Mini PC Won't Turn On After Power Outage: The BIOS Fix

Power blips at 4 a.m. Three minutes later the lights are back, your switch is back, your router is back — and your mini PC homelab is still dark. You’re SSHing into nothing because the box never restarted.

The fix is almost always a single BIOS setting that ships disabled on consumer mini PCs. Vendors default to off because their typical customer wants to be present when the PC boots. For a headless 24/7 homelab, that default is exactly wrong.

This guide covers the setting name on every common mini PC vendor, two failure modes the setting alone doesn’t solve, and how to confirm it works without standing next to the breaker box.

The setting you’re looking for

It has different names on different boards but it does the same thing: tells the firmware what to do when AC power is restored after being cut.

The three values you’ll see, in some combination:

1. Power Off (or “Stay Off” / “Last State Off”) — default on almost every consumer mini PC. After AC returns, board stays off until you press the button.
2. Power On (or “Always On” / “Last State On”) — what you want for a homelab box. After AC returns, board powers on automatically.
3. Last State — restores whatever state the machine was in when power was lost. If it was running, it boots; if it was off, it stays off. Acceptable, but Power On is more predictable for a server.

Pick Power On. “Last State” sounds clever but it bites you the first time you shut the box down to add RAM, forget to power it back on, and then lose power three weeks later — it stays off because that was its last state.

Where to find it, vendor by vendor

The wording is inconsistent. These are the names that appear on current shipping hardware, per each vendor’s own BIOS documentation.

Intel NUC (and NUC-clones)

BIOS menu: Power → Secondary Power Settings → After Power Failure. Set to Power On. Intel documents this in their NUC BIOS support article (source). The same submenu exists on most ASUS-branded NUCs that inherited the platform after Intel exited the NUC business.

Beelink, Trigkey, GMKtec, Minisforum (N100/N305/Ryzen mini PCs)

These typically use AMI or Insyde BIOS. Look under Chipset → PCH-IO Configuration → State After G3 or Advanced → APM Configuration → Restore AC Power Loss. Set to S0 State or Power On (wording varies — “S0” is the ACPI sleep-state name for “fully on”). On a few Minisforum units it’s hidden under Advanced → ACPI Configuration.

If you can’t find it in the user-facing BIOS, the option sometimes exists but is hidden — pressing F7 or Ctrl+F1 at the BIOS main screen will sometimes unlock an advanced view on AMI BIOSes. Use that as a last resort.

Dell Wyse 5070 and other Dell thin clients

BIOS menu: Power Management → AC Recovery. Set to Power On. Dell’s admin guide for the 5070 documents this directly in the BIOS settings reference (source). This is the single most-forgotten setting on a Wyse 5070 homelab build — set it before you even install the OS.

HP T620 / T630 / T640 thin clients

BIOS menu: Advanced → Built-In Device Options → After Power Loss (older T620), or Power → Hardware Power Management → AC Recovery (T640). Set to Power On. HP’s terminology shifted between generations — see the HP thin client homelab notes for which model has which menu.

ASUS-branded mini PCs and PN-series

BIOS menu: Advanced → APM Configuration → Restore AC Power Loss. Set to Power On. ASUS publishes a support note for this on their consumer boards (source) and the same setting carries over to their mini PC line.

Lenovo ThinkCentre Tiny / M-series

BIOS menu: Power → After Power Loss (newer units) or Power → Automatic Power On → After Power Loss (older units). Set to Power On.

If your vendor isn’t listed, the search terms that find the right page in any BIOS are: “AC Recovery”, “Restore AC Power Loss”, “After Power Failure”, “After Power Loss”, “State After G3”, or “AC Power Recovery”. One of those phrasings will be there.

Two failure modes the BIOS setting alone doesn’t fix

Setting AC Recovery → Power On handles the common case. There are two scenarios where it isn’t enough.

1. Brownouts and rapid power cycles

If AC drops for less than the firmware considers a real outage — or if power flickers off and back on inside a second or two — some BIOSes treat it as noise and don’t re-trigger the auto-power-on. The box ends up off even though “Power On” is configured.

The fix is upstream of the BIOS: put the mini PC behind a small UPS that smooths brownouts into either a clean ride-through or a clean shutdown. A 600–900VA line-interactive UPS handles the typical homelab mini PC fleet easily and costs $80–$120. The math on what your box actually draws and how long a UPS will hold it is covered in the mini PC power consumption post — most 10W boxes get 30+ minutes of runtime on a small UPS, which is more than enough to survive any blip and gracefully shut down for a real outage. Which unit to buy, and why VA is the wrong number to shop on at these loads, is worked through in UPS sizing for a mini PC homelab.

