At 00:09:27 UTC on 20 April 2024, a miner found block 840,000 and the reward for creating new bitcoin fell from 6.25 to 3.125 BTC. That was the fourth bitcoin halving. The miner of that block still collected 40.75 BTC, because users paid 37.63 BTC in fees to get into it, according to mempool.space block data. The fees alone were twelve times the new subsidy. A token protocol called Runes launched at that exact height, and for one block the most watched supply event in crypto was mostly a fee auction.
Two and a half years later the picture is ordinary again. Block 975,000, mined at 23:36 UTC on 7 November 2026, paid its miner 3.125 BTC of new coins plus 0.034 BTC of fees. New coins made up almost 99 percent of that block’s reward.
Both blocks run the same rule. Every 210,000 blocks, the number of new bitcoins a miner may create is cut in half, and no single party can schedule or delay it. Since the genesis block paid 50 BTC in January 2009, a story told in our lesson on the genesis block and the 2008 financial crisis, the reward has been halved four times. This guide follows that rule from the source code to the three questions people ask about it: how new bitcoin is paid out, why the total stops just under 21 million, and what the halving does and does not do to the price.
The bitcoin halving is one line of code
Bitcoin Core, the software most nodes run, computes the subsidy in a function called GetBlockSubsidy in the file validation.cpp. This is the version on the project’s main branch as read on 8 November 2026:
CAmount GetBlockSubsidy(int nHeight, const Consensus::Params& consensusParams)
{
int halvings = nHeight / consensusParams.nSubsidyHalvingInterval;
// Force block reward to zero when right shift is undefined.
if (halvings >= 64)
return 0;
CAmount nSubsidy = 50 * COIN;
// Subsidy is cut in half every 210,000 blocks which will occur approximately every 4 years.
nSubsidy >>= halvings;
return nSubsidy;
}Read it from the top. The function receives a block height, the position of a block in the chain. It divides that height by the halving interval, which is 210,000 on the main Bitcoin network, and keeps only the whole number. Block 839,999 gives 3 (three halvings so far); block 840,000 gives 4. That integer is the count of halvings that have already happened.
Next comes the starting reward: 50 times COIN. In Bitcoin’s code a coin is not a decimal number. It is 100,000,000 units called satoshis, so the starting reward is the whole number 5,000,000,000. Working in whole satoshis avoids the rounding errors that decimal arithmetic would bring into a system where every node must reach exactly the same total.
The line nSubsidy >>= halvings is a right shift. For a non-programmer, shifting a binary number one place to the right is the same as dividing it by two and dropping any remainder. Shift once and 5,000,000,000 satoshis become 2,500,000,000 (25 BTC). Shift four times, as the code has done since block 840,000, and you get 312,500,000 satoshis, the 3.125 BTC paid today.
Two details matter later. First, dropping the remainder means the reward is always rounded down, never up. Second, the check if (halvings >= 64) exists because shifting a 64-bit number by 64 places or more is undefined in C++. The comment says so in plain words: the guard forces the reward to zero. In practice the subsidy reaches zero long before that, as the supply table below shows.
The interval itself lives in a separate file, chainparams.cpp. It is 210,000 for the main network and for the public test networks. On regtest, a private network developers spin up on their own machine, it is 150 blocks, so they can watch a halving happen in seconds while testing wallets or mining software.
One constant often gets mistaken for the cap. The file amount.h defines MAX_MONEY as 21,000,000 BTC. Its own comment warns that this constant is “not the total money supply”. It is a sanity check: no transaction output may claim more than that amount. The real limit on how many bitcoins can exist comes from the halving function, not from a number written at the top of a file.
That check has a history. On 15 August 2010, users found that block 74,638 contained a transaction creating about 184 billion bitcoins through a number overflow bug, according to the Bitcoin Wiki’s account of the value overflow incident. A fixed client came out within five hours and the chain was rebuilt without that block. The episode shows what the supply rule really is: a check that every node runs on every block, and that only works because nodes refuse blocks that break it.
Subsidy plus fees: what a miner is actually paid
The first transaction in every block is special. Called the coinbase transaction, it has no sender: it pays the miner the block reward, which is the subsidy computed above plus the fees of every transaction in the block. Fees are the difference between what a transaction spends and what it sends, and the miner who includes it keeps that difference. A node rejects a block whose coinbase claims more than subsidy plus fees. A miner may claim less, and a few have, by mistake.
