Bitcoin's issuance falls by half at regular intervals, and the change applies to every miner simultaneously at a predetermined block height. The effect on mining economics is immediate and unusually abrupt.
Issuance follows a schedule, not conditions
Each block creates a fixed number of new coins, and that subsidy halves after a set number of blocks have been produced.
The schedule is written into the rules and does not respond to price, demand or the amount of hardware active on the network.
It is also what produces the supply cap, since a sequence of halvings converges on a finite total.
Revenue falls while costs do not
A miner's income is the subsidy plus transaction fees. Halving the subsidy removes a large share of that revenue overnight.
Electricity contracts, hosting agreements, staff and debt payments continue unchanged, so the entire reduction lands on the margin.
Operations running close to break-even before the change are not running above it afterwards, and the date on which this happens has been known to everyone for years.
Preparation is possible in principle, though it mostly consists of replacing hardware and renegotiating power, both of which take capital that thin margins do not generate.
Difficulty adjusts, but only after the fact
Unprofitable machines are switched off, which lowers total hash rate and slows block production until the next difficulty adjustment.
That adjustment raises the reward per unit of hash rate for whoever remains, partially restoring margins for surviving operations.
The correction happens at intervals rather than immediately, so there is a period during which the network operates with reduced capacity and slower blocks.
Efficiency determines who survives
The variable that decides which machines stay on is cost per unit of computation, which combines hardware efficiency with the price of power.
Operations with cheap long-term power contracts continue through periods that force higher-cost competitors offline.
Each halving therefore concentrates capacity among the most efficient operators and accelerates the retirement of older hardware.
Machines taken offline are not always scrapped. They frequently relocate to regions where power is cheaper, which spreads capacity geographically rather than removing it.
Fees become a larger share over time
As the subsidy shrinks towards nothing across successive halvings, transaction fees must eventually fund the network's security entirely.
Fee revenue is far more variable than the subsidy, since it depends on demand for block space at any given moment rather than on a schedule.
How a security budget behaves when it rests on that variable income is an open question about the design, and it is why fee market development attracts as much attention as it does.