2. The box boots but the OS doesn’t come up

A second silent-failure mode: the hardware powers on after AC returns, but the boot path stalls. Common culprits:

• GRUB waiting for input because the last boot was unclean. Set GRUB_RECORDFAIL_TIMEOUT=5 in /etc/default/grub and run update-grub on Debian/Ubuntu.
• A network-mount blocking systemd boot. An NFS or CIFS mount in /etc/fstab without nofail,x-systemd.device-timeout=10s will hang boot for minutes if the file server isn’t up yet. Add those options.
• A docker-compose stack with hard dependencies. Use restart: unless-stopped on every service — let containers crashloop until their database is up rather than relying on boot order.

The pattern: the box powered on fine, but you can’t tell because you can’t SSH in. The next section covers how to catch this before it happens at 4 a.m.

How to test without unplugging anything

Don’t yank the cord — that’s a real outage simulation but it’s hard on disks and routers. Three softer tests, in increasing realism:

1. Issue a clean shutdown then verify auto-on works at all. sudo systemctl poweroff from SSH, wait until the box is off, then physically toggle the wall switch off and back on. The box should come up on its own. If it doesn’t, the BIOS setting didn’t take — re-check it.

2. Flip the breaker for the outlet. Closer to a real outage. Lets you confirm the UPS rides through (if you have one) and that the box auto-restarts when AC actually drops.

3. Pull the kettle plug from the back of the PSU brick. Realistic and reversible. Wait 30 seconds, plug it back in. The box should come on within a few seconds.

If the box powers on but doesn’t come back to a usable SSH state within ~3 minutes, you have an OS-side boot problem. Connect a monitor and keyboard, reboot once more, and watch what happens — you’ll see GRUB, fsck, a hung systemd unit, or a network-mount stall right on screen.

A related setting worth turning on while you’re in there

Wake-on-LAN is the other thing every homelab box should have configured, for the reverse situation: deliberately powering the box off and waking it later without standing next to it.

In BIOS, look for Wake on LAN, Power On by PCIe, or Resume on LAN and enable it. In Linux, confirm the NIC supports it and that it’s armed for magic-packet wakeup with ethtool -s <iface> wol g (the wol g argument means “wake on magic packet”). The ethtool man page documents the full WOL flag set (source) — g is the one almost everyone wants.

Test it from another box on the same LAN: wakeonlan <mac> (install the wakeonlan package on Debian/Ubuntu). If the box wakes, you’re done. If it doesn’t, check that the BIOS option is actually enabled and that ethtool <iface> reports Wake-on: g rather than Wake-on: d (disabled).

TL;DR

• The setting is AC Recovery → Power On (or whatever your vendor calls “After Power Failure”). Set it before you install the OS, not after the first outage.
• Add a small UPS to smooth brownouts that confuse the BIOS into not re-triggering auto-on.
• Test with a kettle-cord pull, not just a clean reboot. Auto-power-on and clean shutdown are different code paths.
• Fix the boot-path stalls that prevent SSH from coming back: GRUB recordfail, dirty-fsck pauses, blocking NFS/CIFS mounts.
• Enable Wake-on-LAN in BIOS and arm it with ethtool while you’re already in there.

Most “my homelab didn’t come back from the outage” stories trace to one of two things: this BIOS setting was never enabled, or the OS booted but something on the boot path stalled. Both are fixable in ten minutes. Worth doing before you actually need it.

For sizing the UPS you put in front of your fleet, the power consumption math post covers how to translate idle wattage into UPS runtime; if the fleet is still on paper, how to build a low power homelab works the same numbers forward from a watt budget instead. And if you’re still picking the box itself, the N100 buying guide calls out which units expose the AC Recovery option in their stock BIOS without needing an advanced-menu unlock, and N150 vs N100 covers what the 2026 Twin Lake refresh changed if you are buying new.

While the machine is open in front of you, this is not the only firmware default that suits a desktop rather than a server. Wake-on-LAN, the virtualization extensions, boot mode, the C-state limit that decides idle draw, and the radios a headless box never uses are all worth changing in the same sitting — the full pass is in mini PC BIOS settings for a 24/7 homelab.

Related across the network

• MPPT vs PWM Charge Controllers for Homelab Solar — solarhomelabs.com