How to Cut Your Cloud Bill by 62% Using These 4 Forgotten Settings

AWS cost optimization

Here are the 4 settings you can changed to cut your cloud bill by 62% without rewriting any code – works for AWS, Azure and GCP.

e.g. Last month cloud bill was £1,847.

x

This month: £702.

No need to delete servers. no rewrite code. Just changed 4 settings that AWS, GCP and Azure hide in the docs because they make less money when you use them.

If you’re running anything on the cloud as a startup or small business, do these today.

The Bill Before (Real Numbers)

Stack: 3 EC2 instances, 1 RDS, S3 + CloudFront, NAT Gateway

  • EC2 On-Demand: £892
  • RDS: £341
  • Data Transfer / NAT: £287
  • S3 + Other: £327
  • Total: £1,847/mo

Most organizations can reduce AWS costs 20-30% without changing architecture, but with these four, I got 62%.

Setting #1: Switch to ARM / Graviton (20% Instant Saving)

This can take 7 minutes.

AWS, Azure and GCP all offer ARM processors that are 20% cheaper and 40% more efficient. For AWS it’s Graviton.

What to do:
In EC2 console > Launch Template > Changed m6i.large to m7g.large (same specs, ARM).

For RDS: Modified instance to db.t4g.medium.

Zero code changes – Node.js, Python, Go, Java all run natively on ARM now. Only old x86 binaries break.

Saving: £178/month

how to lower GCP bill

Setting #2: Turn On Savings Plans – Not Reserved Instances (Up to 72% Off)

This is the one AWS recommends but hides the math.

Savings Plans reduce your bill by up to 72% vs On-Demand in exchange for a 1 or 3 year hourly spend commitment. There are Compute, EC2 and SageMaker Savings Plans.

Mistake you can made: Buying Reserved Instances for specific instances. Too rigid.

What to do instead:
AWS Billing > Savings Plans > Buy Compute Savings Plan > $5/hour commitment for 1 year. Covers ANY EC2, Fargate, Lambda.

Same trick exists: GCP calls it Committed Use Discounts – 37% off for 1 year. Azure calls it Reserved Instances.

Saving: £312/month

Setting #3: Kill the $0.045/GB Tax – VPC Endpoints + Free Multicloud

This is a great tip.

Unsuspecting teams rack up serious bills moving data out of AWS. NAT Gateway was charging for every S3 request.

Two fixes:

A) VPC Endpoint for S3 (2 minute fix):

aws ec2 create-vpc-endpoint \
  --vpc-id vpc-xxx \
  --service-name com.amazonaws.us-east-1.s3

Without endpoints, S3 traffic routes through NAT Gateways, incurring data transfer costs. With endpoint, it’s free and private. Plus CloudFront caching reduces origin requests.

B) AWS Interconnect – Multicloud (New in 2026):
AWS now charges nothing per GB to send traffic to GCP or Oracle, just an hourly port charge. There’s a free tier of one 500 Mbps port per provider.

Move analytics to BigQuery and stopped paying AWS egress.

Saving: £287/month – entire data transfer line can go to £12.

Setting #4: Auto-Shutdown Non-Production (The 39% in 12 Weeks Trick)

One SaaS org cut AWS costs by 39% in 12 weeks using this phased approach.

Dev + staging ran 24/7. They only needed to run 9am-6pm Mon-Fri.

What to do:

  • AWS Instance Scheduler – free tool from AWS
  • Tag: Schedule=office-hours
  • Lambda turns off dev RDS + EC2 at 7pm, on at 8am
  • Savings: 65% of non-prod hours

Also: Set up billing alerts in AWS Budgets to alert when you exceed thresholds. you can set them at £25, £100, £500.

Saving: £368/month

After: The New Bill

  • EC2 Graviton + Savings Plan: £402 (was £892)
  • RDS Graviton: £198 (was £341)
  • Data Transfer: £12 (was £287)
  • S3 + Other: £90 (was £327) – lifecycle policy to Intelligent-Tiering
  • Total: £702/mo – Down 62%

Your 15-Minute Checklist (Do This Today)

  1. In Cost Explorer, filter by “Data Transfer” – if >10% of bill, add VPC Endpoints
  2. Change one EC2 type to Graviton / ARM – test, then roll all
  3. Buy Compute Savings Plan for 1 year – don’t overcommit, start at 60% of current usage
  4. Tag dev/staging auto-shutdown and turn on Instance Scheduler
  5. Move static assets to Cloudflare R2 or CDN – 10x cheaper than S3 egress

Bonus: Free Tier Comparison 2026

If you’re still pre-launch, you shouldn’t be paying at all:

  • GCP: Most generous always-free – e2-micro, Firestore, 2M Cloud Functions
  • AWS: t2/t3.micro 750hrs, Lambda 1M req
  • Azure: B1s 750hrs, Functions 1M
  • Oracle: 4-core ARM server forever free – best kept secret

Startup credits: GCP up to $200K, Azure up to $150K, AWS Activate up to $100K.