When Mining’s Waste Heat Isn’t Wasted: Energy Mix 2026

Every argument about crypto and energy begins in the same place: a figure for how much electricity mining burns, and a fight over how clean it is. That is the energy mix, and it matters. But it describes only one side of the meter, the side marked input. It says nothing about what the energy does on its way out. For most of the last decade there was nothing to say, because the answer was identical everywhere: the heat left the building through a wall of fans and disappeared into the air.

In 2026 that stopped being true in enough places to matter. Miners in Finland are feeding district-heating grids that warm tens of thousands of residents. An operator above the Arctic Circle heats a converted hospital and dries fish with machines that happen to secure Bitcoin on the side. Agricultural engineers are modelling greenhouses that grow produce on the exhaust. None of it lowers the headline consumption number by a single watt. What it changes is the accounting of net harm, and it reframes a question the mix alone cannot answer: is this load purely extractive, or can one kilowatt-hour do two jobs?

This is a guide to both ledgers, the one every headline cites and the one almost no one does. It covers what the mix actually looks like now, why the published figures disagree by almost a factor of two, why proof-of-stake sits this fight out entirely, and why the most interesting energy story in mining this year is happening at the outflow pipe rather than the power line.

Two meters, not one

Three separate quantities tend to get blurred together in casual coverage. The first is consumption: the total electricity a network draws, measured in terawatt-hours per year. The second is the mix: the share of that electricity coming from gas, coal, hydro, wind, solar or nuclear. The third, and the one that almost never appears, is net effect: consumption and its carbon, minus whatever the system displaces or reuses, plus the externalities a power figure hides, such as water and electronic waste.

Consumption and mix are input-side metrics. They tell you how big the load is and how dirty the feedstock is. They are necessary, and Bitcoin’s critics are right that they are often reported carelessly. But they are not the whole meter. A gigawatt is a gigawatt whether the heat it produces vanishes into a Texas night or warms an apartment block in Lapland. The difference between those two outcomes is invisible to the mix and visible only on the output ledger. The argument here is simple: in 2026 the output ledger finally has entries on it, and ignoring them distorts the picture in both directions, too flattering for the industry’s marketers and too damning for its harshest critics.

What the mix actually looks like in 2026

Start with the input side, because it is the part with real data behind it. In its April 2025 study, the Cambridge Centre for Alternative Finance put Bitcoin’s annual electricity draw at about 138 terawatt-hours, roughly 0.5% of global electricity, with around 52% coming from sustainable sources (close to 43% renewables plus almost 10% nuclear). Natural gas was the single largest fuel at about 38%, coal was down near 9%, and the network’s footprint came to 39.8 million tonnes of CO2 equivalent. The study drew on direct reporting from 49 mining firms across 23 countries, covering roughly 48% of network hashrate.

Alexander Neumueller, who leads the centre’s digital-assets research, framed the methodology as a deliberate move away from guesswork, saying the report “directly addresses a persistent data gap by relying on direct practitioner insights rather than abstractions.” A preliminary update presented in late 2025 pushed the numbers higher: annual consumption of roughly 190 terawatt-hours, up about 38% in eighteen months, with low-carbon sources rising to 59.4% and hydropower overtaking gas as the largest single source. Emissions rose about 20%, to roughly 48 million tonnes, a smaller jump than consumption because the feedstock kept getting cleaner. The trajectory, in other words, is more power and a greener share at the same time.

One reason the total keeps climbing is that mining consumption tracks revenue, not transaction count. When the Bitcoin price rises, more machines become profitable and more power gets switched on, which is why energy models key off price and block rewards rather than usage. It is also why energy-based valuation frameworks such as stock-to-flow and production-cost models keep circling the same variable: the electricity bill is the floor under the whole enterprise.

Why the number is a range, not a reading

No one can walk up to a global, permissionless network and read its meter. Every consumption figure is a model, and the two best-known models disagree by almost half. Cambridge works bottom-up, combining machine-level efficiency data with operator surveys. Digiconomist, run by the economist Alex de Vries, works top-down from miner economics, on the logic that the richer mining gets, the more power it can afford to burn. That model currently puts the network near 204 terawatt-hours a year, comparable to Thailand, with about 114 million tonnes of CO2, 22.7 kilotonnes of electronic waste and 3,222 gigalitres of water attached to it. The method assumes electricity runs to roughly 75% of miner income, about 10.2 billion of an estimated 13.6 billion dollars, at a conservative five cents per kilowatt-hour.

