toolbox
Home / Startups & Digital Business / Spot Instance Cost Forecast Estimator

Spot Instance Cost Forecast Estimator

The Spot Instance Cost Forecast Estimator projects spend for a spot-compute fleet across a selected number of months. It combines the expected instance count, average spot price, monthly runtime, and retry overhead, then adds optional fixed monthly costs that belong to the same workload. The forecast is useful for budgeting batch, analytics, rendering, and other interruption-tolerant workloads where the spot price itself is only one component of expected spend.

Because spot prices and availability can change, the calculator uses the average rate you enter rather than presenting a fixed provider price. For forward planning, use a conservative rate derived from historical billing or an internal budget assumption. The retry-overhead field lets you model extra billable hours caused by interruptions without pretending that every purchased hour is productive. Review several scenarios rather than relying on a single point estimate when spot volatility is material.

Inputs

units
$/hr
hr
%
$
mo
Result
—
Forecast spot workload cost
Monthly spot compute—
Monthly total—
Monthly retry overhead cost—
Other costs over period—

1. Enter average fleet size
Use the average number of concurrently running spot instances expected during the modeled month.

2. Set an average spot rate
Use a rate that reflects the instance pools and regions you intend to use, preferably from historical effective pricing.

3. Enter runtime per instance
Estimate the billed runtime for one average fleet instance during a month before retry overhead.

4. Add retry overhead
Express extra billable runtime from interruptions as a percentage of base runtime.

5. Include other monthly costs
Add workload-specific fixed costs such as orchestration or storage only when they belong in the same forecast.

6. Choose the forecast horizon
The calculator holds the entered assumptions constant across the selected months.

Base monthly spot cost = Instances × Runtime per instance × Spot hourly rateRetry overhead cost = Base monthly spot cost × Retry overhead rateMonthly total = Base monthly spot cost + Retry overhead cost + Other monthly costsForecast cost = Monthly total × Forecast months

Where:

  • Instances — average spot fleet size
  • Runtime per instance — base billed hours per instance per month
  • Spot hourly rate — average cost per instance-hour
  • Retry overhead rate — extra compute percentage attributed to interruptions
  • Other monthly costs — fixed workload costs added each month
  • Forecast months — number of months projected

Assumptions: Fleet size, rates, runtime, retry overhead, and other monthly costs remain constant through the forecast. The model does not predict future spot-market prices.

What the result means

The main result is the estimated total workload cost over the selected forecast horizon under the entered average spot assumptions.

For volatile workloads or markets, run low, base, and high scenarios using different rates, fleet sizes, and interruption overhead.

Given

  • 25 average spot instances
  • $0.19 average hourly rate
  • 480 base hours per instance per month
  • 10% retry overhead
  • $220 other monthly costs
  • 4-month forecast

Calculation
Base monthly spot cost = 25 × 480 × $0.19 = $2,280
Retry overhead cost = $2,280 × 10% = $228
Monthly total = $2,280 + $228 + $220 = $2,728
Forecast cost = $2,728 × 4 = $10,912

Result
$10,912 forecast spot workload cost

The model budgets $2,728 per month and $10,912 across four months under the stated spot and interruption assumptions.

Does this forecast predict future spot prices?

No. It applies the average hourly rate you enter. Use historical data and scenario ranges if future spot prices are uncertain.

Should retry overhead be added to runtime instead of cost?

Either approach can be mathematically equivalent when all retry hours use the same average rate. This calculator expresses the extra runtime as a percentage of base compute cost for simplicity.

What belongs in other monthly costs?

Include only recurring costs directly associated with this spot workload, such as checkpoint storage or orchestration, if you want them in the forecast.

How do I model a fleet that grows every month?

This page holds fleet size constant. For a growing fleet, run separate periods or use a month-by-month planning model.

Why compare this with an on-demand forecast?

The comparison reveals whether the lower spot rate still delivers meaningful savings after interruptions and supporting costs are included.