You don't need a consultant or a paid platform to shave real money off your AWS bill. Most savings come from a handful of boring, low-risk housekeeping steps. Here they are, in order of safest-first.
AWS bills compound quietly. A $40/month forgotten load balancer is $480/year; a right-sized-down instance can be $2,000/year back in your pocket. None of it requires re-architecting — it's finding what you're paying for and not using. The steps below are ordered so you can start with the zero-risk ones today and work up to the commitments only when you're sure.
Open CloudWatch and check average CPU on your EC2 instances over the last 2 weeks. Anything steady under ~20% CPU is almost certainly oversized. Move spiky workloads to burstable t3/t4g instances; move flat-but-low workloads down a family.
Saving: often 30–60% of that instance's cost. A c5.4xlarge → c5.xlarge move is 4x cheaper for the same code.
Risk: low, but test after downsizing — watch error rates for 24h.
Unattached EBS volumes, Elastic IPs with no instance, load balancers serving near-zero traffic, and old NAT gateways are the classic "set and forget" charges. These are the 6 patterns we break down in our AWS waste patterns post — and they're exactly what the free CloudSaver scan flags automatically.
Saving: varies, but idle resources are usually 5–15% of a messy bill.
Risk: low — snapshot before deleting EBS, release only EIPs you recognize.
If you have instances that run 24/7 and aren't going away, a Compute Savings Plan locks in a discount vs on-demand. AWS advertises up to 72% savings vs on-demand for 1- or 3-year commitments (AWS official figure). Start with a small commitment covering only what you're certain is permanent.
Saving: up to 72% on covered usage (AWS official, for 3-yr Compute Savings Plans).
Risk: medium — you're committing to pay for that usage for the term. Under-commit rather than over-commit.
Data rarely accessed but sitting in the Standard tier is overpaying. Add an S3 lifecycle policy to move objects to Infrequent Access after 30 days and Glacier after 90–180 days. Enable Intelligent-Tiering for buckets with unknown access patterns.
Saving: IA is ~40–50% cheaper than Standard; Glacier is ~90% cheaper for archive.
Risk: low — retrieval from Glacier has latency, so don't tier hot data.
Anything that can survive interruption — CI runners, image processing, data pipelines, dev/test — belongs on Spot. AWS advertises up to 90% off vs on-demand for Spot (AWS official figure). Use Spot Fleets or an auto-scaling group with a fallback to on-demand.
Saving: up to 90% on eligible workloads (AWS official).
Risk: medium — jobs must handle interruption. Never put a production database on Spot.
Dev, staging, and demo environments are often billed 24/7 but used 40 hours a week. Use the AWS Instance Scheduler (Lambda + DynamoDB) or a simple cron to stop them nights and weekends. That alone cuts their cost by ~65%.
Saving: ~65% on those environments (they run ~35% of the week instead of 100%).
Risk: low — just make sure nobody's demo breaks Monday morning.
EBS snapshots older than your retention window and AMIs nobody launches anymore are pure storage cost. Set a retention policy (30–90 days for most teams) and delete the rest. Check the EBS Snapshot console sorted by creation date.
Saving: typically $10–50/month per stale account, more at scale.
Risk: low — keep at least one known-good snapshot before pruning.
Every account is different, so treat this as a range, not a promise: a typical small-to-mid account that has never been optimized can usually recover 20–40% of its bill with steps 1–2 and 4–7 (the low-risk ones), and more if it commits to Savings Plans (step 3). That's the same 27–32% waste band Flexera reports — you're not doing anything exotic, you're just cleaning up what was already leaking.
CloudSaver reads your AWS billing CSV in your browser, flags the waste patterns above, and shows what you can recover. No signup, no upload to a server.
Scan my AWS bill free →Want a founder-reviewed deep dive into your specific bill? Email support@getcloudsaver.com.
Not if you go in order. Steps 1, 2, 4, 6, 7 are low-risk if you snapshot/verify first. Savings Plans (step 3) and Spot (step 5) need a bit more care — commit only to what you're sure about.
No. The patterns above are visible in the CSV you can export from the AWS Billing Console today. A free scanner just speeds it up and catches items you might miss by hand.
Those are AWS's own published maximum discounts for Compute Savings Plans and Spot vs on-demand. Your actual saving depends on commitment length and how much of your workload is eligible — most teams see less than the maximum, which is still substantial.