Those are not small differences, and they are not dishonesty. They are what happens when a figure has to be inferred rather than measured. A third voice, the data analyst Daniel Batten, lands close to Cambridge on sustainability but argues the standard ledgers miss a credit: miners that burn waste methane are destroying a greenhouse gas far more potent than the CO2 they emit, which on his accounting can push some sites net carbon-negative. The practical takeaway for a reader is to treat any single Bitcoin energy number with suspicion and to ask which model produced it. The honest range is wide, and, as the rest of this piece argues, the output-side entries are not captured by any of them.

Proof-of-stake moved the denominator to the floor

Much of the heat in the crypto-energy argument comes from treating all of crypto as if it mines. It does not. Ethereum, the second-largest network, stopped mining in September 2022 when it switched to proof-of-stake in the event known as the Merge. Validators replaced miners; instead of racing to solve a hash puzzle, they are chosen to propose blocks in proportion to the stake they lock up. The power draw of that job is trivial. Ethereum’s own figures put the network at roughly 0.0026 terawatt-hours a year, about 2,601 megawatt-hours, a cut of more than 99.98% in electricity and about 99.99% in carbon from its pre-Merge draw of around 21 terawatt-hours. A single validator runs on tens of watts, the order of a couple of light bulbs.

That one design choice is why the energy debate is, in practice, a Bitcoin debate. It also kills the output-side story before it begins for proof-of-stake: there is no meaningful waste heat to reuse when the whole network draws less power than a small office park. The flip side is that proof-of-stake carries its own concentration and custody questions that proof-of-work does not, which is a large part of why liquid-staking providers such as Lido, Rocket Pool and Frax draw so much scrutiny. Ethereum’s engineers also make a point worth borrowing: they reject the popular per-transaction energy metric outright, noting that “the energy required to propose and validate a block is independent of the number of transactions within it,” so dividing network power by transaction count invents a relationship that does not exist.

The one number nobody disputes

Here is the physical fact the entire output ledger rests on, and unlike the consumption figures it is not a model. A Bitcoin miner does no mechanical work and emits almost no light. By the first law of thermodynamics, essentially all of the electricity it draws has to leave the machine as heat. An ASIC is, in energy terms, a very expensive space heater that happens to earn Bitcoin while it runs. The heat is not a byproduct you can engineer away or make more efficient; it is close to 100% of the input, every second the machine is switched on.

For fifteen years that heat was pure loss, dumped outdoors by immersion baths or fan walls because capturing it was more trouble than it was worth. The shift in 2026 is not a new technology. It is a change in where miners choose to sit, and a recognition that the heat becomes free the moment you decide to catch it rather than exhaust it. The catch is that low-grade heat does not travel well, so the value only appears when the miner is placed next to something that needs warming. That constraint, more than any breakthrough, explains the map of where heat reuse is actually happening.

Finland turned the exhaust into a utility

Finland is the clearest proof of concept, because it has two things most mining jurisdictions lack: ubiquitous district-heating networks and winters cold enough to make heat a product worth buying. MARA Holdings, one of the largest listed miners, runs two Finnish sites feeding those networks. Its Satakunta installation delivers about 3.5 megawatts of heat at 55 to 78 degrees Celsius, hot enough to feed the grid directly without heat pumps in milder months; its Seinäjoki site adds a megawatt at 50 to 60 degrees plus a quarter-megawatt of heat-pump capacity that can lift output toward the 95 degrees peak winter demand requires. Both were, by the company’s account, energised in under 30 days.

The human scale is what makes the Finnish case stick. A January 2026 Grist investigation reported that MARA’s heat reaches about 80,000 residents across the two districts, and that each megawatt of recycled heat avoids somewhere between 455 and 720 tonnes of CO2 a year, depending on what it replaces. In Seinäjoki the mining heat displaced fuels the utility had burned for years. Mikko Mursula, head of the district-heating unit at the local utility Seinäjoen Energia, put the economics plainly: “With that heat, we replaced peat, oil, and wood,” and said it significantly lowered the total heat-production cost. Adam Swick, MARA’s chief strategy officer, described the design goal as invisibility: “That’s kind of the goal, that nobody’s impacted.”

MARA is not alone. Terahash Energy runs a Finnish heat project it calls Genesis, and Hashlabs operates six sites built around the same idea. Alen Makhmetov, a Hashlabs co-founder, framed heat offtake less as charity than as survival strategy, saying he wants to make the business “much more sustainable, robust, and long term.” That instinct is also commercial: with electricity costs and taxes on large computing loads climbing across the Nordics, a second revenue stream from heat can be the difference between a viable site and a shuttered one.

