AWS Cost Forecast & Rightsizer
Forecast annual AWS compute spend based on growth rates and estimate savings from purchasing Compute Savings Plans or Reserved Instances (RIs).
Formula
Cost = (CommittedSpend × (1 - DiscountRate)) + (ExcessUsage × OnDemandRate)
We project your growth to estimate total On-Demand spend. Then we apply the Savings Plan discount to the percentage of usage you choose to cover (e.g., 80%). Any usage above that coverage target is billed at standard On-Demand rates.
Worked Examples
Example 1: Startup Growth
Problem:$10k/mo, 5% MoM Growth, 1yr Partial Upfront (30% off)
Solution:Forecast On-Demand: ~$159k. With SP: ~$125k.
Result:$34k Annual Savings
Example 2: Stable Enterprise
Problem:$50k/mo, 0% Growth, 3yr All Upfront (55% off)
Solution:Forecast OD: $600k. With SP: $270k.
Result:$330k Annual Savings
Frequently Asked Questions
What is a Compute Savings Plan?
It's a flexible pricing model that offers low prices (up to 66% off) in exchange for a commitment to a consistent amount of usage (e.g., $10/hour) for a 1 or 3-year term. It applies to EC2, Fargate, and Lambda.
How does it differ from Reserved Instances (RIs)?
RIs require you to commit to specific instance types or regions. Savings Plans are more flexible—they apply automatically to any usage globally (for Compute SPs), reducing the risk of unused commitments.
What is 'Coverage Target'?
The percentage of your current usage you want to cover with a commitment. 100% is risky because if your usage drops, you still pay. 70-80% is a safe 'Rightsized' target.
Should I do 1 Year or 3 Years?
3 Years offers ~2x the savings but locks you in. Only choose 3 Years if your business is very stable. Startups usually stick to 1 Year to maintain agility.
What does 'Payment Option' mean?
'All Upfront' gives the best discount but requires cash now. 'No Upfront' requires no cash but has a slightly lower discount and is billed monthly. 'Partial' is a middle ground.
Does this include Spot Instances?
No. Spot Instances are already heavily discounted (up to 90%) but can be interrupted. Savings Plans do not apply to Spot usage.
What if my usage drops?
With a Savings Plan, you pay the committed hourly rate regardless of usage. If you commit to $10/hr and only use $5/hr, you still pay $10. This is why rightsizing coverage is crucial.
Does this forecast Storage (S3) costs?
No, AWS Cost Forecast & Rightsizer focuses on Compute (EC2/Fargate/Lambda) where Savings Plans apply. S3 has its own tiering logic (Intelligent Tiering).
How accurate is the forecast?
It uses a simple compound monthly growth rate (MoM). Real cloud spend is often spiky. Use this for directional budgeting, not exact accounting.
Can I stack Savings Plans?
Yes! A common strategy is to buy a base layer (e.g., 50% coverage) on a 3-year term, and layer 1-year plans on top as you grow. This 'laddering' minimizes risk.
Background & Theory
FinOps Fundamentals
Cloud cost management (FinOps) is about balancing Speed, Cost, and Quality. Savings Plans are the primary lever for Cost without sacrificing Speed or Quality.
Rightsizing Strategy
The goal isn't to cover 100% of usage. Usage fluctuates. A "Waterline" strategy involves:
- Base Load: Covered by 3-Year Savings Plans (Max discount).
- Predictable Growth: Covered by 1-Year Savings Plans (Flexibility).
- Variable Spikes: On-Demand or Spot Instances.
Payment Options
- All Upfront: Capital-intensive but max ROI. Good for cash-rich companies.
- Partial Upfront: Balance of cash flow and discount. Usually 50% down.
- No Upfront: Pure OpEx. Zero risk to cash balance, but lower total savings.
Why Forecast?
Cloud spend typically compounds. A $10k bill growing at 5% MoM becomes $18k/month in a year. Committing to a Savings Plan based on *today's* spend might leave you under-covered in 6 months. This calculator helps you see that gap.
History
The On-Demand Era
When AWS launched EC2 in 2006, the revolutionary idea was "pay for what you use." This On-Demand model killed the need for upfront capital expenditure (CapEx) on servers. However, as scale grew, the convenience premium became expensive.
Reserved Instances (2009)
To help customers lower bills, AWS introduced Reserved Instances (RIs). You committed to a specific instance type (e.g., m3.large) in a specific region (us-east-1) for a term. It saved money but was operationally painful. If you changed instance families, your RI was wasted.
Savings Plans (2019)
AWS solved the flexibility problem with Savings Plans. Instead of committing to *hardware*, you commit to *spend* (e.g., $5/hour). This automatically applies to any instance family, size, or region (for Compute SPs). This decoupled financial planning from engineering choices, creating the modern FinOps standard.
Common Misconceptions
- Myth: "I need to assign Savings Plans to servers." Reality: AWS applies them automatically to the usage that yields the highest discount.
- Myth: "Spot is always cheaper." Reality: Spot is cheaper, but SPs are guaranteed availability. Use SP for base load, Spot for fault-tolerant bursts.