[Intermediate] — the numbers, what the Merge actually changed, and the criticisms that survive it.

The proof-of-work numbers

Bitcoin’s electricity consumption is on the order of 100–150 TWh per year — comparable to a mid-sized country. Estimates vary by methodology (the Cambridge CBECI and Digiconomist differ meaningfully), but the order of magnitude isn’t disputed.

The uncomfortable part is that this is by design, not inefficiency.

The energy expenditure is the security. An attacker must out-spend the honest network, so the cost of attack is proportional to the ongoing cost of operation. You cannot make proof-of-work cheap without making it weak → consensus.

There is no optimisation available. Better hardware raises hashrate, difficulty adjusts, and consumption returns to whatever the block reward supports. The energy use is a function of the reward, not of efficiency.

The Merge — what actually changed

Ethereum’s move to proof-of-stake in September 2022 cut its energy use by roughly 99.95%, from tens of TWh per year to a few GWh — comparable to a large office building rather than a country.

This is the single largest emissions reduction of any computing platform, and it happened in one day. It’s also the strongest evidence against the claim that this energy use is inherent to blockchains: it isn’t, it’s inherent to proof-of-work, and one major chain simply stopped.

Which reframes the debate. Most remaining energy use is Bitcoin, which will not change — its conservatism is deliberate, and its community regards PoW as essential rather than incidental → Bitcoin.

The arguments, weighed

“It uses renewables.” Partly true — a substantial and growing share, since miners chase cheap power and stranded renewables are cheap. The counter is real too: demand that can be sited anywhere still competes for clean generation that would otherwise displace fossil fuels elsewhere. Using renewable energy isn’t the same as being carbon neutral when the grid as a whole is not.

“It monetises stranded and flared energy.” Genuinely true in specific cases: flared gas at oil wells, curtailed hydro. A real efficiency gain where it happens, and a small share of total consumption.

“It’s grid-stabilising.” Miners can shut down within seconds, making them useful demand response — demonstrated in ERCOT (Texas). Real, and it also means miners are paid to not compute, which is a strange thing to defend as efficiency.

“Banking uses more energy.” The comparison is usually constructed unfavourably (including bank branches and ATMs against Bitcoin’s mining alone) and, more importantly, it’s a comparison of systems with vastly different transaction volumes. Per-transaction figures are also misleading in the other direction, since energy tracks the block reward, not transaction count. The honest framing is: Bitcoin consumes a country’s worth of electricity to secure a settlement network, and whether that’s worth it depends on how much you value what it settles.

“E-waste.” Less discussed and real: ASICs are single-purpose and obsolete within a few years, producing tens of thousands of tonnes annually. Unlike energy, this has no renewable offset.

The other externalities

Wealth concentration. Early adopters hold disproportionate supply, and proof-of-stake compounds it structurally — stake earns stake, so relative holdings grow without any productive activity. Whether this is worse than existing financial concentration is arguable; that it’s a redistribution upward is not.

Fraud harm. Billions annually, concentrated on the least sophisticated participants. Pig butchering operations involve human trafficking — a serious harm the field rarely counts as its own → the fraud surface.

Ransomware. Crypto did not create ransomware, but pseudonymous, irreversible cross-border payment made it dramatically more scalable. This is a genuine causal contribution to a real and growing harm, and it is the criticism the field engages with least honestly.

Opportunity cost. A great deal of excellent engineering talent has gone into financial infrastructure for trading tokens. This is a legitimate criticism that doesn’t depend on any technical claim being wrong.

Holding it honestly

Both of these are true:

The energy criticism has been substantially answered for proof-of-stake chains. Ethereum uses negligible energy, and continuing to cite pre-2022 figures for it is simply out of date. Most new chains are PoS.

It has not been answered for Bitcoin, and won’t be. The consumption is deliberate, and defences that begin “but banking…” are deflections. The honest position is that Bitcoin’s energy use is real, large, permanent, and a cost you are trading against whatever value you place on a censorship-resistant settlement network.

And the non-energy externalities are the ones with the weakest defences — fraud harm, ransomware enablement and wealth concentration are less discussed than energy and, arguably, matter more.

Key insight

“Blockchains waste energy” was always really “proof-of-work costs energy, and that cost is the security.” Ethereum proved the criticism was contingent by eliminating 99.95% of its consumption overnight; Bitcoin demonstrates it’s permanent where PoW is retained. The field’s remaining hard externalities aren’t environmental — they’re the fraud, the ransomware and the concentration, which have no equivalent technical fix waiting.

Source: [reference] — figures approximate, Aug 2026. Cambridge CBECI is the standard reference for Bitcoin consumption.

↑ Back to The Honest Assessment