Bitcoin (BTC): a full reference on the coin and its mining

Bitcoin is the first decentralised payment network built on proof of work and at the same time the most liquid digital asset in existence. The network went live on 3 January 2009 and its technical description was published on 31 October 2008 under the pseudonym Satoshi Nakamoto. Since then the protocol has run without interruption, with no single operator and with no way to issue more coins than planned.

The purpose of the network is simple: to let two participants pass value to each other without an intermediary and without trusting a third party. The order of transactions is fixed by mining: participants spend electricity searching through SHA-256 hashes and are paid in new coins and fees for it. That work is exactly what makes rewriting history economically pointless.

For a hardware owner Bitcoin remains the main network, though not the easiest one. At the end of July 2026 the hashrate is holding around 900 exahashes per second, while fees deliver less than one percent of the block reward, so miner income is effectively equal to the subsidy of 3.125 BTC. That means brutal competition on efficiency: machines worse than 25 joules per terahash do not even cover their electricity at European tariffs.

This page is put together as a reference work: network parameters, the history of protocol decisions, a ranking of current ASIC models from our catalogue, pools, wallets, mining economics and a separate breakdown of the figures that large aggregators still publish in an outdated form.

• 2008The technical description of the network is published on 31 October. It sets out the chain of blocks, the proof of work and the longest chain rule.
• 2009The genesis block is mined on 3 January. In the first months the network runs on ordinary processors and the block reward is 50 BTC.
• 2010The first known purchase of a real good with the coin, on 22 May. In the same year mining moves to graphics cards.
• 2012First halving: the reward drops from 50 to 25 BTC. The first chips built specifically for SHA-256 appear.
• 2016Second halving: 25 BTC becomes 12.5 BTC. The ASIC market is already formed and home mining on graphics cards finally loses any point.
• 2017SegWit is activated: the signature moves out of the main part of a transaction and a block weight of 4,000,000 units replaces the hard limit in megabytes.
• 2018Lightning Network launches on the main chain. Small payments move off the first layer into channels.
• 2020Third halving: the reward falls to 6.25 BTC. A wave of institutional interest in the asset begins.
• 2021Taproot is activated at block 709,632 on 14 November: Schnorr signatures and addresses in the bc1p format. In September El Salvador gives the coin the status of legal tender.
• 2023The Ordinals protocol appears: arbitrary data starts to be written into the witness part of transactions. In some periods fees rise sharply.
• 2024Spot exchange traded funds launch in the United States on 11 January. The fourth halving comes on 20 April at block 840,000 and the reward drops to 3.125 BTC. The same month brings the launch of the Runes protocol: individual blocks earn more than 10 BTC in fees, above the subsidy itself.
• 2025A strategic bitcoin reserve is set up in the United States. On 6 October an all-time high price is reached, after which the market loses around 19 billion dollars to liquidations. On 13 October Bitcoin Core version 30 is released: the OP_RETURN limit rises from 80 bytes to 100,000 bytes and part of the node operators move to the Knots build in response.
• 2026In February the price goes below 61,000 dollars and the drawdown from the high reaches roughly 52 percent. In May pools controlling around 75 percent of the hashrate back an open standard for block construction. In March a study on quantum computing lowers the estimate of the qubits required to below 500,000, which speeds up work on the post quantum proposal BIP-360.
• 2026Two independent events are announced for August: the BIP-110 soft fork limiting arbitrary data, whose miner support in mid July stood below 1 percent, and a separate eCash hard fork at block 964,000. Neither of them is an agreed upgrade of the network.

The network stores no account balances, only a list of unspent outputs. Every transaction refers to earlier outputs, spends them and creates new ones. To spend an output you have to present a signature from the key that output is addressed to. The total of all unspent outputs is the supply of the coin.

Transactions land in a shared waiting pool and miners assemble a block out of it. A block is capped at 4,000,000 weight units, so not everything fits: transactions with a higher fee per unit of weight take priority. The average block size stays around 1.6 megabytes and the base throughput of the first layer runs from 3 to 7 transactions per second.