Above the Arctic Circle, the heat is the product

The Finnish sites are mines that sell heat. A project in Kirkenes, a fishing town on Norway’s far northeastern coast, inverts the order. Sazmining built a roughly 2.5-megawatt installation inside a 350-square-metre former hospital, using hydro-cooled machines in a closed loop that moves heat through a radiator into a second loop running through the building. CoinDesk reported the original plan in late 2024, when the operator set out to heat a building in the Arctic village rather than simply mine there. The build eliminated an oil boiler, turned the old hospital into a business centre, and now covers around 30% of the site’s heating, with some of the warmth used to dry fish, a staple of the local economy.

In a place where winter temperatures sit near minus 20 degrees Celsius for long stretches, heat is not a nice-to-have. Sazmining’s chief executive, Kent Halliburton, has argued that building heat-first changes the operator’s political standing: a company providing a tangible local service is far harder to shut down than one that only exports hashes. That is the output ledger showing up as regulatory resilience rather than carbon math. It also previews the business model heat reuse points toward, in which the Bitcoin earned is the subsidy that makes cheap, clean heat possible, not the other way around.

Greenhouses, pools, and the limits of the idea

Heat is fungible once you have it, so the list of offtakers keeps growing: greenhouses, swimming pools, commercial laundries, timber kilns and residential heaters have all been paired with miners. Agricultural engineers have gone as far as modelling a mining-integrated greenhouse with predictive controls that schedule hashing around crop heat demand. The appeal is obvious: the grower gets heat that would otherwise cost money, and the miner gets a site with a reason to exist beyond the Bitcoin price.

The limits are just as real, and they are the reason heat reuse is not a silver bullet. Miner heat is low-grade, in the 50 to 80 degrees range, and it does not travel; you cannot pipe it across a country, so demand has to sit next to the machines. That favours compact Nordic district grids and penalises the sprawling, remote sites that make up most of the industry. Demand is also seasonal, collapsing in summer. And the decisive problem is scale. A network drawing an average of more than 20 gigawatts around the clock dwarfs the few megawatts reclaimed at each heat site; add up every project running today and it is still a rounding error against the whole. Heat reuse genuinely lowers net harm where it is deployed. It does not, and cannot yet, offset the network.

The other output: emissions that never happen

Heat is the most tangible output-side entry, but it is not the only one. The second is avoided emissions, and it turns on a gas far more damaging than carbon dioxide. Oil fields and landfills release methane, which traps roughly 80 times as much heat as CO2 over a twenty-year horizon. Much of it is simply flared or vented because there is no pipeline to carry it anywhere useful. A miner is a buyer that can show up at a remote wellhead with no grid connection, burn that methane in a generator, convert it into the less-potent CO2, and earn Bitcoin from the power in the process.

This is the strongest version of the industry’s green case, and also the most contested. Advocates like Batten argue that when a site destroys methane that would otherwise have escaped, the climate math can run negative, and that treating such a site the same as a coal-fired one is simply wrong. Skeptics counter that the effect is real but narrow, that it is used as a halo for a network still running mostly on grid power, and that paying miners to show up can prolong the life of fossil infrastructure that should be capped instead. Both can be true at once. Flared-gas mining is a genuine improvement at the specific sites that do it, and a small share of a network whose centre of gravity is still gas and, in places, coal.

The critics are not buying the halo

Every output-side claim runs into the same objection: does any of it justify the load in the first place? Johanna Fornberg, a senior research specialist at Greenpeace, gave the sharpest version of it in the Grist reporting on Finland’s heated towns. Shown the district-heating success, she was unmoved: “Does that actually provide justification for the industry? I would argue it does not.” Her point is that reusing heat improves a system that, in her view, should not be drawing the power at all.

De Vries presses the externalities that heat reuse does nothing about. His model attaches thousands of gigalitres of water and more than twenty kilotonnes of electronic waste to the network each year, the latter because application-specific mining chips have short useful lives and little salvage value. Those figures lean on a per-transaction framing that is itself disputed, for exactly the reason Ethereum’s engineers flagged: network energy tracks hashrate and block rewards, not transaction volume, so per-transaction water or waste numbers can mislead. The fair reading sits between the camps. Heat reuse and methane capture are real improvements where they occur, and they are also marketing devices where they do not, deployed far more often in press releases than in megawatts.

Heat reuse meets the AI megawatt war

The force reshaping mining’s mix in 2026 is not policy or carbon targets. It is competition for the same cheap, clean power from artificial-intelligence and high-performance computing. The International Energy Agency projects that data-centre electricity demand will roughly double, from about 485 terawatt-hours in 2025 to around 945 terawatt-hours by 2030, close to 3% of global supply, with AI the fastest-growing slice. Miners sitting on interconnection rights and power contracts have become acquisition targets and landlords: the sector has signed more than 70 billion dollars of AI and HPC deals, including Riot Platforms leasing 191 megawatts to the AI developer Anthropic in a twenty-year agreement valued at about 9.1 billion dollars.

