughly doubles small instances | Bundled into the Windows hourly rate |
The public IPv4 charge is the one that surprises people most, because it is new and applies to every address whether used or not; we cover it in detail in our guide to Elastic IP charges. Data transfer is the one that scales with success: a site that serves 1 TB a month pays about $81 for egress, more than the server itself on a small instance.
A Worked Example: Two Real Servers
A small business website on a t3.small: $15 for compute, $2.40 for a 30 GB disk, $3.65 for the IP, and about $9 for 200 GB of traffic beyond the free 100 GB. Total: about $30 a month. Switching to a t4g.small (Graviton) saves about $3; a one-year Savings Plan saves another $5 or so on the compute line. Lightsail's $12 Small bundle would cover a similar workload with 3 TB of transfer included, and is the simpler choice for a site this size.
A production application server on a t3.large behind a load balancer: $61 for compute, $8 for a 100 GB disk, $16 for the load balancer plus about $11 for its three IPv4 addresses, $33 for a NAT Gateway, and about $36 for 500 GB of traffic. Total: roughly $165 a month, of which the server itself is 37 percent. A one-year Compute Savings Plan brings the compute line to about $40 and the total to about $145. The lesson in both examples is the same: the instance price is the number everyone quotes and less than half of what you pay.
EC2 or Lightsail?
Amazon Lightsail is AWS's answer to the simple-hosting question. It bundles compute, SSD storage, a data transfer allowance, and a static IP into a fixed monthly price: from $3.50 a month for an IPv6-only nano instance (512 MB RAM, 20 GB SSD, 1 TB transfer), $5 with a public IPv4 address, and $12 for the Small bundle with 2 GB RAM and 3 TB of transfer. For a blog, a small website, a WordPress install, or a side project, it is cheaper, more predictable, and far simpler than EC2. EC2 wins the moment you need to pick instance sizes freely, scale automatically, use Savings Plans or Reserved Instances, attach to a VPC and the wider AWS ecosystem, or run anything that outgrows a single fixed bundle. Many teams start on Lightsail and graduate.
How to Pay Less for an AWS Server
- Right-size first. Most servers run far below their capacity; stepping down one size halves the compute line.
- Use Graviton where your stack supports it, for roughly 20 percent off comparable Intel instances.
- Schedule non-production servers to stop outside working hours; a dev server running 8 hours on weekdays costs a quarter of one running all month.
- Drop public IPs on servers that sit behind a load balancer, and release idle Elastic IPs.
- Put a CDN in front of high-traffic sites, since the S3-to-CloudFront leg is free and edge caching cuts origin egress.
- Then commit the steady baseline to a Savings Plan, and only then.
These are the EC2-specific versions of the levers in our AWS cost optimization guide, and keeping them applied over time depends on seeing server cost by resource and by owner, which is what Opslyft's cost visibility is built to do.
The Bottom Line
An Amazon web server costs whatever its instance rate says only if you ignore the disk, the address, and the traffic attached to it, and the invoice will not let you. Small servers start at about $7.59 a month for compute and land around $30 in practice; production servers with load balancing and NAT run into the low hundreds; and the free tier, for new accounts, is now a credit balance that runs out rather than a free server that lasts a year. Pick the smallest instance that does the job, prefer Graviton, mind the IP and egress lines, consider Lightsail for anything simple, and commit only once you know what you actually need. Do that and an AWS server is one of the best values in computing; skip it and the bill quietly becomes twice the price you thought you agreed to.
FAQs
In US East, a small Amazon EC2 server costs about $7.59 a month for a t3.micro (2 vCPU, 1 GB RAM), $15 for a t3.small, $30 for a t3.medium (2 vCPU, 4 GB), and $61 for a t3.large, running all month on Linux at On-Demand rates, before storage, IP address, and data transfer. A realistic small web server with disk, a public IP, and modest traffic lands around $30 to $45 a month. Windows adds a license charge.
Amazon Lightsail is the cheapest way to run a server on AWS, from $3.50 a month for an IPv6-only nano instance or $5 a month with a public IPv4 address, with storage and a data transfer allowance bundled in. On EC2 itself, a t3.micro at about $7.59 a month plus storage and IP is the cheapest practical option; Graviton t4g instances are cheaper still per unit of performance.
For accounts created on or after July 15, 2025, AWS gives $100 in credits at sign-up and up to $100 more for completing onboarding tasks, usable across services for up to six months on the Free plan, rather than a fixed free server. Accounts created before that date keep the legacy 12-month free tier of 750 hours a month of a t2.micro, t3.micro, or t4g.micro instance. Either way, a server is free only while the credits or allowance last.
EC2 bills per second (with a one-minute minimum) at an hourly rate set by the instance size, operating system, region, and pricing model, with Linux as the base rate and Windows adding a license charge. The compute rate covers CPU and memory only; the boot disk (EBS), public IPv4 address, data transfer out beyond 100 GB a month, snapshots, and any load balancer bill separately.
Savings Plans and Reserved Instances cut the compute rate by up to about 72 percent for a one- or three-year commitment, with Savings Plans offering more flexibility across instance families and regions. Spot Instances cut it by up to about 90 percent for interruptible work such as batch jobs and dev environments. Commit only after right-sizing, or the discount locks in an oversized server.
The most common are the EBS boot volume (about $0.08 per GB-month for gp3), the public IPv4 address ($3.65 a month), data transfer out beyond the 100 GB monthly free allowance (about $0.09 per GB), EBS snapshots (about $0.05 per GB-month), and, for production, a load balancer (about $16 a month plus usage) and a NAT Gateway (about $33 a month plus per-GB fees). Together these often add 30 to 50 percent to the compute rate.
Use Lightsail for a simple website, blog, or small application where you want a fixed monthly price with compute, storage, and transfer bundled and minimal configuration. Use EC2 when you need to choose instance sizes freely, scale automatically, use Reserved Instances or Savings Plans, or integrate with the wider AWS ecosystem. Many teams start on Lightsail and move to EC2 as needs grow.