A block once found spreads across the network and every node checks it independently. The node verifies the signatures, the absence of double spends, the correctness of the reward and the fact that the block hash is below the current target. That is why a miner cannot write extra coins to himself: a block with the wrong reward is simply discarded.

1. The wallet signs a transactionThe key owner builds the transfer and signs it. Nothing secret goes out to the network at that moment, only the signature and public data are published.
2. The transaction spreads across nodesEvery node checks it against its own rules and puts it in its own waiting pool. The pools on different nodes can differ.
3. The pool builds a block templateThe mining pool picks the transactions with the highest fee per unit of weight and hands out the job to the connected machines.
4. The ASIC searches through hashesThe hardware changes the service fields of the header and computes a double SHA-256 until the result comes out below the target value.
5. The block is publishedThe block once found goes out to the network. Its first transaction creates the reward: a subsidy of 3.125 BTC plus the fees of every transfer included.
6. Nodes accept the blockEvery node checks the block on its own and adds it to the chain. A faulty or greedy block is discarded without discussion.
7. Difficulty is recalculatedEvery 2,016 blocks the network compares the actual time with the expected two weeks and adjusts the target so that a block is found in 10 minutes again.

SHA-256 is a cryptographic hash function that turns data of any length into a 256 bit result. Bitcoin applies it twice in a row to the block header of 80 bytes. The task of a miner is to pick values for the service fields so that the resulting hash comes out below the target. There is no inverse operation, the only way is brute force.

The function needs no memory and consists of simple logic operations, so it is easy to freeze into silicon. Because of that ASIC hardware pushed out everything else as early as 2013. The gap is enormous: a modern machine delivers hundreds of terahashes at an efficiency of about 10 joules per terahash, while a graphics card works at the level of tens of megahashes and spends thousands of times more energy on the same job.

Mining Bitcoin on a graphics card or a processor makes no sense today at any tariff: the income will be indistinguishable from zero while the electricity bill will be entirely real. Any service that promises otherwise is either selling a cloud contract or counting with numbers that do not exist.

A distinct feature of SHA-256 is AsicBoost, a technique that saves part of the computation by reusing the intermediate state of the function. The open variant works by iterating over bits of the version field, it is described in BIP-320 and pools support it as standard. The covert variant is incompatible with SegWit and is not used in practice.

The strengths of the algorithm: verification is extremely simple, the component base is mature and the hardware fleet is the largest in the industry. The weaknesses follow straight from the strengths: entry demands specialised hardware, and competition runs only on the price of a kilowatt hour and on chip efficiency, where being one generation behind already means working at a loss.

Full description of the SHA-256 algorithm →

The issuance cap equals 21,000,000 BTC and it is written into the block validation rules. By the end of June 2026 about 20,047,000 coins had been mined, roughly 95.5 percent of the limit, leaving less than a million. The last satoshis are calculated to be issued around the year 2140.

The block reward halves every 210,000 blocks. The stages already passed: 50 BTC from 2009, 25 BTC from 2012, 12.5 BTC from 2016, 6.25 BTC from 2020 and 3.125 BTC from 20 April 2024, block 840,000. The next halving is tied to block 1,050,000 and will cut the subsidy to 1.5625 BTC. As of the end of July 2026 about 89,600 blocks were left before it, with 2028 as the expected year. Nobody knows the exact date: it depends on how fast blocks are actually found.

Annual issuance at the current reward comes to roughly 164,000 BTC, so supply inflation holds around 0.8 percent a year and keeps falling. There is no coin burning in the protocol: the only proven reduction of supply is one satoshi lost through a quirk of block 124,724. Estimates of the volume of coins lost in wallets without keys turn up often, but no verifiable figure exists.

