If you take one idea from this site, make it this one: idle watts decide your homelab electricity bill, not peak load. Spec sheets and product pages love to quote the big load number, but that is almost never what you pay for.
The short answer: always-on homelab gear spends roughly 95% of its life at idle, so the idle figure dominates the annual cost. Estimating from the peak load number can overstate your real bill by 2-4x. For anything running 24/7, idle is the number that matters.
Why idle dominates a 24/7 bill
A homelab device that is “on” is mostly waiting. A Proxmox host serving a handful of containers, a NAS holding backups, a firewall routing a home connection — these spend the overwhelming majority of every day near idle, with brief bursts of activity. Electricity cost is energy over time, so the state a machine sits in for 23+ hours a day is the state that bills you.
That is also why our methodology and the device cost pages lead with a draw weighted toward idle rather than the headline load figure.
A worked example: the same box, two estimates
Take an Intel N100 mini-PC from our wattage database: it idles around 7 W and peaks near 22 W under full load. Running 24/7 at the US average of $0.17/kWh:
| Estimate basis | Watts used | US $/yr (0.17) | Germany €/yr (0.40) |
|---|---|---|---|
| Idle (realistic for 24/7) | 7 W | $10.42 | €24.53 |
| Peak load (worst case) | 22 W | $32.76 | €77.09 |
| Difference | 15 W | $22.34 | €52.56 |
Source: HomelabWatts, data as of 2026-06-14. Check any device in the calculator.
Using the load number triples the estimate. The mini-PC almost never runs at 22 W for a full day, so the $10/year idle figure is far closer to the truth. Get this wrong on a rack server and the error is in real money — a 110 W idle / 250 W load server is the difference between ~$164/year and ~$372/year.
When load actually matters
Idle isn’t always the right number. Use load (or measure carefully) when hardware is genuinely busy most of the day:
- A GPU transcoding Plex/Jellyfin continuously or running local AI inference for hours.
- A CPU under sustained render, compile or simulation load.
- A NAS during long scrubs, resilvers or nightly backup windows — though even then it returns to idle most of the day.
For everything else — routers, switches, NAS boxes, light virtualization hosts — idle is the figure that decides the bill. See how much it costs to run a NAS 24/7 for storage specifically.
The most honest number: measured average
The cleanest way to sidestep the idle-vs-load debate is to measure accumulated energy rather than a snapshot of watts:
- Leave a plug-in meter or smart plug on the device for a week.
- Read the kWh total, divide by the hours, and you get the true average wattage — idle, load and everything in between, automatically.
That measured average is what you should feed into the cost calculator with your local kWh price. If you can’t measure, a 70% idle / 30% load weighting is a reasonable stand-in for typical always-on gear. For the full how-to, see measuring your homelab’s real power draw.
The bottom line
Peak load is a marketing number; idle is your bill. Estimate 24/7 cost from idle (or a measured average), reserve the load figure for genuinely busy hardware, and you’ll stop over-worrying about gear that costs a few dollars a year — and start spotting the one idle-heavy box that quietly dominates your power draw.