Bitcoin and Energy

Where you're going: Bitcoin consumes a lot of electricity. That is true, measurable, and by design. This chapter walks through how much, where it comes from, what it buys, what the critics get right, and why "Bitcoin wastes energy" and "Bitcoin is an environmental free lunch" are both wrong.

Energy is the objection that comes up first, with everyone. Before quantum, before "criminals use it," before volatility, the person across the table has heard that Bitcoin boils oceans, and they have heard it from people they trust. Most of what they have heard is somewhere between stale and rhetorically engineered. Some of it is correct. The honest version is more interesting than either camp wants it to be.

The number, and the comparison trick

On Cambridge's figures as of October 2026, Bitcoin mining draws about 140 to 160 TWh of electricity a year, depending on the method. The University of Cambridge's 2025 industry survey found 138 TWh, and its Bitcoin Electricity Consumption Index read about 162 TWh in October 2026. Those are the most methodologically careful public estimates. The index moves with price and hash rate, and it put the whole of 2025 higher, at about 189 TWh. The world used 31,779 TWh of electricity in 2025, by Ember's count, so Bitcoin's share is roughly half of one percent, or about 0.6 percent on the index's 2025 total. It is a lot of electricity.

Now watch what happens to that number in the wild. "More electricity than Argentina": true, and designed to alarm. "About as much as gold mining": roughly true, depending on whose gold estimate you take, and designed to flatter. "More than all of Google": true (Google reported using about 44 TWh in 2025), and meaningless without asking what each one produces for the energy.

Every one of those comparisons is a choice of denominator, and the choice is doing the arguing. A country sounds outrageous because countries contain people. The comparison is not analysis. It is framing wearing analysis's clothes.

The honest baseline: Bitcoin uses a measurable, material amount of the world's electricity. Under one percent of the global total, more than small countries, and well short of some single industries: the world's aluminum smelters reported using about 990 TWh in 2025. Hold that number still and ask the real questions: what does it buy, where does it come from, and what would that energy be doing otherwise.

The energy is the security

Start with what the electricity purchases, because it is not "transactions."

Bitcoin's entire security model is that rewriting history costs more than anyone can profitably spend. Proof-of-work converts electricity into exactly that: every block carries proof that real-world joules were burned to produce it, and an attacker who wants to rewrite the chain must outspend the entire honest network, in real electricity, for as long as the attack runs. The energy bill is not overhead on the security. The energy bill IS the security.

This is why "make Bitcoin use less energy" is a different request than it sounds. Cut the energy and you cut the cost of attacking the ledger by the same proportion. A Bitcoin that was cheap to run would be a Bitcoin that was cheap to rewrite.

The obvious follow-up, "then why not proof-of-stake, like everything else," deserves one honest paragraph. Proof-of-stake replaces the external, physical cost with an internal, financial one: security is posted as locked capital instead of burned electricity. That is not free; it is a different trade, with different failure modes: wealth concentration compounding into control, and no physical anchor binding the ledger to the world outside its own token. Bitcoiners think the external anchor is the entire point. Reasonable people disagree about that trade. What is not reasonable is pretending it costs nothing.

What miners actually buy

Almost everything surprising about Bitcoin's energy traces back to a single fact:

A miner is a rare kind of industrial-scale electricity buyer: location-agnostic, interruptible in seconds, and indifferent to when the power arrives.

The middle claim is the one people doubt, so here is the evidence. NERC, the body that writes reliability standards for the North American grid, reviewed the new generation of very large loads in 2025. It found that mining and AI facilities can cycle their consumption on and off in under a minute, and its report includes a trace of one mining site in North America dropping 298 megawatts in 25 seconds on instruction. NERC files that speed under risks, not gifts, and it is right to: a load that can vanish in seconds can come back just as fast, and the grid has to be ready for both.

A factory needs workers, roads, and customers. A cloud data center is selling uptime, so its draw stays nearly flat whatever the grid is doing. An aluminum smelter cannot simply switch off: a few hours without current and the molten bath in its pots starts to freeze. When one smelter in Bahrain lost current for eight hours in 2024, restarting the line took 44 days, and its engineers believe that was the fastest recovery ever from such an outage. A miner needs electricity and an internet connection. It can sit at a hydro dam in the back country, a flare stack in an oil field, or a wind farm at the end of a congested transmission line, wherever a megawatt is cheapest because nobody else can use it.

That is why mining migrates to stranded power. Not because miners are environmentalists, but because stranded electrons are the cheapest electrons on earth, and mining margins are thin. The incentive does the work; no virtue required.

The same property runs in reverse, and that is where it gets interesting for everyone else.

