Understanding CI/CD Costs

This guide explains how GitHub Actions billing works and how CICosts helps you understand and manage your costs.

GitHub Actions Billing Model

How GitHub Charges

GitHub Actions charges based on:

1. Minutes used - Actual runtime of your jobs
2. Runner type - Larger runners cost more per minute
3. Operating system - Windows and macOS cost more than Linux

Billing Calculation

Cost = Minutes × Rate per Minute

Where:

• Minutes are rounded up to the nearest minute
• Rate depends on runner size and OS

Runner Pricing

Runner	Linux	Windows	macOS
2-core (default)	$0.008/min	$0.016/min	N/A
4-core	$0.016/min	$0.032/min	$0.08/min
8-core	$0.032/min	$0.064/min	$0.12/min
16-core	$0.064/min	$0.128/min	$0.24/min
32-core	$0.128/min	$0.256/min	N/A
64-core	$0.256/min	$0.512/min	N/A

Free Minutes

GitHub includes free minutes with some plans:

Plan	Free Minutes (Linux)
Free	2,000 minutes
Pro	3,000 minutes
Team	3,000 minutes
Enterprise	50,000 minutes

Windows and macOS minutes are charged at higher multipliers:

• Windows: 2x multiplier
• macOS: 10x multiplier

What CICosts Tracks

Gross Costs

CICosts calculates gross costs - what you would pay without free minutes:

Gross Cost = All minutes × Runner rate

This helps you understand true resource consumption regardless of your plan.

Why Gross Costs?

1. Consistency - Compare costs across plans
2. Forecasting - Predict costs if you exceed free minutes
3. Optimization - Find waste even within free tier
4. Planning - Budget for growth

Mapping to Actual Bill

Your actual GitHub bill may be lower:

Actual Bill = max(0, Gross Cost - Free Minute Value)

CICosts shows both gross and estimated actual costs.

Understanding Your Spending

Where Money Goes

Typical cost distribution:

Category	Typical %
CI/Test workflows	40-50%
Build/Deploy	20-30%
Scheduled jobs	10-20%
Other	10-20%

The 80/20 Rule

Usually, 20% of workflows cause 80% of costs. CICosts helps you identify:

• Which workflows are expensive
• Why they're expensive
• How to optimize them

Cost Factors

Job Duration

Longer jobs cost more:

# This costs more
- run: npm test # 10 minutes = $0.08

# Than this
- run: npm test --parallel # 5 minutes = $0.04

Runner Size

Larger runners cost more per minute:

# Expensive
runs-on: ubuntu-latest-32-core  # $0.128/min

# Cheaper
runs-on: ubuntu-latest  # $0.008/min

But larger runners can reduce total time:

Scenario	Runner	Time	Cost
A	2-core	30 min	$0.24
B	8-core	8 min	$0.26
C	16-core	5 min	$0.32

Operating System

macOS is significantly more expensive:

# Most expensive
runs-on: macos-latest  # $0.08/min (4-core)

# Much cheaper
runs-on: ubuntu-latest  # $0.008/min (2-core)

Matrix Builds

Matrix builds multiply costs:

strategy:
  matrix:
    os: [ubuntu, windows, macos]  # 3 jobs
    node: [16, 18, 20]            # × 3 versions
                                   # = 9 total jobs

This 10-minute workflow costs:

• 1 Ubuntu job: $0.08
• 9 matrix jobs: $0.72+

Workflow Triggers

Trigger frequency affects costs:

Trigger	Typical Cost
push (main only)	Low
push (all branches)	Medium
pull_request	Medium
schedule (hourly)	High
workflow_dispatch	Varies

Reading CICosts Data

Dashboard Metrics

Metric	What It Means
Today	Cost so far today
vs Average	Comparison to your typical day
This Month	MTD spending
Projected	Estimated month-end total

Trend Analysis

Look for:

• Spikes - Unusual high-cost days
• Patterns - Weekly or monthly cycles
• Trends - Increasing or decreasing over time

Per-Workflow Analysis

For each workflow, check:

• Average cost per run
• Success rate (failed runs still cost)
• Run frequency
• Duration trend

Key Questions to Ask

1. Which workflows cost the most?

   • Check the "Top Workflows" view

2. Are we using appropriate runners?

   • Check runner type distribution

3. Are matrix builds necessary?

   • Compare matrix size vs. value

4. How much do failures cost?

   • Check failure rate and cost

5. Are scheduled jobs worth it?

   • Compare scheduled job costs to value

---

Next: Optimizing Workflows →