The second part of the reward is fees. They are meant as a replacement for the vanishing subsidy, but so far they do not work as one: as of mid July 2026 fees delivered 0.69 percent of the total block reward, on average about 0.02 BTC per block. For comparison, in April 2024 on the wave of the Runes launch individual blocks brought in more than 10 BTC in fees, above the subsidy itself. The spread between those two regimes is the main open question in the long term economics of the network.

The full table of SHA-256 models from our catalogue, sorted by efficiency: the lower the joules per terahash, the longer a machine stays profitable as difficulty rises.

Browse miners in the catalogue →

The ranking is built on real efficiency from the catalogue, not on how popular a model is. Payback estimates start from a hashprice of about 30 dollars per petahash a day, recorded in mid July 2026.

The stock Bitmain firmware lets you start a machine and pick one of the preset modes, but it gives no control over voltage and frequency on individual boards. Because of that you cannot bring a particular unit to its own optimum: chips differ even within a single batch.

Third party builds solve exactly that task. Through HashCore Toolkit the supported Antminer models of the S19 and S21 series receive firmware with manual tuning of frequencies and voltages per chain, automatic mode selection and full telemetry for every hash board. The developer fee is 2.8 percent.

The key mode of such builds is AsicBoost. The technique saves part of the SHA-256 work by reusing the intermediate state of the function: bits of the version field in the header are iterated over, not only the nonce. The open variant is described in BIP-320, it is compatible with SegWit and all major pools support it as standard. In our catalogue that support is marked on Bitmain models of the S19, S21 and T21 series.

The economic point of reflashing is direct: at an unchanged tariff, cutting the draw per terahash pushes back the moment when a machine goes into the red. With fees at around 0.7 percent of the reward and a subsidy of 3.125 BTC, a miner is left with almost no other levers beyond the price of electricity and hardware efficiency.

One limitation is worth understanding in advance: reflashing voids the manufacturer warranty, and wrong voltage settings lead to degradation of the boards. Work from a backup of the factory firmware and raise the overclock gradually, watching temperature and the number of rejected shares.

You can only start mining Bitcoin with a SHA-256 ASIC. The minimum kit: the machine itself, its standard power supply, a separate power line, wired internet and a pool account. No node setup is required: the block template comes from the pool.

The electrical supply is the first constraint. A modern air cooled machine draws from 3.1 to 6.6 kW, which means it calls for a separate line and a breaker of 16 to 32 amperes. A household socket does not hold that load, and one breaker for several machines is a direct route to a fire. Hydro models such as the A4 Ultra Hydro, with a draw above 8 kW, also need a ready loop with a heat exchanger.

The second constraint is noise and heat. Industrial air cooled models run at around 75 decibels and above, comparable to a vacuum cleaner left on around the clock. You cannot put such a machine in a living space, and that is not a question of comfort but of conflict with neighbours and with the rules. Hydro and immersion variants are quieter, but they demand infrastructure.

Home mining in 2026 makes sense in two scenarios only: learning on low power devices such as the Avalon Nano 3 at 140 W, and heat reuse, where the machine replaces the heating. As a source of income at a household tariff in Spain a home Bitcoin ASIC is loss making, and that is worth accepting before the purchase rather than after it.

The industrial scenario is built around the tariff. At the hashprice of mid July 2026 revenue by efficiency tier came to about 109 dollars per megawatt hour for machines better than 14 J/TH and about 41 dollars for the range of 25 to 38 J/TH. That difference of two and a half times explains why the old fleet is switched off first.

Miner income is made of four multipliers: the hashrate of the machine, the current network difficulty, the price of the coin and the fees in blocks. The first two decide how many BTC you receive, the other two decide what that is worth. You can control directly only hardware efficiency and the price of a kilowatt hour.

A convenient way to fold all of those values into one is hashprice, revenue per petahash a day. In mid July 2026 it held at about 30.9 dollars per PH/s a day. From there the arithmetic is easy: a machine at 270 TH/s brings in on the order of 8 dollars a day before electricity, while its own draw at 3.645 kW comes to about 87 kilowatt hours over the same day. Anything above roughly 0.09 dollars per kilowatt hour eats that revenue whole.