The grid receipts

In August 2023, during a Texas heat wave, Riot Platforms (one of the largest miners on the ERCOT grid) earned $31.7 million in a single month mostly for NOT mining: $24.2 million selling its pre-purchased power back to the grid at crisis prices, and $7.4 million in demand-response credits. The bitcoin it mined that month was worth about $8.9 million. The filing is public. During the peak-demand periods, Riot curtailed more than 95 percent of its load, handing those megawatts back to the grid.

Three mechanisms, worth knowing because they are not the same thing:

1. Price response. When power costs more than the bitcoin it would mine, the rational miner shuts off. Whether it can also sell the power it is not using depends on the contract. Riot buys a block of fixed-price power under a long-term agreement, so when spot prices spike it can stand down and sell that power back to the grid for credits. A miner buying at the spot price just stops, and saves the bill.
2. Ancillary services. The grid operator pays the miner for the option to shed its load on command. The miner gets paid whether or not the call comes. This is where the seconds matter: a miner registered with ERCOT as a controllable load resource runs under the grid operator's dispatch control, and when it sells regulation service it has to answer ERCOT's signals within five seconds, the same standard a power plant meets. The grid gets a dispatchable buffer the size of a small city.
3. Peak-avoidance programs. Power down during the handful of annual system peaks, pay dramatically less for transmission all year.

Now the criticism, because it is not stupid: the head of Environment Texas, an environmental group, called the $31.7 million absurd, saying ordinary Texans got nothing for cutting their own use, and argued the grid operator should pay households to conserve instead. The honest answer is that grid operators pay many large loads and generators for flexibility, because options on demand are genuinely valuable to a grid that cannot store much; a miner is an unusually fast version of something ERCOT was already buying. The honest concession inside the answer: a grid that builds around interruptible mining demand is making a bet about how that demand behaves in a crisis, and the bet is only as good as the contracts. Both things are true, and that is the actual debate.

Methane: the one place mining might be net-negative

Methane is a far more potent greenhouse gas than CO2 over its first decades in the atmosphere. The world's landfills vent it continuously; oil fields flare it incompletely. Both are emissions with no business model attached, which is exactly the gap a location-agnostic, containerized electricity buyer can fill. Park generators and ASICs at the gas, combust it completely (turning methane into less-harmful CO2 plus electricity), and the mining revenue funds the capture.

The strongest quantified version of this argument comes from Daniel Batten's research, which estimates a surprisingly small amount of methane-fed mining would offset the network's entire footprint. Disclosure where it is due: Batten is an investor in this sector and an advocate, not a neutral academic. Treat his numbers as the well-argued bull case, not settled science.

What is not in dispute: the projects exist, they are profitable without subsidies, and the gas they burn was being vented or flared anyway. The honest status: real, growing, and still a small slice of total hash rate. Promising is the right word. Proven-at-scale is not, yet.

Heat is not waste if someone is cold

An ASIC converts nearly every watt it draws into heat. For most of Bitcoin's history that heat went out a fan into the sky. In cold places with district-heating infrastructure, it has started going somewhere better: in Finland, miners are piped into municipal district-heating systems, warming homes in a town of eleven thousand; a Nordic project is scaling toward heating roughly 2,800 homes with hot water from hashboards; greenhouse pilots preheat boiler water in Canadian winters; and you can now buy a home miner that is, frankly, a space heater with a payout address.

This is a modest claim, and it should stay modest. Heat reuse fits cold climates with existing pipe infrastructure; it will not absorb the Texas summer hash rate. It is a real and growing channel where geography cooperates, and the energy gets used twice: once for security, once for soup.

What the critics get right

The strongest forms of the critique are real, and they deserve to be stated as plainly as the defense. Here they are.

The local externalities are real, and they have a town's name on them. In Granbury, Texas, residents living next to Marathon's (MARA) mining site at the Wolf Hollow gas plant sued over constant, health-disrupting noise, documented by national press and entirely believable to anyone who has stood next to an air-cooled mining container. A 24/7 industrial fan wall is a bad neighbor. Mining companies that site carelessly deserve the lawsuits they get, and the industry's social license depends on the operators who site well.

Not all mining runs on stranded hydro. The sustainable-mix estimates range from about a quarter to about 60 percent, depending on who is measuring and whether nuclear counts; Cambridge's 2025 industry survey puts it at 52.4 percent. That also means close to half of the network, even on Cambridge's figure, runs on grids where the extra demand is met by burning coal or natural gas. The mix is improving because the economics push that way, but "improving" and "clean" are different words.

