Counting the Real Cost of a Four-Hour Outage
A worked example using a 64-unit fleet, KEPCO time-of-use rates, and an unplanned hashboard swap at the wrong hour.
Operators often ask us to quantify the difference between a planned four-hour maintenance window and an unplanned failure. Here is a real calculation from a Changwon facility we audited in early 2025.
The Setup
- Fleet: 64 Antminer S19 Pro units, each rated at 0.11 PH/s
- Power contract: KEPCO industrial time-of-use, Gyeongsangnam-do
- Peak rate: ₩142/kWh (weekdays 10:00–12:00, 13:00–17:00)
- Off-peak rate: ₩78/kWh (weekdays 23:00–09:00, weekends all day)
- Average revenue: ₩0.018 per TH/s per hour (network conditions at time of audit)
Scenario A — Planned Window
Four units taken offline for hashboard inspection during Sunday 01:00–05:00.
- Lost hashrate: 4 × 110 TH/s = 0.44 PH/s for 4 hours
- Revenue loss: 0.44 × 4 × ₩0.018 × 1000 = ₩31.68 (simplified TH-to-revenue conversion)
- Electricity saved during downtime: negligible for four units
- Labor: one technician, 4 hours at standard rate (no overtime premium)
Total estimated cost: ₩31.68 revenue loss + standard labor
Scenario B — Unplanned Failure
Same four units fail at 14:30 on a Wednesday due to undetected thermal paste degradation.
- Response time: 90 minutes before technician arrives
- Repair time: 3.5 hours (hashboard swap + thermal paste reapplication)
- Total offline: 5 hours during peak-rate period
- Revenue loss: 4 × 110 TH/s × 5 hours × peak-adjusted rate ≈ ₩52
- Emergency labor: 1.5× overtime rate for afternoon dispatch
- Adjacent units thermally stressed, reducing fleet efficiency by ~2% for 6 hours
Total estimated cost: ₩52 revenue loss + overtime labor + ~₩18 thermal cascade loss
The Difference
The unplanned failure cost roughly 2.5× more than the planned window for the same four units and the same repair task. Scale that across a fleet experiencing three unplanned events per month and the annual difference exceeds ₩2,000,000 in a 64-unit operation.
Takeaway
The repair itself takes the same time either way. The cost difference comes from when it happens and whether adjacent units compensate for the offline hashrate. Maintenance windows exist to move Scenario B into Scenario A.
If you do not know your per-unit downtime cost, start with your power bill and pool payout records. We build these calculations into every audit report.