On this page
What miners actually do
Somewhere in west Texas there is a building the size of several football pitches, filled floor to ceiling with identical metal boxes. Each box contains hundreds of chips that do one thing: compute the SHA-256 hash function, a few hundred trillion times a second. The noise is like standing behind a jet. The building uses as much electricity as a small city. Every ten minutes or so, somewhere in the world, one box in one building like this gets lucky.
That is bitcoin mining. Miners collect pending transactions into a candidate block and then search for a number that makes the block’s hash fall below the network’s target. The search is pure trial and error. The miner who finds a valid hash broadcasts the block, every node checks it in an instant, and the miner is paid: 3.125 newly created bitcoin plus the fees attached to the transactions inside.
Mining does two jobs at once. It is how new coins enter circulation, on a schedule no one can accelerate. And it is what makes the ledger expensive to tamper with, because rewriting history means redoing the work. If you have not tried it yet, the mining demo lets you find a block by hand, at a difficulty unimaginably lower than the real thing.
From laptops to ASICs
Satoshi mined on an ordinary computer and assumed everyone would. The idea of one CPU, one vote lasted about eighteen months.
- 2009: CPUs. Any desktop could mine 50 coins a block. Hal Finney turned his off because the fan was noisy.
- 2010: graphics cards. Gamers’ GPUs turned out to be hundreds of times faster. Laszlo Hanyecz, of pizza fame, was among the first to write the code.
- 2011: FPGAs. Reprogrammable chips offered better efficiency for a brief period.
- 2013: ASICs. Application-specific integrated circuits, chips designed to do nothing but SHA-256, arrived and made everything else obsolete within months. They have been the only viable way to mine since.
Three manufacturers make most of the world’s machines: Bitmain and MicroBT, both Chinese, and Canaan. Efficiency is measured in joules per terahash. The Cambridge Centre for Alternative Finance put the fleet average at about 28 J/TH in mid-2024; the newest machines use well under 20. A machine that is a few points less efficient than its neighbours becomes unprofitable first when the price falls or the reward halves, which is why the industry is in a permanent upgrade race.
You can run a machine, but you will almost certainly pay more for electricity than you earn unless your power is unusually cheap. Hobbyists do it for heat, for education, or for the tiny chance of “solo” mining a whole block. Treat any service promising easy mining profits with suspicion; cloud mining offers are a perennial scam.
Pools: sharing the lottery
A single machine might wait centuries to find a block on its own. So miners pool their effort. Each contributes hashrate to a shared attempt, and when anyone in the pool finds a block the reward is split according to work contributed. The first pool, run by a Czech programmer known as Slush, started in late 2010. Today almost all mining goes through a pool.
Three pools producing more than 60 percent of blocks sounds alarming, and it is the centralisation concern raised most often by bitcoin’s own developers. Two points soften it. Pools do not own the machines; miners can and do switch pools in minutes if one misbehaves. And a pool that attacked the network would destroy the value of the very coins it is paid in.
The subtler worry is over who builds the block. In most pools the operator, not the individual miner, chooses which transactions to include, which makes a few companies potential points of pressure for censorship. Newer protocols such as Stratum V2 and the DATUM system used by the OCEAN pool hand that choice back to individual miners. Adoption is growing but still small.
Where the hashrate lives
Mining goes where electricity is cheapest and the law is tolerable. Until 2021 that mostly meant China, which hosted somewhere between two-thirds and three-quarters of the world’s hashrate, much of it chasing surplus hydropower in Sichuan in summer and coal in Xinjiang in winter. Then Beijing banned it.
The United States inherited the lead, with Texas at its centre thanks to cheap wind and gas power and a grid that pays large users to switch off at peak times. Its share has slipped over the past year as American miners redirect power to AI. Russia’s share is the highest on record, built on Siberian hydro and gas. China never went to zero; mining continued quietly despite the ban. Paraguay, with a vast surplus from the Itaipu dam, has become the fourth-largest location. Bhutan, Ethiopia, the UAE and Kazakhstan all host significant operations.
The economics of a miner
A miner’s revenue per unit of hashrate is called hashprice. It rises with the bitcoin price and with fees, and falls as more machines join the network. Its costs are electricity, which Cambridge found makes up more than 80 percent of cash operating costs, plus machines that wear out or become obsolete in a few years.
