Electricity is the one commodity we still mostly can’t store cheaply, and moving it long distances is expensive. We have to build grids for the peak moment of the peak day, and renewables produce on nature’s schedule, not ours. So every grid wastes energy constantly. Wind blows hard at 3am when demand is lowest. A hydro dam in a wet season, far from any city, spills water past its turbines. Oil wells flare gas [burn it off at the wellhead, because piping it anywhere isn’t worth the cost]. In the UK we pay wind farms to switch off when the grid can’t absorb what they produce, then pay other plants to make up the shortfall. Managing those constraints cost about £1.7 billion in 2024/25, much of it driven by wind, and all of it on everyone’s bills. Stranded power [energy with no profitable route to demand, because of where or when it’s produced] is everywhere, every day.
Why miners hunt the energy nobody wants
A bitcoin miner’s product is digital and sells at one global price no matter where it’s made. The main input is electricity. The hardware fits in a shipping container, can sit next to a dam in Malawi or a wind farm in Texas, and can switch off in seconds without spoiling anything. No other industrial load works like that. A smelter can’t follow cheap power around, and a data centre can’t go dark every evening.
So competition drives miners toward the cheapest watt on earth. And the cheapest watt on earth is almost always the one nobody else can use, the 3am wind, the spilled hydro, the flared gas. Miners don’t go there because they’re green. They go there because they’re greedy, and the greed points in the right direction.
What the energy buys
Is bitcoin’s energy use a bug or a feature? It’s a feature. The energy cost is the mechanism itself, not a side effect. Bitcoin spends energy the way a vault spends steel.
Bitcoin’s record is secured by proof of work [miners burn electricity racing to solve a puzzle, and the winner earns the right to add the next block of transactions]. To rewrite that history, you’d have to redo the work against the entire network, continuously, for as long as you wanted your version to stand. So the cost of attacking bitcoin is physical and ongoing, rather than legal or political. That’s the design. Changing the record means outspending the rest of the world’s miners on energy. The rules are guarded by something even cheaper. Everyone running the software checks every block against the rules and discards any block that breaks them, so a miner who spends a fortune producing an invalid block has bought an expensive chain nobody else accepts. The energy makes attacking the record expensive. The refusal to accept a bad block makes attacking the rules futile.
More energy doesn’t create more bitcoin. The difficulty adjustment [an automatic rule that keeps blocks arriving roughly every ten minutes by making the puzzle harder as more computing power joins] means that when energy floods in, the supply schedule doesn’t move an inch. Extra energy buys extra certainty that the settled record stands, not extra coins. The energy also anchors issuance itself. Every coin that exists cost someone a real electricity bill, so nobody can conjure supply by decree. That’s what “bounded by energy” means. The money is tied to physics instead of to policy.
The cost is the security. Make the work cheap and rewriting the record gets cheap with it. The main alternative, proof of stake [where the right to write the ledger goes with how many coins you lock up], is cheaper because it drops the outside cost. An attacker on bitcoin has to keep buying electricity nobody will refund. An attacker on proof of stake buys influence once, in the coin itself, and the largest holders hold it already. That recreates money governed by whoever already has the most of it. That’s the system we’re trying to leave, rebuilt with new words.
The word “waste” assumes the conclusion. Energy spent on something people value isn’t waste. We don’t call tumble dryers or Christmas lights waste, because we’ve agreed the output matters. Miners pay real money for power and survive only if the market values what they produce. A miner running on expensive electricity goes bankrupt, no slogans required, and in downturns some do. So the fact that bitcoin uses energy settles nothing. Everything does. The question is whether monetary rules that nobody can cheat are worth paying for.
Buyer of last resort, and what it builds
Mining is the strangest electricity customer ever invented, and that strangeness is what lets it fund new power generation and steady the grid. A buyer of last resort takes the supply nobody else wants, which puts a floor under a market. That’s the role mining plays for electricity, and it changes what gets built.
A power project lives or dies on its worst hours. When a wind farm can only sell power during the hours the grid happens to want it, a big slice of its output earns nothing from a customer, and the people financing it charge more because unsold power is risk. Our constraint payments are one answer to that, but they pay a farm for not producing, out of a levy on everyone’s bills. A customer is a different thing. Now give that wind farm a bid on every unit of electricity it produces, including the 3am surplus, from a buyer who takes whatever nobody else has bought. Its revenue floor rises. Its financing gets cheaper because the risk fell. Projects that were marginal become buildable. So more generation gets built than otherwise would have been, and the effect is strongest where power is cheapest to produce, which increasingly means sun and wind.