The two parts of the reward behave very differently. The subsidy is fixed by height and known decades in advance. Fees depend on how many people want to get a transaction into the next block, which our guide to the mempool and transaction fees explains in detail. Over the 144 blocks up to height 975,046 (roughly the last day, as of 09:45 UTC on 8 November 2026), miners earned 452.82 BTC in total, of which 2.82 BTC came from fees, or 0.62 percent, per mempool.space reward statistics.
How that reward turns into a business, with pools, payout schemes, machines and electricity, is covered in our bitcoin mining guide. For this lesson one sentence is enough: a halving cuts a miner’s income in BTC per block by close to half overnight, while costs stay the same.
Why blocks, not dates, trigger a halving
Nothing in the code mentions a calendar. Bitcoin aims for one block every ten minutes on average, so 210,000 blocks take about 2,100,000 minutes, which is close to four years. That is where the “every four years” in the source comment comes from, and it is only an average.
Real block times drift because the hashrate, the total computing power of miners, keeps changing, and the difficulty only adjusts every 2,016 blocks to bring the pace back towards ten minutes. Our lesson on hashrate and mining difficulty explains that loop. In practice blocks have come slightly faster than ten minutes since the last halving: between block 840,000 (20 April 2024, 00:09 UTC) and block 975,046 (8 November 2026, 09:45 UTC), the average was about 9 minutes 56 seconds, by our calculation from mempool.space block timestamps. That small gap is why halving dates arrive a few days or weeks early.
Four halvings since 2012: dates, blocks and rewards
Each halving happened at an exact height, and the timestamps below come from the blocks themselves as recorded by mempool.space. Daily issuance assumes the target of 144 blocks a day.
| Halving | Block height | Date and time of the block (UTC) | Subsidy before | Subsidy after | New BTC per day after (144 blocks) |
|---|---|---|---|---|---|
| 1st | 210,000 | 28 November 2012, 15:24:38 | 50 BTC | 25 BTC | 3,600 |
| 2nd | 420,000 | 9 July 2016, 16:46:13 | 25 BTC | 12.5 BTC | 1,800 |
| 3rd | 630,000 | 11 May 2020, 19:23:43 | 12.5 BTC | 6.25 BTC | 900 |
| 4th | 840,000 | 20 April 2024, 00:09:27 | 6.25 BTC | 3.125 BTC | 450 |
| 5th (next) | 1,050,000 | Estimated around April 2028 | 3.125 BTC | 1.5625 BTC | 225 |
The 2012 halving came when mining was still done largely by hobbyists, and the 2016 one was the first with real economic weight behind it: bitcoin traded around 650 dollars and mining had moved to ASIC machines, as the Cambridge Centre for Alternative Finance (CCAF) and Fii recall in their October 2026 report Funding Bitcoin’s Future. The 2020 halving fell in the middle of the COVID-19 pandemic.
The 2024 halving is the one that taught the most about fees. The CCAF report puts the fees of block 840,000 at about 2.4 million dollars, roughly 92 percent of its total reward, and says fees exceeded the new 3.125 BTC subsidy for 104 consecutive blocks around the Runes launch. On Coin Metrics’ daily figures cited in the same report, fees were 74.8 percent of miner revenue on 20 April 2024. The spike did not last: the daily mean fee per transaction fell about 73 percent the next day, and by 4 May it was down to about 4.5 dollars.
Why 21 million, and why the real total is 20,999,999.9769
The 21 million figure is not written into the rules as a cap. It falls out of the schedule. In the first era, 210,000 blocks of 50 BTC create 10.5 million coins. Each later era creates half as much as the one before: 5.25 million, then 2.625 million, and so on. Add up a series that keeps halving and the sum gets closer and closer to twice the first term, so the total approaches 2 x 10.5 million, which is 21 million.
Satoshi Nakamoto stated the number from the start. In the 8 January 2009 email announcing the first release, posted to the cryptography mailing list, the message reads: “Total circulation will be 21,000,000 coins.” It goes on to say the coins will be distributed to nodes when they make blocks, “with the amount cut in half every 4 years”, and lists the first eras: 10,500,000 coins, then 5,250,000, then 2,625,000.
