Local vs API cost calculator

Whether local AI is cheaper depends entirely on how much you use it. Enter your figures below. This tool is meant to be genuinely useful to someone deciding against local AI as much as for it, so it will happily tell you when the API is the cheaper option.

The local machine

Start from a reference build:

Hardware cost

What you would pay for the machine or GPU.Power draw under load

Hours of active use per day

hrs

Electricity price

£/kWh

Defaults to a rough UK average. Use your own tariff.

Your usage

Input tokens per month

Output tokens per month

The API you would compare against

Input price

$/M tok

Output price

$/M tok

USD to GBP rate

Used to convert the API's USD pricing into pounds.

Local, per month to run£13.13plus £1,600 upfront

API, per month£82.95

Local breaks even after about 23 months, after which it costs less than the API. Whether that is soon enough to be worth it is your call.

Cumulative cost over time

After	Local total	API total	Difference
1 year	£1,758	£995	cost £762
2 years	£1,915	£1,991	save £75.61
3 years	£2,073	£2,986	save £913

What this calculation assumes, honestly

This ignores the value of your own time, which is often the largest real cost of running locally. Setup, maintenance, and troubleshooting are not free.
Hardware keeps resale value and can be used for other things. This calculation treats the full purchase price as sunk, which overstates the true cost of the local option.
API pricing has trended downward over time and may keep falling. A break-even far in the future is less reliable than one soon, because the comparison itself will move.
Local and cloud models are not always capability-equivalent. A cheaper local setup that cannot do the job is not actually cheaper.
Electricity, power draw, and usage are estimates. Real figures vary with your tariff, your hardware, and how hard you actually push it.
API rates are entered in USD per million tokens and converted at a fixed rate you can edit. Exchange rates move; treat the converted figures as indicative.

Is running AI locally cheaper than an API?

It depends entirely on how much you use it. The calculator compares the upfront hardware cost plus running electricity against the per-token price of an API, and reports a break-even point. The more you use a model, the sooner local hardware pays for itself; at low volumes, an API is usually cheaper and the hardware may never break even.

The honest caveats matter and are built into the tool: the calculation ignores your time, which is often the largest real cost; hardware keeps resale value and can be used for other things; API prices have trended down over time; and a local model is not always capability-equivalent to a frontier API. The defaults are deliberately neutral, so the tool will tell you plainly when the API is the cheaper option.