How much does running a local LLM add to my power bill?
Running a model draws real power, but on most bills the surprise is not generation — it is the machine idling around the clock waiting to be used.
The short answer
Computed from this tool’s default settings — power draw under load and the rest as most people start. Change them below for your own case.
About $6.24 a month, of which $3.83 is actual generation. Modest next to almost any API bill at the same volume.
LLM Electricity Cost Calculator
Monthly and yearly electricity, from power draw and hours.
About $6.24 a month, of which $3.83 is actual generation. Modest next to almost any API bill at the same volume.
What each setting changes
Every input moves the result for a reason. This is what each one does and where to find the value for your own machine.
| Setting | Default | What it changes |
|---|---|---|
| Power draw under load | 350 W | Board power while generating, not idle. Card TDP is a good starting point. |
| Electricity price | 0.18 per kWh | Your actual tariff. US average is near $0.17, UK near £0.25, India near ₹8. |
| Hours generating per day | 2 hours | Anywhere from 1 to 24 hours. |
| Idle draw | 20 W | A machine left on all day draws this for the other hours. |
Real answers across hours generating per day
The same calculation run at a range of settings, with everything else left at its default. These are computed by the tool itself, not written by hand.
| Hours generating per day | Per month | Per day | Per year | Generating |
|---|---|---|---|---|
| 1 | $4.43 | $0.15 | $53 | $1.92/mo |
| 7 | $15.27 | $0.50 | $183 | $13.41/mo |
| 13 | $26.10 | $0.86 | $313 | $24.90/mo |
| 18 | $35.13 | $1.16 | $422 | $34.47/mo |
| 24 | $45.96 | $1.51 | $552 | $45.96/mo |
Idling costs more than generating here — $2.52 against $1.92 a month. Sleeping the machine between sessions saves more than any amount of tuning.
How this is calculated
No lookup tables and no invented constants. Here is the arithmetic, so you can check it against your own numbers.
Energy is watts × hours ÷ 1000, priced per kilowatt-hour. The number people get wrong is idle: a workstation left on around the clock spends most of its year idling, and that can exceed the generating cost outright.
These are well-founded engineering estimates, not benchmark results. Your quantisation, runtime and context length all move the real number, and usable memory is an assumption rather than a specification. See the full methodology for every assumption behind these figures.
Who this is for
The three situations that bring people to this calculation.
Budgeting
Put a number on what local AI adds to the bill.
Always-on servers
See what idle actually costs over a year.
Comparing hardware
Weigh a power-hungry card against an efficient one.
How to use this calculator
Four steps, no account, nothing leaves your browser.
Set power draw under load
Start at the top of the panel. Every figure recalculates as you change it — there is no submit button, because watching the number move is the point.
Adjust the rest to match your setup
3 further settings: electricity price, hours generating per day, idle draw. Defaults are the common case, so change only what differs for you.
Read the headline, then the table
The large figure answers the question. The table underneath shows how the answer changes across nearby settings, which is usually where the decision actually gets made.
Check it against the method
The arithmetic is written out above. If a number looks wrong for your hardware, the assumptions are the first place to look — usable memory and quantisation are the two that vary most.
How much does running a local LLM add to my power bill: common questions
The questions people ask about this, answered without hedging.
How much electricity does a local LLM use?
A 350W card generating for two hours a day is under a kilowatt-hour daily — usually a few dollars a month. The generating cost is rarely the problem.
Does idle power really matter?
Often more than generation. A machine left on 24/7 spends most of the year idle, and at 20–60W that adds up to more than short bursts of full load.
How do I reduce it?
Sleep the machine between sessions, and consider a power limit — capping a card at 70% of its board power typically costs only a few percent of inference speed, because inference is memory bound rather than compute bound.
Is a Mac cheaper to run?
Considerably. Apple Silicon draws tens of watts where a discrete card draws hundreds, which is why Macs win on energy per token even when they lose on raw speed.
Related local AI tools
All 50 run on the same arithmetic, so answers across them agree.
More on cost
Will it fit
Context
Speed
Quantisation
Runtime & errors
Fine-tuning
Searches this page answers
Different ways of asking the same question, all resolved above.
Now find the models that fit
Sizing is only half the problem. Model Radar takes your hardware and shows which models actually run on it, ranked by what they are good at — the same arithmetic as this page, applied to every model worth running.