The figure came back in the announcement of version 0.3 on 6 July 2010: “Bitcoin’s total circulation is limited to 21 million coins.”
What the record does not contain is a reason. No explanation by Satoshi can be found in his published emails and forum posts for why the starting reward is 50 BTC, why the interval is 210,000 blocks or why the total should be 21 million rather than 10 or 100 million. The Bitcoin Wiki puts it this way: “Satoshi has never really justified or explained many of these constants.” The whitepaper offers an analogy rather than a calculation: its section on incentives compares the steady addition of new coins to gold miners spending resources to add gold to circulation, according to the text of the paper. The explanations that circulate, such as a match with the four-year halving rhythm or a total number of satoshis close to the capacity of a 64-bit floating-point number, are labelled speculation on the same wiki page. Unless a text by Satoshi turns up, they remain guesses.
The integer arithmetic adds a twist. Because each era’s subsidy is a whole number of satoshis rounded down, the series does not reach 21 million. We summed 210,000 times the subsidy of every era in the code and got 2,099,999,997,690,000 satoshis, which is 20,999,999.9769 BTC. The 2,310,000 satoshis between that and 21 million (0.0231 BTC) will never exist. The Bitcoin Wiki’s supply page gives the same figure.
| Era | First block | Subsidy per block | BTC created in the era | Total at the end of the era | Share of 21 million |
|---|---|---|---|---|---|
| 0 | 0 | 50 BTC | 10,500,000 | 10,500,000 | 50% |
| 1 | 210,000 | 25 BTC | 5,250,000 | 15,750,000 | 75% |
| 2 | 420,000 | 12.5 BTC | 2,625,000 | 18,375,000 | 87.5% |
| 3 | 630,000 | 6.25 BTC | 1,312,500 | 19,687,500 | 93.75% |
| 4 (today) | 840,000 | 3.125 BTC | 656,250 | 20,343,750 | 96.875% |
| 5 | 1,050,000 | 1.5625 BTC | 328,125 | 20,671,875 | 98.4375% |
| 6 | 1,260,000 | 0.78125 BTC | 164,062.5 | 20,835,937.5 | 99.22% |
| 32 (last) | 6,720,000 | 1 satoshi | 0.0021 | 20,999,999.9769 | about 100% |
| 33 and after | 6,930,000 | 0 | 0 | 20,999,999.9769 | 100% |
The table is computed from the code’s rule and ignores coins that were never claimed. Era 32 pays one satoshi per block, the smallest unit there is. At block 6,930,000 the 33rd halving shifts that single satoshi away and the subsidy becomes zero for good. When will that be? The Bitcoin Wiki estimates the last bitcoin will be mined around 2140, and its May 2140 estimate assumes mining power stays constant from block 367,500 onwards. Nobody can date a block a century ahead with any precision.
The real total will be a little lower than 20,999,999.9769 BTC, because some rewards were never created or can never move. The 50 BTC of the genesis block cannot be spent. In block 124,724 a miner claimed one satoshi less than allowed and forfeited the block’s fees. In block 501,726 the miner did not claim the 12.5 BTC reward at all, and mempool.space records a reward of zero for that block. Two early blocks reused a previous coinbase identifier, which made an earlier reward unspendable, according to the Bitcoin Wiki.
How many bitcoins exist on 8 November 2026
Blockchair’s Bitcoin statistics showed 20,109,492.92 BTC in circulation at block 975,046, mined at 09:45 UTC on 8 November 2026. That is 95.76 percent of 21 million. About 890,500 BTC remain to be issued, spread over more than a century, yet about 63 percent of them will be created before block 1,260,000, the sixth halving, expected around 2032. New issuance now runs at about 450 BTC a day, which is roughly 0.82 percent of the existing supply per year by our calculation.
Put differently, most of the work of the halving schedule is already done. Each future halving removes half of a flow that is already small compared with the stock. The fifth will cut yearly issuance from about 164,000 BTC to about 82,000 BTC; the sixth, around 2032, to about 41,000 BTC.
Lost coins shrink the usable supply
Coins that exist are not all coins that can move. Private keys get lost, and some early miners abandoned wallets that were worth little at the time. Nobody can tell from the chain whether an old coin is lost or simply held. In a market analysis of 18 March 2020, Chainalysis worked with an estimate of 3.75 million lost bitcoin out of the 18.27 million mined at the time, leaving 14.52 million it called available. That is an estimate from one firm, made more than six years ago, not a measurement.