Every four years the main source of revenue is cut in half overnight. The April 2024 halving took the subsidy from 6.25 to 3.125 BTC. Fees were supposed to fill the gap, and for one spectacular block they did: the launch of the Runes protocol at the moment of the halving pushed fees in that block above 37 BTC. Since then fees have mostly been a rounding error, under one percent of miner revenue through much of 2026.
The squeeze shows in the data. Network hashrate reached a record of roughly 1,110 exahashes per second in November 2025. Difficulty peaked at about 156 trillion on 29 October 2025. As the price fell through the first half of 2026, older machines were switched off: difficulty dropped by about a fifth by late August, and hashprice hit a low in June. Both have since recovered somewhat, with hashrate around 980 EH/s in early October 2026, just under one zettahash.
The next halving, to 1.5625 BTC, is expected in 2028. What happens to mining, and to bitcoin’s security, as the subsidy keeps shrinking is one of the most important open questions about the system. It is covered under security budget.
The energy debate
Bitcoin uses a great deal of electricity. On that, everyone agrees. The Cambridge Centre for Alternative Finance, the most widely cited academic source, estimated in its April 2025 industry report that mining consumes about 138 terawatt-hours a year, around half a percent of global electricity use and comparable to a mid-sized country, producing an estimated 39.8 million tonnes of CO₂-equivalent.
A vast footprint for a small payment system
Critics, among them the researcher Alex de Vries, argue that bitcoin processes a few hundred thousand transactions a day for the electricity of a nation, and that the design guarantees this: the more valuable bitcoin becomes, the more energy is worth spending to mine it. Proof-of-stake networks secure billions of dollars with a tiny fraction of the power. Gas is now mining’s largest single energy source. Mining also produces electronic waste as machines are retired, and communities near large sites have complained of noise, water use and higher local power prices. On this view the energy is simply the cost of a lottery.
A flexible buyer of energy nobody else wants
Supporters reply that energy use is not waste if it buys something people value, and that proof of work is the only security model with a seventeen-year record at this scale. Miners are unusual customers: they can set up anywhere, use power at any hour and switch off in seconds. That lets them soak up surplus wind, solar and hydro that would otherwise be curtailed, earn money by powering down when a grid is stressed, and burn flared or leaking methane that would otherwise go straight into the atmosphere. The share of mining on sustainable sources has risen sharply, and per-transaction comparisons are misleading because energy secures the whole ledger, not individual payments.
Both sides are describing real things. Mining is a large and growing energy consumer whose mix has become cleaner. Whether the trade is worth it depends on a judgement the data cannot make for you: how much value you place on a monetary network that no government controls.
The public miners
A dozen or so mining companies are listed on stock exchanges, mostly Nasdaq. The largest by hashrate include MARA, CleanSpark, Riot Platforms, IREN, Cango and Core Scientific, along with Bitdeer, founded by Bitmain co-founder Jihan Wu, and American Bitcoin, the Hut 8 subsidiary co-founded by Eric Trump. MARA alone reported 66.4 EH/s of installed capacity at the end of 2025.
For years the listed miners split into two schools: those who sold every coin to pay the bills, and those who held as much as possible and borrowed against it. The 2026 downturn blurred the line. MARA, the largest holder among miners, sold about 15,000 BTC in March to fund new lines of business. Their holdings appear in the table on the holders page.
The AI pivot
A bitcoin mine and an AI data centre need the same scarce things: a large power connection, land, cooling and permits that can take years to obtain. Miners already have them. Since 2024, as AI companies scrambled for electricity, miners discovered that renting their sites for AI computing could pay more per megawatt, on long contracts, than mining bitcoin with it.
- Core Scientific contracted about 590 megawatts to the AI cloud company CoreWeave over twelve years, a deal expected to bring in roughly $10 billion.
- IREN has rebuilt itself largely around AI cloud services.
- MARA said in March 2026 that it was expanding beyond pure mining into energy and AI infrastructure, and its chief executive has said AI can earn more from a unit of power than mining.
- CleanSpark, Riot, Hut 8 and others have announced similar moves.
Analysts link part of the decline in American, Canadian and Emirati hashrate over the past year to this shift. For bitcoin, the effect is mixed. Some hashrate leaves, and the difficulty adjustment absorbs it. The miners that remain are financially stronger for having a second income. But it also means that bitcoin now competes for electricity with an industry that can outbid it, which sharpens the long-run question of who will secure the network when the subsidy is small. The people running these companies are profiled in the directory.