TL;DR: Bitcoin uses ~173 TWh/yr (Cambridge CBECI, 2025), about 0.75% of global electricity. Real and material. The honest case is not that the number is small; it is that the energy buys something (security), the demand is uniquely flexible, and the loudest critiques split cleanly into the ones with answers and the ones without.
The comparison trick. "More than Argentina," "less than tumble dryers," "more than gold mining". Every one is a chosen denominator. Gold mining's own estimates run 132-265 TWh depending on scope, a range that straddles Bitcoin's number. Whether Bitcoin "uses more than gold" is decided entirely by how you count gold.
The energy is the security. Proof-of-work converts electricity into the cost of rewriting history. Cut the energy and you cut the attack cost by the same proportion. This is why "just make it use less" is not the free lunch it sounds like.
What miners actually buy. A miner is the ONLY industrial-scale load that is location-agnostic, interruptible in seconds, and indifferent to when the power flows. So it migrates to stranded hydro, flared methane, curtailed wind: the cheapest electrons on earth because nobody else can use them.
The receipt. August 2023, ERCOT heat wave: Riot Platforms earned $31.7M mostly powering DOWN (power resold + demand-response credits) versus $8.9M of bitcoin mined that month, curtailing 95%+ of load in seconds. SEC-filed, public.
What the critics get right. The Granbury TX noise lawsuit is real. Not all mining runs on stranded renewables, some grids burn coal and gas. Water and e-waste footprints deserve straight answers. The strong forms of the critique have answers; the weak forms get called FUD because they do not. Naming the real ones is the point.
Full chapter with sources:
https://www.learnbitcoin.com/rabbit-hole/energy