Satoshi had a short answer to the worry, in a forum thread called “Dying bitcoins” on 21 June 2010: “Lost coins only make everyone else’s coins worth slightly more. Think of it as a donation to everyone.”
For a holder the practical lesson is simpler. The supply limit is enforced by code; whether your own coins stay part of the usable supply depends on how you keep your keys.
The next bitcoin halving: block 1,050,000, around April 2028
The fifth halving will happen at block 1,050,000, when the subsidy falls from 3.125 to 1.5625 BTC. That height is certain. The date is not, and any countdown that shows a precise minute is an estimate that moves with every block found.
At 09:45 UTC on 8 November 2026 the chain stood at block 975,046, so 74,954 blocks remained. Here is what that gives at different average block times, by our own calculation:
| Average time per block from now | Estimated date of block 1,050,000 (UTC) |
|---|---|
| 9 min 56.5 s (pace since April 2024) | about 8 April 2028 |
| 10 min exactly (the target) | about 11 April 2028 |
| 10 min 15 s | about 24 April 2028 |
| 10 min 30 s | about 7 May 2028 |
The honest statement is therefore “around April 2028, possibly early May”. If the hashrate keeps growing, blocks come a little faster than ten minutes on average and the date moves earlier. If miners switch machines off, for example when a halving squeezes their margins, blocks slow until the next difficulty adjustment and the date slips. Online trackers using their own averages differ by days to weeks for the same reason.
After block 1,050,000, issuance drops to about 225 BTC a day. By then roughly 20,343,750 BTC will have been created according to the schedule (a little less once unclaimed rewards are subtracted), and yearly issuance will fall to about 0.4 percent of the supply.
Is 2028 the last halving? No. Four have happened, and 28 more will still leave a non-zero subsidy, the last of them (the 32nd) setting it at a single satoshi per block. The 33rd halving, at block 6,930,000, brings it to zero. The code’s 64-halving guard is a safety net that the schedule never reaches in practice.
Who pays for security when the subsidy fades
Miners spend real electricity to produce blocks, and that spending is what makes rewriting Bitcoin’s history expensive. The money that pays for it is sometimes called the security budget. Today it comes almost entirely from new coins, and the halving schedule shrinks that source on a fixed timetable.
Satoshi described the end state in section 6 of the whitepaper: once a set number of coins is in circulation, “the incentive can transition entirely to transaction fees and be completely inflation free.” Our guide to the Bitcoin whitepaper places that sentence in the rest of the design. The question is whether fees will be large enough when that happens.
The CCAF report published in October 2026 is the most complete recent study of that question. It finds that fees made up 0.67 percent of miner revenue over the 90 days to 4 October 2026, and that daily changes in fees were 11.7 times as volatile as changes in the price over that window. It quotes Bitcoin Core developer James O’Beirne, speaking in April 2025, warning that “two halvings from now you’re going to have a pretty drastically reduced subsidy”, and MARA Holdings chief executive Fred Thiel, quoted by CoinDesk in November 2025: “If bitcoin doesn’t grow at 50% or more annually, the math gets very tough after 2028.”
The other side has evidence too. In the same report’s halving-era averages, the daily subsidy measured in dollars rose from about 39,000 dollars in 2009-2012 to about 37.8 million dollars in 2024-2026, because the price rise more than offset each cut in BTC. The authors add that the empirical studies they cite have not found the fee-based instabilities that theory predicts on Bitcoin’s main network, and that none appeared in the fee-dominated blocks around the 2024 halving.
The arithmetic behind the debate is simple. To keep the dollar value of the subsidy flat across one halving, the price must double over about four years, which the CCAF puts at roughly 19 percent a year. If it does not, fees have to fill the gap. Whether they will is open, and the answer will not be known for decades. A cheaper network to attack is the worry at the end of that road; 51 percent attacks are the subject of a later lesson of this course.
Could the 21 million limit simply be raised? Only by changing the consensus rules that every node enforces, which would require node operators, miners and users to accept the new software. The CCAF notes that changes to issuance face additional consensus requirements. No such change has been adopted as of November 2026.
