Keel

How do you reduce AWS costs at a small startup?

By the Keel team · Last checked · 4 min read

Short answer

Reduce AWS costs at a small startup in a fixed order: first find out where the money goes, then remove what nobody uses, then resize what is too big, then move to cheaper options such as gp3 storage and Graviton instances, and only then commit to Savings Plans. Doing it in that order stops you locking in a discount on waste.

  • 29%

    Estimated share of cloud spend that is waste

    Source: Flexera, 2026

  • 20%

    Lower cost per GB for gp3 than gp2 EBS volumes

    Source: AWS, Oct 2026

  • Up to 20%

    Lower cost of Graviton EC2 instances than comparable x86, per AWS

    Source: AWS Graviton, Oct 2026

Why do small startups overspend on AWS?

Small startups overspend on AWS mostly because nobody owns the bill: resources get created for a test, a demo or a migration and are never removed, and defaults chosen in a hurry stay forever.

Flexera's 2026 State of the Cloud report estimates that 29% of cloud spend is waste, and that figure rose for the first time in five years.

The good news is that the first round of savings rarely needs clever engineering. It needs someone to look. The guide on why cloud waste is rising covers the wider picture.

Where should a startup start when reducing AWS costs?

Start by finding where the money goes, using Cost Explorer grouped by service and then by usage type, before changing anything.

Cutting blind usually means saving a little on a small line while the real cost carries on.

The step-by-step method is in how to work out which service is driving your AWS bill. Write down your top five cost lines; that list becomes the plan.

What are the quickest ways to reduce AWS costs?

The quickest ways to reduce AWS costs are removing things nobody uses and turning off non-production environments when nobody is working.

These changes carry little risk and often show up on the next bill.

Quick AWS cost reductions for a small startup
ChangeWhat it savesRisk
Delete unattached EBS volumes and old snapshotsStorage you pay for every monthLow, after checking nothing restores from them
Release unused Elastic IPs and public IPv4 addressesHourly public IPv4 chargesLow
Stop dev and staging outside working hoursInstance and database hours overnight and at weekendsLow for non-production
Set retention on CloudWatch log groupsLog storage that grows foreverLow
Remove a NAT gateway you do not need, or share oneHourly NAT gateway and data processing chargesMedium: check private subnets still reach what they need
Move gp2 volumes to gp3gp3 costs 20% less per GB than gp2, per AWSLow; the change is made in place

How do you rightsize AWS instances and databases safely?

Rightsize AWS instances and databases by looking at real CPU, memory and connection use over several weeks, then stepping down one size at a time and watching.

AWS Compute Optimizer gives rightsizing recommendations for EC2 instances, Auto Scaling groups, RDS and more, at no additional charge, based on recent CloudWatch metrics.

  • Check peak use, not the average: a database that idles all day but spikes at 9am needs the headroom.
  • Change one thing at a time, ideally through infrastructure as code, so you can roll back.
  • Rightsize staging and preview environments first; they are usually the most oversized.
  • For containers on ECS, look at task CPU and memory reservations as well as the instances or Fargate size.

Is switching to AWS Graviton worth it for a small startup?

Switching to AWS Graviton is often worth it for a small startup whose code runs on Linux in containers or managed runtimes, because AWS says Graviton-based EC2 instances provide up to 40% better performance at up to 20% lower cost than comparable x86 instances.

Managed services such as RDS also offer Graviton instance types.

The work is mostly in your build: container images need to be built for arm64, and any native dependencies need arm64 versions. Try it on staging first and compare real response times and costs before moving production.

When should a startup buy AWS Savings Plans?

A startup should buy AWS Savings Plans only after cleaning up and rightsizing, and only for the part of its usage that has been steady for a few months.

Committing first locks in a discount on resources you might have deleted next month.

Start small, with a commitment that covers the floor of your usage rather than the average, and review it each quarter. Savings Plans or Reserved Instances compares the options for a team spending under £10k a month.

How do you keep AWS costs down after the first clean-up?

Keep AWS costs down after the first clean-up with three habits: a budget and anomaly alerts, tags on every resource, and a short monthly review of the bill.

Without them, waste builds up again within months.

How to stop a surprise AWS bill covers the alerts, and the monthly AWS cost review checklist covers the review.

Has your startup done the basics of AWS cost reduction?

Tick each item that is true for your account today.

  • You know your top five AWS cost lines by usage type.
  • No unattached volumes, old snapshots or idle IP addresses remain.
  • Non-production environments stop outside working hours.
  • All EBS volumes are gp3, not gp2.
  • Instances and databases were sized from real metrics in the last six months.
  • Any Savings Plan covers only steady, cleaned-up usage.

Decision point: Five or six ticks: further savings need deeper engineering. Three or fewer: there is probably easy money on the table this month.

Sources

  1. Flexera, 2026 State of the Cloud Report (2026)
  2. AWS docs, Analyzing your costs with AWS Cost Explorer (checked October 2026)
  3. AWS, AWS Compute Optimizer pricing (checked October 2026)
  4. AWS, Amazon EBS General Purpose volumes (checked October 2026)
  5. AWS, AWS Graviton Fast Start (checked October 2026)
  6. AWS docs, Compute Savings Plans and Reserved Instances (checked October 2026)
  7. AWS, Amazon VPC pricing (NAT gateway) (checked October 2026)
  8. AWS News Blog, New charge for public IPv4 addresses (July 2023)

Keep reading

Keel is an AWS release and recovery operations service for SaaS teams of 3 to 15 developers, provided by Bargainn Services Ltd, a UK company registered at Companies House (16497746). Registered office: 3rd Floor, 86-90 Paul Street, London, EC2A 4NE, United Kingdom.