A running ASIC miner has no power button and no graceful shutdown sequence — it is built to hash continuously and it powers on the instant it sees line voltage. That design has two consequences every operator has to plan for: an abrupt power loss and a lost pool connection are not the same event, and the way you bring the machine back on is where most avoidable damage happens. This guide covers what actually occurs inside the unit during each, and the correct procedure for shutting down, riding out an outage, and re-energizing without cooking your gear.
What actually happens when the power drops
ASIC miners — Antminer, Whatsminer, Avalon — ship without a physical on/off switch. You start and stop them by making or breaking the AC feed at the PSU input, the PDU, or the breaker. A hard power loss is therefore a normal shutdown mechanically; the miner does not corrupt anything by losing power mid-hash. The real risks are thermal and electrical, and they land at two moments: the cool-down after shutdown and the re-energization when power returns.
Condensation is the slow killer
A full-size miner dumps 3,000–5,000+ W of heat. When it stops, hashboards and PSU internals drop from ~60–80°C toward ambient fast. In a humid space (an unheated garage, a shed, a shipping-container hashcenter) that temperature swing pulls the dew point across the boards and moisture condenses on bare copper, chip pads, and connector pins. Repeated wet/dry cycles corrode traces and pins over weeks — a classic cause of dead domains and intermittent hashboards that looks like a “random” failure months later.
- Keep airflow moving through the room, or keep the space climate-controlled, so surfaces don’t sit below the dew point after shutdown.
- If a miner has been off and cold in a humid environment, let it acclimate and run its fans before it comes under full load, rather than slamming it to 100% immediately.
Bringing miners back: inrush is the danger, not “surges”
The single most damaging habit is flipping one main breaker to restore an entire rack at once. Every PSU draws a large inrush current at the instant it energizes as its bulk capacitors charge. One machine is fine; a dozen switching on together can trip breakers, sag the feed, and stress PSU input stages. Restore power in a staggered sequence.
- Confirm utility power is stable and back for good — do not re-energize during a flapping brownout, when voltage is sagging and recovering repeatedly. Dirty, low voltage is harder on PSUs than a clean full outage. Wait for solid voltage.
- Power miners on one at a time, pausing a few seconds between each, so inrush and load ramp are spread out instead of stacking.
- Watch the first machines start their fans and boot before you continue down the rack.
- Give the whole rack a few minutes, then check each unit reached its pool and is hashing at expected rate — not stuck with dead boards or missing ASIC chip counts.
Circuit sizing matters here. Bitmain APW-series PSUs (APW9/APW9+/APW8, APW3) are rated for full output at 200–240 V. They will run on a 110/120 V circuit, but at reduced output — and a single 15 A/120 V branch cannot supply the full draw of a large miner regardless. Size the circuit to the machine’s real wattage at 200–240 V, and never try to bring a whole hashcenter back on one undersized circuit.
Network disconnection is a different problem
Losing the pool connection is not a power event and it does not shut the miner off. When a miner can’t reach its pool it keeps running, keeps its fans and control board powered while the hashboards idle, and retries the primary pool (then any configured backups) on a loop. In other words: it still draws power — fans, control board, idling boards — while producing zero revenue. It does not “shut down for protection” after a set number of retries — that is a myth. A control-board watchdog may reboot the controller if the firmware itself hangs, but a plain loss of internet keeps the machine online and retrying into a dead connection indefinitely.
So a network outage costs you money the whole time it lasts, with no revenue to show for it. The decision to power down during a network-only outage is economic, not mechanical:
- Brief blip (router reboot, ISP hiccup): leave it running. Reconnecting a booted, warm miner is instant and avoids an unnecessary cool-down/condensation cycle and another inrush event.
- Extended outage with no ETA: the machine is just drawing power (fans, control board, idle boards) for nothing. Weigh the electricity you’re burning against the minor wear of a shutdown and restart, plus the condensation risk in your environment. If the outage will clearly run long and power is expensive, cut it. When you do, cut the power at the PSU/PDU — there is no “disconnect the network first” step that protects anything.
Before assuming the internet is down, rule out the miner itself: a unit that dropped its pool because of a failing hashboard, a bad PSU, or a config problem looks identical to a network outage from the dashboard. If a single miner is offline while its neighbors are fine, it’s the miner, not the network — work it as a hardware fault.
The one thing you must never do
Never open a PSU shell to “check” it after an outage. The lethal-voltage hazard in a miner lives inside the sealed PSU: the mains-side and PFC bulk capacitors hold a dangerous charge long after the plug is pulled. The DC output side (the 12–21 V bus feeding the hashboards) is a low-voltage rail — the danger there is not shock but the enormous available current, which will happily vaporize a dropped tool or ring and throw molten metal. Treat the DC bus with respect for its current, and treat the PSU interior as off-limits.
How to confirm the recovery worked
- Each miner’s web dashboard shows it connected to the pool (accepted shares climbing, not just “connecting”).
- Hashrate is at spec for the model and every hashboard reports its full ASIC chip count — a board that came back with missing chips or a dead voltage domain took damage or has a seated-connector issue.
- Board and chip temperatures are climbing normally and fans are ramping under PID control, with no thermal-protection or over-temp warnings in the log.
- No repeated controller reboots in the kernel/system log, which would point to a watchdog tripping on a firmware or hardware fault rather than a clean recovery.
Quick reference
- Shutting down: cut AC at the PSU/PDU input; there’s no switch and none is needed.
- Humidity: keep air moving so cooling boards don’t cross the dew point — condensation is the long-term corrosion risk.
- Restoring power: wait for stable voltage, then power up one machine at a time; never flip a whole rack on one breaker.
- Network down: the miner keeps drawing power (fans, control, idle boards) for nothing — power it down only if the outage is clearly long. It won’t self-protect by shutting off.
- Safety: never open the PSU; the DC bus danger is current, not voltage.
Related: If a hashboard came back with missing chips or a dead domain after an outage, walk it through the ASIC fault finder, source what you need from ASIC repair parts, or start a repair with us. Model-specific power and startup details are in the miner manuals.