In 2026 fees stopped being a noticeable addition: their share fell to a level below one percent of the reward. You cannot plan around spikes like the one in April 2024, those are rare episodes. Planning has to start from the subsidy.

Budget separately for two risk factors. The first is difficulty growth: it is recalculated every two weeks and in a rising market phase it eats income faster than the price grows. The second is the halving of 2028, which will cut the subsidy in half in a single block. A machine that works on a thin margin today will turn loss making after it without any change in tariff.

Open the profitability calculator →

For a miner the rule is simple: the address the pool sends payouts to must belong to you completely. An exchange address is convenient, but if the account is blocked or the venue runs into trouble, access to what you mined disappears along with it. A hardware wallet plus a seed phrase written down on paper closes that risk for sensible money.

Pay attention to the address format. The modern formats bc1q (SegWit) and bc1p (Taproot) take up less weight in a block, so spending from them costs less than from addresses of the 1... kind. Make sure the pool supports the format you choose before you save it in the payout settings.

Bitcoin trades on every large venue. As of 31 July 2026 the daily trading volume was around 26.8 billion dollars, the most liquid market in the industry. Among the venues available in the EU the most used are Kraken, Bitstamp, Coinbase, Bitvavo and Bit2Me, and among the global ones Binance, OKX and Bybit.

A separate layer of liquidity is the spot exchange traded funds in the United States, launched in January 2024. By the end of April 2026 their combined assets exceeded 100 billion dollars, the funds held from 1.29 to 1.3 million BTC, that is 6 to 7 percent of the whole supply. Cumulative net inflow since launch stood at 58.72 billion dollars as of 4 May 2026. The inflow for 2026 turned out an order of magnitude smaller than for 2025: around 536 million dollars against 25 billion.

We trade hardware, not the coin. This page is a reference: choose the venue yourself, looking at its licence in your jurisdiction, the depth of its order book and its terms for withdrawal into fiat. Check the withdrawal address character by character and do not keep more on an exchange than you are ready to lose.

For buying hardware the rules are different: look at the warranty term, the presence of service in the EU and at whether the machine ships with factory firmware. Second hand ASIC units often arrive already reflashed, and restoring the stock build can be far from trivial.

The main scenario on the first layer is transferring large amounts and final settlement. A throughput of 3 to 7 transactions per second is not designed for everyday payments, but settlement is final and does not depend on an intermediary. The fee in a calm market is measured in cents, though in periods of load it climbs into tens of dollars.

Small payments have moved to the second layer. Lightning Network as of May 2026 held about 4,898 BTC of public capacity in 41,080 channels across 17,438 nodes. The peak came in December 2025, about 5,637 BTC, after which capacity dropped by roughly 12 percent. Monthly payment turnover passed 1 billion dollars for the first time in February 2026. Since March 2026 stablecoins also travel through the channels via Taproot Assets.

The second large scenario is a reserve asset. In 2025 a strategic bitcoin reserve was set up in the United States, and corporate treasuries keep accumulating the coin: by the count at the end of March 2026 about 145 public companies held it, and the largest holder at that moment had 762,099 BTC. Figures on corporate holders go stale quickly and estimates from different trackers diverge.

A new vector of 2026 is the use of the coin as collateral in traditional lending: in the United States rules are being prepared for accepting crypto assets as security for a mortgage. For a miner this matters indirectly, through demand: the more institutional holding scenarios there are, the steadier the price his revenue is counted from.

The security of the network rests on the cost of an attack. To rewrite history you have to keep out hashing the rest of the network constantly, and that is on the order of 900 exahashes per second and a comparable electricity bill. A 51 percent attack does not allow stealing coins from other users or creating extra ones: the most it allows is reversing your own recent transactions. That is exactly why exchanges wait for confirmations.