The water and e-waste critiques deserve straight answers. The water-footprint estimate published in Cell Reports Sustainability is a legitimate piece of accounting, not FUD: it put the fresh water used to generate Bitcoin's electricity at about 1.6 trillion liters in 2021, and it estimated on-site cooling water separately, for large US mines only. The best-known e-waste estimate, from de Vries and Stoll in 2021, put it at up to 30.7 thousand tonnes a year, comparable to the small IT equipment waste of a country like the Netherlands. It assumes a machine is scrapped once it stops being profitable, which the authors put at less than 1.29 years on average. Cambridge's 2025 survey of miners points the other way: most retired machines were expected to be resold, repurposed or recycled, leaving about 2.3 thousand tonnes of e-waste in 2024, though Cambridge calls its own method markedly simplified. Either way, the core of the critique stands: ASICs are single-purpose hardware, and only about a quarter of the miners Cambridge surveyed had a recycling program in place.

And the aggregate is the aggregate. Roughly half a percent of the world's electricity is real consumption with a real footprint, whatever its mix. The honest case for it has to be that what it buys is worth it, not that the number is small.

The discipline, same as our energy-fud entry puts it: engage the strong forms of the critique. The strong forms have answers. The weak forms get called FUD because they don't.

The per-transaction fallacy

One bad statistic deserves its own section, because you will see it everywhere: "each Bitcoin transaction uses as much energy as a household does in a month."

This is a stock-and-flow error wearing a lab coat. Mining energy secures the ledger (all of it, every balance, the whole history), and its level is set by the bitcoin price and the block subsidy, not by how many transactions flow through. Run twice the transactions and the energy draw does not double; run zero transactions for an hour and the miners do not power down. Dividing total energy by transaction count produces a number with the units of nonsense.

It also misses where transactions actually happen. A single on-chain transaction can settle a batch of thousands of payments, close a Lightning channel that carried months of them, or move an exchange's entire internal ledger. Per-payment energy is unknowable and shrinking; per-ledger security is the thing being bought; the famous statistic measures neither.

The honest uncertainty

Things we do not know, stated plainly:

• What secures the chain when the subsidy runs out. Mining is paid mostly by new issuance, which halves every four years. Whether transaction fees alone will fund an adequate security budget decades from now is a real open question, arguably Bitcoin's deepest, and energy spend is exactly where it will show up.
• Whether the mix numbers converge. Self-reported industry surveys and academic models still disagree by ten points or more. Better disclosure is narrowing it; it has not closed.
• Whether methane mitigation scales beyond the advocacy-stage projects, or stays a rounding error with great marketing.
• Whether the social license holds. Granbury-style fights determine where mining can physically exist. An industry that wins grid arguments and loses zoning ones still loses.

These uncertainties do not change the frame. They sharpen it: the energy story is a live negotiation between incentives that genuinely point somewhere good and an industry that does not get graded on intentions.

Closing

Bitcoin spends energy on purpose. The spend is the security; the security is the product. The honest debate is whether what it buys (final settlement, neutral money, a ledger nobody can quietly rewrite) is worth a measurable slice of the world's electricity, and who gets to decide.

Argue that question with real numbers. We just gave you ours.

---

Sources and further reading

• Cambridge Bitcoin Electricity Consumption Index: https://ccaf.io/cbnsi/cbeci
• Riot Platforms August 2023 production report (SEC filing): https://www.sec.gov/Archives/edgar/data/1167419/000155837023015517/riot-20230906xex99d1.htm
• Earthjustice on the Granbury suit: https://earthjustice.org/press/2024/granbury-residents-sue-local-bitcoin-mine-over-health-threatening-noise-pollution
• Daniel Batten's methane research (advocacy, disclosed): https://batcoinz.com/quantifying-the-potential-impact-of-bitcoin-mining-on-global-methane-emissions/
• Grist on Finnish district heating: https://grist.org/buildings/bitcoin-cryptocurrency-district-heat-finland/
• Nakamoto Institute, "Bitcoin Does Not Waste Energy": https://nakamotoinstitute.org/mempool/bitcoin-does-not-waste-energy/
• Mongabay's critical survey (read the other side): https://news.mongabay.com/2025/06/bitcoin-boom-comes-with-huge-intensifying-environmental-footprint/
• NERC's 2025 review of large loads, mining included (the risk chapters are the useful part): https://www.nerc.com/globalassets/who-we-are/standing-committees/rstc/3_doc_white-paper-characteristics-and-risks-of-emerging-large-loads.pdf

Glossary cluster on this site:

• Energy FUD
• Proof-of-Work
• Mining
• Difficulty
• Block Subsidy
• Halving