Efficiency decides how many halvings a machine survives

Joules per terahash looks like a spec-sheet detail. It is the one number that decides whether a machine is still worth switching on after the next halving, and it explains why an old machine and a new one can both be good buys.

A polished silicon wafer patterned with hundreds of dies, lit at an angle

What joules per terahash measures

Every bitcoin miner does the same job. It computes hashes, as fast as it can, and the network pays for them in bitcoin. The only thing that separates one machine from another over its working life is how much electricity each of those hashes costs.

Joules per terahash — J/TH — is that cost. A joule is a watt for one second, and a terahash per second is a trillion hashes every second, so J/TH is simply watts drawn divided by terahashes produced. A machine drawing 3,500 watts to produce 200 TH/s runs at 17.5 J/TH. One drawing 5,000 watts for 500 TH/s runs at 10.

Turned into a daily figure it becomes concrete. Multiply J/TH by 0.024 and you have the kilowatt-hours one terahash burns in a day. At 17.5 J/TH that is 0.42 kWh per terahash per day; at 10 J/TH it is 0.24. Multiply by your electricity rate and you have what each terahash costs you to run, every day, for as long as it runs.

A mining machine sells exactly one thing, hashes, and it buys exactly one thing to make them: electricity.

The break-even line

The other side of the ledger is hashprice: what one unit of hashrate earns in a day. It is set by the network, not by the machine — every terahash on earth earns the same amount at any given moment, whatever box it came out of. Hashprice is usually quoted in dollars per petahash per day, and you can follow it on our hashprice chart.

Put the two sides together and every machine has a line. Above it, the machine earns more than its electricity costs. Below it, it does not. The line is just efficiency multiplied by the power rate, and it is worth seeing a few of them.

At 6¢ a kilowatt-hour a 30 J/TH machine needs hashprice above about $43 per petahash per day to clear its power. A 20 J/TH machine needs about $29. A 15 J/TH machine needs about $22, and a 10 J/TH machine about $14. At 3.9¢ the same four lines sit at roughly $28, $19, $14 and $9. Those are before cooling overhead and downtime, which move them a little, but the shape does not change: the line sits where efficiency and power price put it, and nothing else moves it.

Our break-even electricity chart turns this around and shows the highest power price a given efficiency can afford on any day in the record.

Two levers that multiply

Look at those numbers again and something useful appears. The cost of running a terahash is efficiency multiplied by the power price, so the two are interchangeable in the arithmetic.

A 30 J/TH machine on 3.9¢ power pays exactly as much electricity per terahash as a 19.5 J/TH machine on 6¢ power. Cheaper power does for an old machine what a newer generation of silicon does for a new one. That single fact explains most of how a mining fleet is run: the newest, most efficient machines go where uptime is highest, and older machines keep working for years by moving somewhere the power is cheaper.

What a halving does to the line

Bitcoin pays miners a fixed subsidy for every block, plus the fees in it. Every 210,000 blocks — roughly four years — the subsidy halves. It started at 50 bitcoin a block, fell to 25 in November 2012, 12.5 in July 2016, 6.25 in May 2020 and 3.125 in April 2024. The next halving, to 1.5625, is expected around April 2028. Our halving page counts down to it.

On the day it happens, every terahash on the network suddenly earns about half as much bitcoin as it did the day before (fees aside). Hashprice steps down overnight. The break-even line of every machine stays exactly where it was, because a halving changes nothing about a machine’s efficiency or its electricity price.

So a halving does not retire machines by age. It retires them by distance from the line. A machine comfortably above it before the halving is still above it afterwards. A machine that was only just clearing its power is suddenly below it. That is why efficiency is not a nicety — it is the measure of how far above the line a machine sits, and therefore how many halvings it can absorb.

A halving does not move the break-even line. It moves the ground underneath it.

The slow squeeze between halvings

Halvings are the steps. Between them there is also a slope.

The network adjusts its difficulty every 2,016 blocks, about every two weeks, so that blocks keep arriving roughly every ten minutes however much hashrate is pointed at them. As newer, more efficient machines are switched on, total hashrate climbs, difficulty follows it, and each existing terahash earns a slightly smaller share of the same rewards. You can watch it happen on the difficulty chart.

So left to itself, hashprice drifts down between halvings and drops at each one. Rallies in the price of bitcoin push it back up for long stretches, but across bitcoin’s history the long-run direction of hashprice has been down. What has kept mining worth doing is the hardware underneath it — and a machine’s efficiency is what decides how long it keeps paying its own way while that happens.

The silicon has kept pace with the halvings

Here is the part that makes the long-run picture hang together. Mainstream mining hardware from 2016 used around 100 joules per terahash. The mainstream machines of 2020 used between about 30 and 34. By late 2023 the new flagship was at 17.5, and the most efficient machine announced in 2026 is rated below 9.

That is roughly a halving of energy per hash with every generation — about as fast as the subsidy itself halves. It is not a law of nature, and each step gets harder as chip processes approach their limits. But it is why mining has not simply become unprofitable halving after halving: the network’s best machines keep producing twice the hashes per joule, and the operators running them keep making coins for a sensible cost.

It also tells you what a machine is up against. Its competition is not the machine it replaced. It is whatever the network plugs in next.

An hourglass on a dark surface, sand running through the neck — The halving runs on a published schedule. Efficiency decides how many turns of it a machine lasts.

Below the line is not the end

A machine that falls below the line at one site has not stopped being useful. It has stopped being useful at that price of electricity, and the same box keeps working for years at a cheaper one. That is the whole logic of moving older machines to cheaper power, and it is covered in Machines are the roof, not the building.

And a machine that nothing will pay for is switched off, not thrown away. Its coins are already mined and kept. When a rally lifts hashprice back above its line, it switches back on. Old generations can spend their last years this way, working in the good stretches and resting in the bad ones, without ever running at a loss.

Efficiency against entry price

None of this means the most efficient machine is always the best buy. Two rules pull in opposite directions, and both hold.

Efficiency decides how many halvings a machine survives. It is the number that decides whether a machine is still worth switching on in 2029.

The entry price decides the return. A generation that has been on the market for a year or two often costs far less than it did at launch, and the hashrate it produces has not changed at all. Bought at a settled price, a proven machine can return more in its first year than the newest one, which carries a launch premium for the extra runway. Our miner price chart shows how quickly each generation settles.

The right machine is the one whose runway matches your plan. If you are buying to mine through the 2028 halving and beyond, buy efficiency. If you are buying for the next two years at cheap power, the settled price of a slightly older generation may do more for you.

The one question to ask

You can answer it with a pencil. Take today’s hashprice and roughly halve it for the halving. Work out the machine’s line: its J/TH, times 0.024, times your power price, times a thousand for the per-petahash figure. If the halved hashprice still sits comfortably above the line, the machine survives the next halving even if the price of bitcoin does nothing at all. If it does not, the machine needs either a rising price or cheaper power to get through it.

That is the entire reason efficiency is worth paying for. It buys distance from the line, and distance from the line is what a halving takes away.

Will it still clear its power after the next halving, at the rate you will actually pay?

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.