That bidding war cuts against the mix on paper, because AI tends to outbid miners for the cleanest, most reliable power and pushes mining toward whatever is left. But it also sharpens the value of the output ledger. A site that sells heat, or that can power down in seconds when the grid is stressed, has revenue and goodwill that a flat-out AI inference hall cannot easily copy. The same logic drives the decentralised GPU-compute markets such as Akash, where the open question is always whether raw capacity can be turned into durable value. For marginal miners being outbid on electricity, heat offtake and grid flexibility are becoming part of how they stay solvent, not an ESG footnote.

What regulators count, and what they ignore

Official disclosure rules decide which numbers the public ever sees, and right now they measure only the input side. In the European Union, the sustainability rules under MiCA, set out in Commission Delegated Regulation (EU) 2025/422, make one indicator always mandatory: the total annual electricity of a crypto-asset’s consensus mechanism, in kilowatt-hours. Cross 500,000 kilowatt-hours a year and the supplementary fields kick in, including renewable share, energy intensity per transaction and greenhouse-gas emissions. Energy mix, electronic waste and water are optional. Those rules have applied to crypto white papers since late December 2025.

Read that list again and notice what is missing. Nothing in it credits reused heat, destroyed methane or demand-response flexibility. The official European scorecard measures input harm and is silent on net effect, which means the exact entries growing fastest in 2026 are invisible to it. The United States measures even less. The Securities and Exchange Commission adopted a climate-disclosure rule in March 2024, saw it stayed in the courts, and in 2026 moved to rescind it, leaving federal climate reporting effectively dormant. So the green number is mandatory but input-only in Europe and voluntary in America, and in neither place does the paperwork know that a mine in Lapland is heating an apartment block.

What it changes for the energy-mix debate

None of this is an argument that Bitcoin’s energy use is small, or that the output ledger cancels the input one. The network still draws the power of a mid-sized country, most of it still comes from the grid, and a large share of that grid is still gas. The argument is narrower and, for once, cuts both ways. The mix is necessary but not sufficient. The better 2026 question is net effect on the energy system: consumption and its carbon, minus heat reuse and avoided methane, plus water and electronic waste, all of it shaped by how flexibly the load runs. On that fuller accounting some sites look worse than the headline and a few look far better.

Remember what all that power is actually for. The electricity buys security, a multibillion-dollar-a-year wall of hashing that makes rewriting Bitcoin’s ledger prohibitively expensive, and that even a future quantum computer could not cheaply out-mine. With Bitcoin trading around 82,000 dollars in mid-October 2026 and the hashrate hovering around a zettahash per second after a volatile year, the revenue that funds the whole machine is healthy enough to keep the lights on. The miners most likely to survive the AI squeeze and the next halving are the ones already pricing in a second product; whether that is enough to move the aggregate mix, or to justify the load at all, is the fight the price-target models never quite settle. What is clear is that the exhaust pipe has stopped being a dead end and started being a revenue line.

Frequently Asked Questions

Does Bitcoin mining waste heat actually heat homes?

Yes. In Finland, MARA’s two pilot sites feed district-heating networks that serve about 80,000 residents, and an operator in Kirkenes, Norway covers roughly 30% of a converted building’s heating. Almost all of the electricity a miner draws leaves the machine as heat, so capturing it is a question of plumbing rather than new physics.

How much energy does Bitcoin use in 2026?

Estimates range from about 190 terawatt-hours, in Cambridge’s late-2025 preliminary update, to more than 200 terawatt-hours in Digiconomist’s economic model, roughly the annual electricity use of a mid-sized country. The spread exists because no one can meter a global, permissionless network directly, so every figure is a model.

Is Ethereum part of the energy debate?

Barely. Ethereum switched from mining to proof-of-stake in September 2022 and cut its electricity use by more than 99.98%, to roughly 0.0026 terawatt-hours a year. That is why the crypto energy argument is, in practice, a Bitcoin argument.

Does reusing waste heat make Bitcoin green?

Not on its own. Heat reuse improves the specific sites that do it, but the megawatts involved are a tiny fraction of a network that averages more than 20 gigawatts. It lowers net harm where it is deployed; it does not offset the whole network.

Do regulators give miners credit for reusing heat?

No. The European Union’s MiCA disclosure rules require issuers to report electricity use, renewable share and emissions, but not heat reuse, avoided methane or grid flexibility. In the United States, the SEC’s climate-disclosure rule has been stayed and is being rolled back, so the figure there is voluntary.

By the HOGE Wire mining and energy desk.