What is a hashing algorithm? Mining algorithms, and the machines married to them

A hashing algorithm takes any input and produces a fixed-length fingerprint — change one character of the input and the whole fingerprint changes unrecognisably, and there is no way to run it backwards. Mining is the brute-force use of that one-way property: your machine guesses trillions of inputs per second until one produces a fingerprint below the network's target. Find one, win the block.

That is the definition. The part that costs people money is what follows from it: every coin picks one algorithm, every ASIC is built for one algorithm, and the two are married for life. A Bitcoin miner can never mine Kaspa. A Kaspa miner can never mine anything but Kaspa and its kin. When you buy an ASIC you are not buying a computer — you are buying a bet on one algorithm's economy, welded into silicon.

So this page is organised the way the decision actually works: every algorithm with a functioning ASIC economy in 2026, what mining it currently pays, and which machines serve it. Rates verified against live network data in September 2026.

Why different algorithms exist at all

Bitcoin chose SHA-256 in 2009. Nearly every algorithm since was designed to be hostile to the specialised hardware that conquered SHA-256 — Scrypt added memory requirements, Ethash made memory bandwidth the bottleneck, Equihash bound performance to memory capacity, RandomX went furthest and required a general-purpose CPU.

The scoreboard after fifteen years: the hardware won every round. Scrypt ASICs arrived by 2014. Ethash ASICs by 2018. Equihash the same year. And in 2026 Bitmain ships the Antminer X9 — a commercial ASIC for RandomX, the algorithm specifically engineered to make ASICs impossible. ASIC-resistance turned out to be a delay, never a defence. The practical lesson for a buyer: judge an algorithm by the health of the coins on it, not by its design ideals.

Every algorithm with a live ASIC economy, and what it pays

Do not compare those payout numbers to each other directly — a terahash of SHA-256 and a terahash of kHeavyHash are entirely different quantities of work from entirely different silicon, and the electricity behind each unit differs just as much. The number that makes algorithms comparable is the break-even power price of the best machine on each, which is what our profitability ranking computes across everything we stock.

Two details in the table repay attention. Scrypt is really a Dogecoin economy — merge-mining means one hash earns LTC and DOGE simultaneously, and DOGE is about 85% of the revenue; our Scrypt guide unpacks that. And Equihash's rate looks small until you notice Zcash's 2026 price run made the cheapest Equihash hardware the fastest-payback machines we list.

The algorithms that died — and what they teach

Every dead algorithm is the same story: the machine outlived its market. Silicon lasts five years; a coin's mining economy carries no such warranty. Ethash is the cautionary giant — billions of dollars of hardware lost its purpose at a scheduled software upgrade, and the survivors now crowd onto Ethereum Classic, one chain doing the work of two. This is the risk you accept with any single-coin machine, it is priced into their faster paybacks, and it is why our used-hardware guide treats old altcoin miners more harshly than old Bitcoin miners.

Bitcoin's SHA-256 sits at the other pole: fifteen years, one algorithm, zero changes, and by far the deepest liquidity behind the coin it secures. You pay for that safety in longer paybacks — the barbell between SHA-256 safety and altcoin-ASIC upside is the real portfolio decision, and our buyer's guide works through it.

How the guess actually works, in thirty seconds

Your miner assembles a candidate block header, hashes it, and checks whether the result is below the difficulty target — effectively, whether the fingerprint starts with enough zeros. If not, it changes a throwaway number called the nonce and hashes again. An S21 XP repeats this 270 trillion times per second. Difficulty self-adjusts so that, worldwide, someone succeeds about every ten minutes on Bitcoin — one lottery draw per hash, no memory, no progress, no "getting closer," which is exactly the maths behind solo-mining odds in our one-bitcoin guide.

The same property is why a hash proves work was done: there is no shortcut to a winning fingerprint except raw guessing, so a valid block is receipts for electricity spent. That conversion rate between electricity and expected coins is the entire economics of mining — quantified as hashprice, which has its own guide.

Choosing an algorithm as a buyer

Start from your power price, not the algorithm. At domestic rates nothing here is profitable and the algorithm question is moot — the fix is cheaper power, not a cleverer coin. Then check the coin's liquidity — daily volume in the hundreds of millions (BTC, LTC/DOGE) means you can always sell what you mine; a few million (ALPH) means you are also making a market bet. Then buy the efficiency leader for that algorithm, because when revenue-per-unit falls — and on every algorithm, it structurally does — the least efficient machines die first.

And hold the one-way rule in mind at the moment of purchase: the coin can abandon your machine, but your machine can never abandon the coin.

Rates derived from live network data (mempool.space, Minerstat) during September 2026 and cross-checked against our earnings model; they move daily with coin prices and difficulty. Algorithm histories from project documentation and public record. Mining income is never guaranteed.