Mining, difficulty and the arms race
It began on laptops and ended in warehouses near power stations. The competitive dynamics were entirely predictable and are worth understanding as economics.
A miner is paid for producing a valid block. The payment has two parts: newly created coins, called the subsidy, and the fees attached to the transactions included.
The hardware progression
- CPUs. Ordinary computers, in the earliest period. Anyone could participate meaningfully.
- GPUs. Graphics cards perform the same simple operation in parallel, which is exactly what hashing needs.
- FPGAs. Reconfigurable chips, briefly.
- ASICs. Chips manufactured to do nothing but compute SHA-256. Orders of magnitude more efficient, and worthless for anything else.
Once the reward is large enough to justify custom silicon, custom silicon appears, and general-purpose hardware becomes uncompetitive permanently. The early vision of mining on ordinary computers ended not because of a policy decision but because that is what competition does to any process with a fixed reward and a variable cost. Nothing in the protocol prevented it.
The economics
A miner's margin is the value of the reward minus the cost of the electricity and hardware consumed producing it. Because difficulty adjusts to hold the block interval constant, adding capacity does not increase total rewards — it divides the same rewards among more machines.
If the total reward per day is worth $X and difficulty adjusts to whatever capacity shows up, then capacity expands until the marginal miner's daily cost approaches $X ÷ (their share). In a competitive market with no barriers, mining profit tends toward the cost of the cheapest available electricity. This is why miners cluster around stranded hydro, flared gas and surplus grid capacity — it is the only durable advantage available.
Pools
A single small miner might find a block once a decade, which is an intolerable income pattern. Pools solve this: many miners combine their capacity, and the pool distributes rewards proportionally to the work each contributed.
This creates a concentration concern that is often overstated and is not nothing. A pool coordinates block construction, but the individual miners can leave instantly if the operator misbehaves — and historically, when a pool has approached a majority share, miners have left voluntarily. The constraint is reputational and it has held so far, which is a weaker guarantee than a protocol rule.
The subsidy runs down
It began at 50 coins per block, and after the halving of April 2024 it is 3.125. Roughly every four years it halves again. At some point it rounds to nothing, and transaction fees must pay for the entire security budget. Whether fees alone can sustain sufficient security is a genuinely open question, acknowledged by the technically literate on all sides of the subject. See the supply schedule.
Hashrate as an indicator
Total network hashing power is published continuously and is sometimes cited as a measure of network health. Read it carefully: hashrate follows price with a lag, because miners expand when margins are good and switch off when they are not. It is largely a consequence of price rather than a predictor of it.