On a grid that’s already heavy with renewables, the problem flips from “not enough supply” to “supply at the wrong times”. Midday solar floods the grid, evening demand spikes after sunset. Mining fits that curve. It soaks the midday glut and switches off for the evening peak. Utilities already pay for this under the name demand response [adjusting electricity use to match grid conditions, usually for payment]. A flexible load at that scale makes overbuilding renewables affordable, and overbuilding plus storage is the working path to a clean grid, not a slogan about one.
The version of this I find most affecting is the smallest. At Bondo in Malawi, a micro-hydro plant powers a village and mines with the electricity the villagers aren’t using at that moment, in a country where the national grid fails routinely and devaluations eat savings. Mining revenue is what lets a plant like that pay for itself instead of waiting on a donor. The villagers’ power comes first. The miner takes what’s left at that moment and drops off when the village wants more, which is why the residents stop losing power rather than start. The same logic scales up. You can raise financing against predictable generation, with mining as the buyer that guarantees the rest until real local demand grows into the capacity. The miner is there to get the wires built. The village is what they’re for.
The buyer of last resort also travels. When China banned mining in 2021, roughly half the network’s computing power went dark, and within months it had relocated and recovered. Demand that can pack up and move disciplines energy markets everywhere at once, because any region with wasted power can now monetise it.
If all of this holds, the implications compound. Energy priced in bitcoin should trend cheaper across cycles, because mining keeps pulling new, cheaper generation into existence. Cheap energy then puts the energy-hungry fixes we’ve shelved back on the table, desalination and carbon removal among them. And oil chokepoints [the straits and canals oil must pass through] lose their grip as more regions make their own power.
The case against mining’s energy use
I’d rather build the case properly than knock down a cheap one. It has five planks.
-
It’s net new demand at national scale. Whatever the mix, bitcoin adds demand that wouldn’t otherwise exist, comparable to a mid-sized country’s electricity use. On any grid where the marginal generator [the plant that switches on to serve the next unit of demand] burns fuel, marginal demand is fossil demand. “We use waste” describes part of the industry, not all of it.
-
The mechanism is fuel-blind. The same revenue floor that rescues a wind farm rescues a coal plant. The protocol doesn’t care what the fuel is, only the price. And the objector can point to real cases, like a gas-fired plant in New York state revived largely to mine, coal-heavy Kazakhstan absorbing hashrate after the China ban, and a small American city freezing new mining after residents’ bills rose. Those happened.
-
The energy budget scales with price, not with need. Nothing in the protocol defines how much security is enough. What the block reward is worth sets the budget, so if the price rises tenfold, the energy spend chases it. “Bitcoin only uses what it needs” is false as stated. It uses what the reward will pay for, with no governor.
-
The green claims can’t be audited. The renewable-share numbers mostly come from the industry itself, and the counterfactual, whether that wind farm would have been built anyway, is unknowable case by case. The industry grades its own homework.
-
The hardware is disposable. Mining machines are single-purpose and newer ones outrun them in a few years, so the old machines pile up as electronic waste. I’ll hold my reply on this one lightly.
That’s the objection at full strength. Now my replies, and what I concede.
Planks one and two are true as mechanics, and I won’t pretend otherwise. What I’d say back is that the incentive bends cleaner over time, because solar and wind keep falling in cost while anything that burns fuel has a permanent floor. Fuel costs money forever. Sunshine doesn’t. And mining is actually a poor match for coal economics. A coal plant wants steady, high-priced demand around the clock, and mining is the customer that vanishes the instant prices rise. It’s about the best customer a wind-and-solar grid could ask for and a mediocre one for baseload fossil [plants built to run flat out around the clock]. That argument doesn’t cover the gas plant, and I won’t stretch it to. A plant that can follow the price is a good partner for a miner, which is why that one happened. Kazakhstan is the case where my own reply gets tested. Miners went there for cheap coal power, the grid couldn’t carry them through winter, and they were the first load cut off. That’s the mechanism working rather than failing, but it took a power crisis to show it, and the households who lost power in the meantime didn’t get a say. On flared gas specifically, the gas is already being burned and wasted, so the right comparison is a flare against a generator, not burning against not burning. A generator burns it more completely, so less methane escapes unburned, and the gas does some work on the way out instead of none. Where the alternative is venting rather than flaring, the gain is larger still, because methane warms the planet far more than the carbon dioxide it becomes. I’m outside my notes on the combustion figures, so treat the size of that gain as my general understanding, not the thesis. Where fossil power is artificially cheap because of subsidy, yes, mining will buy it, like every other industry does. That’s an indictment of the subsidy, not of the load, and the moment the pricing is fixed, mining is the first load to leave, because it has no reason to stay put.
