A machine's price is what it settled to, not what it launched at

Every entry price on this site is a settled price: the median of a machine's first six months on the market, or the median after a large early drop. Almost nobody paid the launch number, so we do not quote it. Here is the rule, and what it means when a review says a machine is at the best entry price it has ever had.

An antique brass two-pan beam balance, perfectly level, with a row of graduated weights arranged in front of it on a wooden counter

The number on the announcement

On the day a new machine is unveiled it has exactly one number attached to it: the manufacturer's list price. That is the figure that goes into every announcement, every recap and every comparison table that appears within hours of the reveal. It is also the smallest amount of information anyone has about what the machine will actually sell for once it exists in quantity.

Launch pricing behaves the way most new hardware pricing behaves. Supply is thin in the first weeks, the buyers ordering first are often paying for the right to be first as much as for the box, and a manufacturer can charge more for a machine nobody else can get yet. A launch batch is also usually the smallest one a generation will ever ship in, so the handful of transactions behind that first number carry far less weight than the thousands that follow once production is running at scale.

None of that makes the launch number fraudulent. A handful of real buyers really did pay it. It simply is not the number that describes what owning the machine costs, because almost nobody else paid it, and a figure that describes a handful of buyers has no business standing in for what a generation actually sells for.

Almost nobody buys a machine the week it is announced.

What the price does in the months after

From there the price usually falls, and it rarely falls in a straight line. Supply catches up with demand over the following weeks, wholesalers who over-ordered start discounting to clear stock, and the market finds a level. Sometimes that fall is gentle. Sometimes it is a step: a rival factory ships a cheaper, similarly efficient generation into the same window, or a manufacturer runs a promotion to move a batch, and the price drops sharply rather than sliding.

Either way, the price settles. It reaches a level that holds for a stretch, roughly stable, before the next generation starts pulling both attention and buyers away and the machine begins its long, slow slide into being an old one. Our miner price chart shows this happening across the whole catalogue: a spike at launch on almost every generation, then a plateau that is nothing like the spike.

That plateau, not the spike, is where our own buying happens. It is why our reviews and field reports so often open with a line like "bought once its price settled" rather than a date tied to the announcement: waiting those first weeks or months out is not caution for its own sake, it is buying the same hashrate once the market has stopped paying a premium for being first.

The rule: the median of the first six months

Every entry price quoted on this site — in a review, a field report, a progress report — is what we call the settled price: the median of the machine's price over its first six months on the market.

We use the median rather than the average on purpose. An average lets one extreme week drag the whole figure toward it — a chaotic launch day, a single panic sale, a short-lived promotion. A median simply asks what price most of the window actually looked like, and reports that. It is a duller number than an average and a more honest one.

Six months is long enough to hold a real volume of trading, not just opening-week noise, and short enough that the figure still describes the machine as a current generation rather than as an old one drifting toward its replacement's launch. It is the same window, applied the same way, to every machine we price.

A launch week and the months that follow it are both real. Averaged together, they describe neither.

A close-up of a brass balance's central pivot and fork-shaped equilibrium pointer, settled dead-centre and at rest — A settled price is not a guess at where a market is heading. It is where it has already come to rest.

When the window holds two different prices

Sometimes the six months are not one market but two. A machine can trade in a tight band for its first weeks and then reprice sharply lower — a cheaper rival generation shipping, a manufacturer clearing a glut — and stay at that lower level for the rest of the window. When that happens, a plain median across the whole six months lands somewhere between the two regimes: a number that describes neither the early price nor the one that followed it, and one that nobody was actually paying on any given day.

So the rule has a second clause. If a machine's price falls by more than 25% at some point inside its first six months, the settled price is the median of the period after that drop, not a blend of before and after.

It is worth seeing why with round numbers. Suppose a machine launches at $90 a terahash and trades in a tight band around $85 for its first six weeks. A rival then ships a comparably efficient generation for less, wholesalers reprice to compete, and by week nine the machine is trading around $60, where it stays for the rest of the six-month window. A median across the whole period lands near $75 — a price nobody was paying by month one and nobody was paying by month five either. The post-drop median takes the median of the $60-ish weeks alone, because that later, cheaper regime is the market a buyer entering the machine today is actually stepping into. The rule is not smoothing away an inconvenient data point. It is choosing which of two real markets the number should describe.