Bitcoin price after each halving: four data points, no rule
Before every halving, the same question comes back: what does it do to the price? The data below come from the mempool.space price history, in US dollars. For the first three halvings that history has one data point per week, so the point used is the nearest one on or before each date, and its date is shown; the 2024 points are taken at 12:00 UTC. These are prices at a given moment, not highs or lows.
| Halving | Price near the halving | About 6 months later | About 1 year later | About 18 months later |
|---|---|---|---|---|
| 28 Nov 2012 | $12.48 (26 Nov 2012) | $133.5 (27 May 2013) | $1,079.9 (28 Nov 2013) | $575.8 (22 May 2014) |
| 9 Jul 2016 | $655.1 (7 Jul 2016) | $784.6 (5 Jan 2017) | $2,372.1 (6 Jul 2017) | $15,185.7 (28 Dec 2017) |
| 11 May 2020 | $9,317.1 (7 May 2020) | $15,708.7 (5 Nov 2020) | $49,502.7 (6 May 2021) | $62,942.9 (28 Oct 2021) |
| 20 Apr 2024 | $63,730 (20 Apr 2024, 12:00 UTC) | $66,854 (17 Oct 2024) | $84,128 (20 Apr 2025) | $111,914 (12 Oct 2025) |
In the first three cycles the price was far higher a year later: about 86 times higher after 2012, 3.6 times after 2016, 5.3 times after 2020. The 2024 cycle was much flatter, up about 32 percent at the one-year mark. The same series showed 124,277 dollars at 12:00 UTC on 6 October 2025 and stood at 75,897 dollars at 12:00 UTC on 7 November 2026, about 39 percent below that October 2025 point. Percentages that circulate online for “the price after the halving” vary widely because each source picks a different starting day and end day; always check both.
There are good reasons not to read a law into this table. A rise that follows a halving is a correlation; it does not show that the halving caused it. Four observations are too few to separate one cause from everything else that moved at the same time, from the COVID-19 pandemic around the 2020 halving to new buyers and shifts in interest rates in other cycles. The schedule is public years in advance, so a market that expects higher prices from a known cut has every chance to move early. And the share of supply affected keeps shrinking: the 2012 halving cut yearly issuance from about 25 percent of the existing supply to about 12.5 percent (at 144 blocks a day), while the 2028 one will trim it from about 0.8 percent to about 0.4 percent.
The best-known attempt to turn halvings into a price formula is the stock-to-flow model, published in 2019 by a pseudonymous analyst known as PlanB. Stock-to-flow divides the existing supply (stock) by the new supply per year (flow); each halving doubles the ratio, and the model mapped that ratio to a price path. By June 2022 the model’s projection of more than 100,000 dollars had not materialised, and bitcoin had fallen below 20,000 dollars, as Decrypt reported on 21 June 2022. Ethereum co-founder Vitalik Buterin wrote that “the ‘halvings cause BTC price rises’ theory is unfalsifiable”. PlanB answered that the original model “certainly had a good run” from March 2019 to March 2022.
A model that only knows supply leaves out demand, and price comes from both. Nothing in this lesson is a forecast or a recommendation to buy or sell.
For readers in the United States, the halving itself creates no tax event for a holder: nothing is received. The IRS treats digital assets as property, not currency, according to its digital assets page, and what matters for tax is when you sell, exchange or otherwise dispose of them. Our guides to crypto taxes in the United States and crypto taxes in the United Kingdom cover the details.
Halving myths that do not survive the code
A few claims come back before every halving. Each one can be checked against what the code and the chain show.
- The halving does not change the ten-minute block target. It changes only how many new coins each block may create.
- Hashrate does not fall because of the halving alone. Some miners may switch off machines that are no longer profitable, and the difficulty adjustment then rebalances block times.
- Halvings are not exactly four years apart. They come every 210,000 blocks, which has meant a little under four years so far.
- The halvings do not end with a final event in 2140. The subsidy shrinks to one satoshi per block in its final era and then to zero at block 6,930,000, by which point issuance is already negligible.
- A price rise is not guaranteed. Four past cycles are a pattern of four, with very different backgrounds and a much weaker first year after 2024.
If you want to see the rule at work, open any block on a block explorer and read its coinbase transaction: the new coins it creates will be 3.125 BTC until block 1,049,999, and 1.5625 BTC from block 1,050,000. Our glossary entry on the bitcoin halving gives the short version, and the bitcoin coin page shows the live supply and market data.
