Cutting AWS spend 25% in 90 days without breaking the product

You're an $80M SaaS company running on AWS. The May bill landed at a 38% year-over-year increase, and the new CFO is asking for a 25% cut in 90 days, with the explicit constraint that the product can't degrade. Engineering's first reaction (everything is essential) is the same thing every engineering org says, and it's wrong about 70% of the time at this scale. The bill has compounded for 18 to 24 months without a real owner. Finance has line-item visibility but no architectural context. Engineering has architectural context but no incentive to optimize until someone asks. The 25% target sounds aggressive. It isn't. Cloud bills at this scale typically have 25 to 35% in defensible cuts available without a product change.

Run option A as the primary track and bring in a FinOps platform (option B) as the measurement layer underneath it. Skip option C for this 90-day window.

The CFO's 25% target lands inside option A's range without product risk. The sequence matters. Start with storage and dead environments because they're the lowest-risk wins: orphaned EBS volumes, untiered S3 (Standard when it should be Intelligent-Tiering or Glacier Deep Archive), CloudWatch logs retained forever, and non-production environments that nobody shut down. Together those typically clear 7 to 12% off the bill in 30 days with no product risk.

Then attack commitment coverage. At your size you should be running 60 to 80% Savings Plan or Reserved Instance coverage on steady-state compute. If you're below 30% (most companies in this position are), the next 8 to 12% comes from a Compute Savings Plan locked at one or three years.

Right-sizing compute is the third pull. It's slower, it requires engineering attention, and it's where most cost programs stall, so do it last when the early wins have built momentum. Together: 22 to 28% off the May number, defensible to the board, no product change.

Assume the May AWS run rate is $1.6M to $1.9M monthly ($19M to $23M annualized) at $80M revenue, which sits in the 24 to 28% of revenue range that's typical for SaaS at this scale.

Three-pull program savings, 90 days:

• Storage tiering and dead-environment cleanup: 7 to 12% off, or $1.6M to $2.7M annualized.
• Savings Plan / RI commitment lift: 8 to 12% off, or $1.8M to $2.7M annualized.
• Right-sizing compute: 7 to 10% off, or $1.6M to $2.3M annualized.
• Combined: $5M to $7.7M annualized, or 22 to 28% off the run rate.

Investment: $80K to $160K in year one (FinOps platform $30K to $60K, consulting $50K to $100K). Net first-year savings: $4.8M to $7.5M.

This is a 30 to 50x ROI in year one, which is normal for a company that hasn't run a real cost program before. The reason the math looks too good is that the bill is too high, not that the savings are exotic.

1. Week 1. Name the owner (a VP of Engineering or a Director of Platform). Stand up a FinOps platform with read-only access to the AWS billing account. Pull a tagging audit.
2. Weeks 2 to 4. Storage and dead-environment sweep. S3 Intelligent-Tiering on every bucket without an explicit policy. Identify and shut down non-production environments left running. Tighten CloudWatch retention.
3. Weeks 4 to 6. Compute baseline analysis. Buy first Compute Savings Plan tranche at 60% of steady-state.
4. Weeks 6 to 10. Right-sizing sweep on EC2, RDS, and ElastiCache. Engineering reviews per service. Implement the easy ones; defer the architecture-level changes.
5. Weeks 10 to 12. Second Savings Plan tranche based on right-sized baseline. Final report to CFO.
6. Quarter two onward. Monthly cost review as a permanent meeting. Tagging discipline as a code-review check.

• This week: ask AWS for a free Cost Optimization Hub review and a Trusted Advisor reservation analysis. Both are included in your support contract and surface 30 to 40% of the wins for free.
• This week: name the engineering owner. Cost programs without a named owner stall by week six. Every time.
• Next week: pull the top five biggest line items from the May invoice with engineering. The conversation about why each one is where it is will tell you which option (A, B, or C) is the real shape of the work, ahead of the consulting engagement.