How much electricity does Bitcoin use each year, in terawatt-hours?

How to get close to this number

Compare it to the data centre figure on this site: a few hundred terawatt-hours for all computing infrastructure worldwide. A single application consuming a substantial fraction of that is the right scale, which puts it in the low hundreds.

The country comparison is the standard framing and the explanation gives it: comparable to the entire annual electricity use of Poland or Argentina. A single protocol matching a mid-sized national grid is the fact that makes this politically contentious rather than merely technical.

The design point in the explanation is the one usually missing from the argument: the energy is spent deliberately. The cost of computation is what secures the network — making an attack expensive is the entire mechanism — so the consumption is not inefficiency to be optimised away but the security model itself. Whether that is a reasonable trade is a separate question from whether it is a bug.