One of the cases I listed isn’t about fuel at all, and it’s the one that should worry you most. A city froze new mining because residents’ bills went up. The miner was on the same cheap firm power [power that’s there around the clock, not just when the wind blows] the town was already using. That’s the opposite of everything I’ve just described. A buyer of last resort takes what nobody else wants and gets out of the way when somebody does, and a load that shares your neighbour’s cheap supply while your neighbour covers the extra cost isn’t that. The design I’m defending is the one where the miner goes last. The town case happens when the price a household pays doesn’t move with the cost of the next unit, so the miner’s demand lands on everyone’s bill instead of its own. Fix that and the miner is the first customer to leave, because its margin is the thinnest in the room. What I can’t offer is a way to make bitcoin fix that pricing, and until a town fixes it, the objection lands.
On plank three, I partly concede the mechanism. The spend does track price, not a defined need. My reply is about what the spend buys. It buys a cost that renews, one an attacker has to keep paying for as long as they want their version of the record to stand. That renewal is the thing proof of stake gives up, because a stake is bought once and stays bought, so a rich enough attacker can buy in and stay in. But whether the long-run budget lands at the right level, especially decades out when new-coin issuance fades and transaction fees have to carry security on their own, is an open question, and I won’t pretend it isn’t.
On plank four, I agree entirely. Argue from mechanism and named, checkable cases, never from industry survey data. That’s what I’ve tried to do here. It’s also why I said earlier that projects which were marginal become buildable, rather than claiming mining built any particular one. The one project I did name, Bondo, I named because you can go and look at it, not because I can prove the village would be dark without the miner. The mechanism is checkable. A guaranteed buyer raises a project’s revenue floor and lowers its financing cost, and cheaper financing gets more projects built. Which specific projects it cleared is exactly the thing nobody can audit, including me.
On plank five, the e-waste cost is real. The partial reply is that machines get resold down the cost curve to cheaper power rather than binned on day one, but treat that as my general understanding, not the thesis.
The energy bill nobody prints
The environmental question is which monetary system forces more energy through the world, not how much electricity bitcoin uses. Bitcoin’s energy cost is on a meter. Somebody gets a bill for it, and where electricity is priced properly that somebody is the person who wants what it buys. The system it replaces spends energy invisibly.
That system anchors demand for money in oil. After 1971, when the dollar’s link to gold ended, pricing the world’s oil in dollars held up demand for the currency, and that arrangement has been defended, when needed, with military force. A monetary system that needs carriers to keep the world wanting its money has an energy bill nobody prints on a chart. Bitcoin prices energy by buying it in the open instead. It’s the first monetary asset that could become a reserve without a navy.
The old system must also push consumption up year after year. A debt-based system fails when prices broadly fall, because the debts are fixed in pounds while incomes fall with prices, so defaults cascade. So when technology makes energy cheaper, the response is to create money and credit until prices rise anyway. Cheaper solar should mean lower bills and less extraction, but the new money lifts the general price level, the saving never reaches you, and the higher oil price keeps tar sands and marginal wells in business. A system that mandates rising prices on a finite planet is mandating more energy through the world, and more materials with it, forever. That’s why I say inflation is climate change in disguise. It’s the same growth engine seen from two sides.
And the same system punishes saving, which pushes savers into consumption and speculation just to outrun the melt in their savings. Money that holds its value lets efficiency show up as falling prices, so people can work less, buy less, and waste less without getting poorer.
So rather than bitcoin’s metered electricity versus zero, the comparison sets a system that pays for its security openly, in metered electricity, against a system that hides its energy cost in forced growth, propped-up oil economics, and the wars it makes cheap to start. That comparison is structural, not measured. Nobody has cleanly measured the fiat system’s total energy cost, and maybe nobody can. I hold it as a strong argument from mechanism, not as a settled number.