The coinbase combines new units and fees
Every valid block begins with a special coinbase transaction. Unlike an ordinary transaction, it spends no older UTXO and creates an output for the miner or pool. Its maximum value consists of the protocol-permitted block subsidy plus the sum of fees from transactions included in the block.
Only the subsidy represents newly issued satoshis. Fees are existing value that ordinary transaction inputs did not assign to their outputs. A miner may claim less, but if the coinbase pays more than the permitted subsidy and fees, Bitcoin Core marks the block invalid.
A halving follows blocks, not the calendar
The subsidy began at 50 BTC per block. After every 210,000 blocks, its satoshi value is shifted down by an integer half. At the ten-minute target average, one period is roughly four years, but the exact date of the next halving is not fixed in a calendar because block times vary.
The reductions continue until integer rounding makes the result zero. BIP 42 removed an old technical bug that could have produced ambiguous behavior in some implementations after a very large number of shifts. The current rule explicitly returns a zero subsidy after 64 intervals.
- 50 BTC was the initial subsidy, not a permanent reward
- the halving interval is 210,000 blocks
- once the subsidy reaches zero, transaction fees remain as miner revenue
Twenty-one million is an accurate shorthand with technical details
The geometric sum of the scheduled subsidies creates the ceiling commonly described as 21 million BTC. Because the protocol uses whole satoshis and discards remainders at each halving, the sum of subsidies actually permitted is slightly below exactly 21 million. In addition, the genesis block's output cannot be spent in the ordinary way.
Lost keys, deliberately destroyed coins and unclaimed rewards further reduce the available stock but do not change the issuance limit. Copying the code or creating another branch does not add units to the bitcoins your node accepts; it produces a separate history or a different asset.
The rule is fixed only while participants keep it
No law of physics prevents people from writing software with a different issuance schedule. Today's nodes would reject such a block, however. Raising the limit would require users to knowingly adopt incompatible rules; those who refuse could remain on the original network. The limit is therefore a technical and social commitment, not a number immutable outside human decisions.
A long-term open question is the security budget as the subsidy declines. The protocol anticipates a growing role for fees, but Bitcoin's future price, demand for block space and the hash rate needed for security cannot be guaranteed. Finite supply addresses issuance predictability; by itself it guarantees neither price nor future security.