How Bitcoin Mining Works

Finding a block: astronomically hard

Across the whole network, miners run on the order of a sextillion (10²¹) hashes every second. It still takes about ten minutes for one of them to get lucky. Every failed guess burned real electricity that nobody gets back.

Source: mempool.space, hashrate and difficulty

Nobody has to trust the result

Verification is two SHA-256 operations and one numeric comparison. A phone does it in microseconds. You never have to trust the miner, replay the search, or take anyone's word for anything.

Clamped at 4x

A single adjustment can never move difficulty by more than a factor of four in either direction. Before any recalculation, the measured timespan is clamped to a quarter of the target at one end and four times it at the other. That caps the damage from timestamp manipulation and smooths out real shocks.

Source: Bitcoin Core, pow.cpp

The coinbase transaction

Every block's first transaction is the coinbase. It has no inputs: it creates the subsidy out of nothing and sweeps up the fees from every other transaction in the block. It also carries an arbitrary data field. That is where Satoshi wrote “The Times 03/Jan/2009 Chancellor on brink of second bailout for banks” into the genesis block. (The exchange named itself after the term, not the other way around.)

Source: Bitcoin Wiki, genesis block

100-block maturity

Newly mined coins cannot be spent for 100 blocks, about 16 to 17 hours. It is a single constant, COINBASE_MATURITY, in the reference implementation. If a competing chain later orphans that block, meaning that chain wins and the block drops out of history, the reward vanishes with it. The rule stops anyone from spending coins that a reorganization might erase. A small detail, and a sign of how carefully the reorg case was thought through.

Source: Bitcoin Core, consensus.h (COINBASE_MATURITY)

PPS and FPPS

Pay-per-share: the pool pays a fixed rate for every valid share, whether or not the pool finds any blocks that day. FPPS adds an average share of transaction fees on top. The pool absorbs all the variance and charges a higher fee for carrying that risk. Revenue comes out smooth and boring, which is exactly what a miner with a power bill wants.

PPLNS

Pay-per-last-N-shares: payouts come only out of blocks the pool actually finds, split across the most recent N shares. Fees are lower and income is lumpier. Loyalty gets rewarded, since hopping between pools forfeits your position in the share window. Over the long run PPLNS and PPS converge; they differ in who holds the risk in the meantime.

Joules per terahash is the whole game

Efficiency is measured in J/TH, energy burned per trillion hashes. The S9 generation ran at about 98 J/TH; the 2024 S21 XP does 13.5 J/TH, roughly seven times better; and the newest 2026 machines push below 10 J/TH. Every generation lowers the electricity price at which the previous one breaks even, and eventually strands it entirely.

Source: ASIC Miner Value, S9 vs S21 XP

Hardware depreciates in two directions

A miner is a wasting asset. As newer machines come online, difficulty rises and your machine earns fewer bitcoin per month even if nothing else changes. Its resale value falls at the same time, for the same reason. Operators underwrite purchases against forecasts of difficulty growth, and they are frequently wrong.

Hashprice: the one number that matters

Hashprice is revenue per unit of hashrate per day, usually quoted in dollars per petahash per day. It folds price, difficulty and fees into one figure, and power contracts and hosting deals get underwritten against it. It falls whenever difficulty outruns price, and spikes when fees do. Every halving cuts it roughly in half overnight.

The honest home-mining math

An S21 XP draws 3,645 W. Run continuously that is roughly 2,600 kWh a month, about $480 at the US residential average of 18.4 cents per kWh. An industrial operator on a $0.04/kWh contract pays around $105 for identical work. Same bitcoin out, roughly four to five times the cost in, and difficulty is set by their economics. Then add a few thousand dollars of hardware, serious fan noise, and 3.6 kW of heat you have to move somewhere.

Sources: ASIC Miner Value, S9 vs S21 XP · EIA Electric Power Monthly, Table 5.6.A

The kernel of truth: capitulation is real

Deep drawdowns, meaning long slides from a price peak, do produce visible hashrate declines and negative difficulty adjustments, and analysts watch them closely. That is a real, observable market dynamic. But it marks miner stress, not a price level anyone is defending. The largest drop on record, about 28% on 3 July 2021, followed China's crackdown: a regulatory event, not a price floor.

Sources: CoinDesk, July 2021 difficulty adjustments · mempool.space, hashrate and difficulty

Miners sell into weakness, not out of it

Miners have fiat-denominated power bills and bitcoin-denominated revenue, so they must sell a portion of production continuously. Under stress they sell treasury reserves too, and distressed operators liquidate outright. So miner behavior adds supply during downturns rather than removing it. That is the opposite of a floor.

When the mempool backs up

During rallies and novel on-chain activity, the mempool (the queue of transactions waiting to be confirmed) fills and fees spike hard. Two things are worth real money in those windows: a wallet that lets you set your own fee rate, and patience.

Quiet blockspace, cheap fees

In slow periods a transaction can confirm for a few cents. Fee markets are volatile, and nothing in the protocol smooths them.

Depth is the defense

A transaction one block deep is far cheaper to reverse than one six blocks deep, because each additional block multiplies the work an attacker must redo. That is exactly why exchanges and custodians set confirmation thresholds, and why large settlements wait for more of them.

Not hypothetical for small chains

Smaller proof-of-work chains with thin hashrate have been 51%-attacked repeatedly. Ethereum Classic was hit three times in August 2020 alone, one attack reorganizing over 7,000 blocks. Bitcoin Gold was attacked in 2018 and again in 2020. Bitcoin's defense is that it is by far the most expensive chain to out-hash. Anyone spending that much would be destroying the value of the asset they get paid in, while holding a warehouse of purpose-built hardware good for nothing else.

Sources: CoinDesk, Ethereum Classic 51% attacks (2020) · Cointelegraph, Bitcoin Gold 51% attack (2020)

The strongest case for

Miners are unusually mobile and interruptible buyers of electricity. They don't need to sit near customers, they run at any hour, and they can shut down within seconds. That pushes them toward power nobody else wants: stranded hydro, curtailed wind and solar, off-peak baseload, gas that would otherwise be flared at the wellhead. Grid operators in Texas and elsewhere pay miners to curtail during demand peaks, which turns the load into a controllable, sheddable buyer. Cambridge's survey of firms covering roughly 48% of global mining put the sustainable share at 52.4%: 42.6% renewables plus 9.8% nuclear. That is up from 37.6% in 2022. It is self-reported survey data, so read it as an estimate with an obvious incentive attached.

Sources: Cambridge Judge Business School summary · Cambridge Digital Mining Industry Report (April 2025)

The strongest case against

The best version of the criticism isn't about the number at all. It is that energy spent here is energy not spent elsewhere. And “it uses power others did not want” does not establish that the activity is worth doing. Local costs are real too: noise complaints near facilities recur, and rapid hardware turnover produces electronic waste. Cambridge's own figures also put network emissions at 39.8 million tonnes of CO₂ equivalent, which is not a small number however you feel about the trade.

Source: Cambridge Judge Business School summary