The weak point sits not in the hardware but in the way it is organised. Four pools control more than 70 percent of the hashrate and the Nakamoto coefficient equals three. An individual machine still stays with its owner, and switching to another pool takes minutes, which is what limits the risk. In May 2026 pools with roughly 75 percent of the hashrate backed an open standard for block construction that returns the choice of transactions to the miner rather than the pool operator.

The second line of defence is the nodes. About 24,500 reachable nodes check every block independently, and nobody can change the validation rules unilaterally. The dispute of 2025 around Bitcoin Core version 30 and the OP_RETURN limit showed it plainly: the operators who disagreed simply moved to the alternative Knots build, whose share passed 21 percent of nodes.

On a horizon of several years the quantum threat is under discussion. A study of March 2026 lowered the estimate of the qubits required to break the signature scheme to below 500,000, which spurred work on the post quantum proposal BIP-360. There is no practical danger today, but that is no reason to reuse addresses.

The price is quoted at the all-time high. The value of about 126,000 dollars is often passed off as the current level. That is the high of 6 October 2025. On 31 July 2026 the price held in a range of roughly 63,000 to 65,000 dollars, with a change over the year of minus 45.1 percent.

The block reward is written as 6.25 BTC. That value was in force until 20 April 2024. From block 840,000 the subsidy equals 3.125 BTC. The error turns up even in calculators, and it overstates the calculated income exactly twofold.

Hashrate is rounded up to one zettahash. The wording about 1 ZH/s or 1,000 EH/s has spread across reference sites. According to the summary of 13 July 2026 the 7 day average was 879 EH/s, down 6.4 percent over the week, and the 30 day average 938 EH/s. The trend is downward, not upward.

Difficulty is given at the level of 136 trillion. An outdated value. The retarget of 11 July 2026 cut difficulty by 5 percent to 127.17 trillion, and by the end of the month the figure was about 126.23 trillion. Over 30 days difficulty fell by roughly 5.7 percent.

Daily transactions are given as a range of 300 to 600 thousand. That range reflects the situation of past years. By the measurement of 23 May 2026 the network was passing more than 730,000 transactions a day, a rise of about 65 percent against the year before.

The fee share is described as 1 to 5 percent of the reward. As of 13 July 2026 fees delivered 0.69 percent of the total reward, on average about 0.02 BTC per block, and that was 18 percent less than a week earlier. Planning income around fees today is not possible.

The date of the halving of 2028 is presented as exact. Concrete dates such as the middle of April 2028 are an estimate obtained by multiplying the remaining blocks by ten minutes. The halving is tied to block 1,050,000, not to the calendar. At the end of July 2026 about 89,600 blocks were left.

Lightning capacity is described as growing without pause. As of May 2026 public capacity was about 4,898 BTC against the peak of 5,637 BTC in December 2025, a decline of roughly 12 percent. The node count is about 17,438 and the channel count about 41,080.

The node count is taken from old snapshots. The figure of 16 to 17 thousand reachable nodes has been drifting through reference sites from past years. The estimate for 2026 is about 24,500 reachable nodes, and counting the unreachable ones sources name from 50,000 to 100,000.

The Taproot share is shown as growing linearly. Taproot usage reached roughly 42 percent in 2024, and by the end of 2025 it had fallen to roughly 20 percent. SegWit meanwhile settled at about 85 percent of transactions.

The forks of August 2026 are called Bitcoin upgrades. As of mid July 2026 the BIP-110 soft fork had miner support below 1 percent, so its activation is unlikely. The eCash hard fork at block 964,000 is a separate chain, not an upgrade of the main network. Presenting them as planned protocol upgrades is wrong.

The efficiency of new machines is understated or overstated. The claimed 9.5 J/TH belongs to the Antminer S23 generation, released in 2026. Lower values do appear in reviews, but there is no confirmation for them. In our catalogue the best figure is 9.45 J/TH on the SealMiner A4 Ultra Hydro.