How to Cut Cloud Costs 30% Without Breaking Anything

Most cloud cost advice is either too vague ("turn off what you don't use") or too risky ("just move everything to spot"). Here is the order I actually work in. Reliability first, savings as a byproduct.

Start with visibility, not cuts

You can't optimize what you can't see. Before touching a single instance, get cost allocation working:

• Enforce a tagging policy (team, environment, service).
• Turn on Cost Explorer and build a per-team dashboard.
• Identify the top 10 line items. They are usually 80% of the bill.

If nobody owns a cost, nobody will reduce it. Attribution is the unlock.

Right-size before you commit

Buying Savings Plans on an oversized fleet just locks in waste. Right-size first:

1. Find idle and over-provisioned compute (CPU or memory utilization under 20%).
2. Move stateless workloads to Graviton and spot where safe.
3. Consolidate under-utilized clusters.

# Quick win: find unattached EBS volumes burning money
aws ec2 describe-volumes \
  --filters Name=status,Values=available \
  --query 'Volumes[].{ID:VolumeId,Size:Size}'

Then commit to your baseline

Once usage is stable, cover your steady-state with Savings Plans or Reserved Instances. Never cover your peak. Aim for roughly 80% coverage so you stay flexible.

Add guardrails so it stays fixed

The savings evaporate without guardrails:

• Budget alerts per team.
• Policy-as-code to block expensive resource types by default.
• A monthly 15-minute cost review.

Do these in order and a 30% reduction is routine, with no late-night incidents along the way.