The 25% threshold is what keeps the rule from firing on ordinary noise. Prices wobble a little most weeks without meaning anything — a slow month, a shipment delay, a wholesaler clearing a small batch. A repricing that large, inside a window that short, is different in kind: it is the market openly saying the earlier price no longer applies. Below that line, the plain six-month median still does its job. Above it, using the whole window would let a fact that stopped being true — the pre-drop price — go on shaping a number that is supposed to describe today's buyer.

A single median across two different markets describes neither of them.

The hashrate never moves

While a machine's price is doing all of this — spiking, sliding, sometimes stepping down hard — the thing you are actually buying does not move at all. Hashrate and efficiency, joules per terahash, are fixed the day the chip design and the board layout are finalised. A machine selling at $85 a terahash in its first month and one selling at $60 a terahash in its sixth are, terahash for terahash, the identical machine.

That is why a lower settled price is not a lesser machine. It is the same machine, at a lower entry cost, and the return that entry cost buys is a separate question from the one efficiency answers. Efficiency decides how many halvings a machine survives — a fact about the silicon that a settled price does not touch either way.

What the settled price does change is the size of the return on the money actually spent. The same hashrate bought for less produces the same coins against a smaller landed cost, which is arithmetic, not opinion — cashflow and the multiple it has paid back are both measured against what left your account, and a lower settled entry lowers that number without lowering anything the machine produces.

Whatever the price does in those six months, the chips inside do not change.

What 'the best entry price it has ever had' means

This is the sentence our reviews reach for once a generation has been on the market long enough to have a real price history, and it means something specific: today's price is at or below the settled price computed this way — the median a real cohort of buyers actually paid across a comparable window, not an arbitrary earlier number chosen because it makes the best story.

It is also a narrower claim than it can sound like, and worth being precise about. It is not a statement that the machine has become worth less. A machine's output — the hashes it produces per watt — never falls; only what it costs to acquire one does, and a fall in that cost is an entry price, not a loss. What changed is how much a buyer has to spend to get exactly the same thing this site has quoted from day one. That is the whole argument behind why the roof gets replaced once its price has settled, not the day it is announced.

It is also a claim you can check. The settled price is a fixed, published rule applied to a public price record, not a judgement call made fresh for each review — run the same six months of prices through the same median and you get the same number we did.

Why the launch price would be the wrong ruler

It is worth asking what would happen if we used the launch number as the benchmark instead. The answer is: every machine would look like a bargain almost immediately, whatever actually happened to its price afterwards, because launch pricing is built to be the highest number a generation ever carries. A comparison against it would flatter every machine equally and say nothing about that particular machine's own history since — whether today's price is genuinely good relative to what real buyers have paid for it, or merely lower than a number that was never representative in the first place.

The same problem runs the other way if the benchmark is picked from the bottom instead of the top. It would be just as easy, and just as misleading, to quote whatever the single cheapest week a machine has ever traded at happened to be — a panic sale, a distressed batch, a week nobody but the most opportunistic buyer actually transacted in. That number flatters nothing; it simply sets an impossible bar that makes every ordinary price look worse than it is. A median drawn from six real months of trading cannot be dragged to either extreme by one week at either end, which is exactly why it is the only number that survives being checked against the record.

The settled price avoids both traps by removing the choice. It is one mechanical rule, applied the same way to every machine in the catalogue, with no discretion to pick whichever earlier number tells the flattering story. A number that cannot be chosen to fit the conclusion is the only kind worth quoting.

Measured against a number nobody paid, every machine looks discounted by its second month.

The short version

Every price on this site is a settled price: the median of a machine's first six months on the market, or the median of the period after a drop of more than 25% inside that window. It is the same rule, applied the same way, to a launch-week flagship and to a generation two years into its life.

Whatever that number is, the hashrate a machine produces never changes with it. When a review says a machine is at the best entry price it has ever had, it is reporting a comparison against what real buyers have actually paid for identical hardware — not against a headline nobody transacted at, and not a claim that the machine is worth less than it was. It is making exactly the coins it always made. It has simply become the cheapest real way anyone has yet been able to get them, and that is why we buy a generation once its price has settled, rather than the week it launches.

The launch number is the one price nobody actually paid. It is the one we do not quote.

Disclosure

Firsthand Bitcoin sells and hosts mining hardware, including this machine. No manufacturer, distributor or affiliate programme paid for or reviewed this page and we take no commission on the links above. Historical figures are computed from daily bitcoin price and network hashprice, each day valued at its own prices. Nothing here is investment advice.