Lucas runs eleven Antminer S21 Pros in a converted workshop outside Leipzig. On paper, every unit is rated at 15 joules per terahash. When he clamped a meter on the feed line last winter, the fleet was drawing closer to 16.4. Same machines, same firmware, a number that did not match the box. That gap, between the J/TH a miner is rated at and the J/TH it actually delivers, is the most misread figure in mining hardware. It also decides more about your profit than the headline hashrate ever will. This guide covers what J/TH measures, how to calculate it, what counts as a good number in 2026, and why the rating on the spec sheet is a ceiling rather than a promise. The hardware side of that story sits in how we build and test every ASIC miner before it ships.
What does J/TH actually measure?
J/TH stands for joules per terahash. It is the amount of electrical energy a miner spends to compute one trillion hashes, which is one terahash of work. The lower the number, the less power the machine burns for the same output.
The logic behind it is simple. Your mining revenue tracks hashrate, because more terahashes mean more chances at the block reward. Your cost tracks power draw, because the meter charges you for every watt. J/TH is the ratio between the two, and it tells you how much of each mined satoshi you actually keep. A miner at 13 J/TH and an older one at 30 J/TH can point at the same pool and mine the same coin.The efficient one keeps roughly twice as much of the reward, because it pays half the electricity for the same hashrate. That single ratio is why a four year old machine that still works can be deeply unprofitable while a current unit prints margin at the same power rate.
The logic behind it is simple. Your mining revenue tracks hashrate, because more terahashes mean more chances at the block reward. Your cost tracks power draw, because the meter charges you for every watt. J/TH is the ratio between the two, and it tells you how much of each mined satoshi you actually keep. A miner at 13 J/TH and an older one at 30 J/TH can point at the same pool and mine the same coin.The efficient one keeps roughly twice as much of the reward, because it pays half the electricity for the same hashrate. That single ratio is why a four year old machine that still works can be deeply unprofitable while a current unit prints margin at the same power rate.
How do you calculate a miner's J/TH?
The formula is one line: divide the miner's power draw in watts by its hashrate in terahashes per second.
J/TH = power (watts) ÷ hashrate (TH/s)
Take the Antminer S21 Pro. It draws 3,510 watts and produces 234 TH/s. Divide 3,510 by 234 and you get 15.0 J/TH. Run the same math on the air-cooled Antminer S23 at 3,498 watts and 318 TH/s and you land at 11.0 J/TH. The S23 does noticeably more work for almost the same power bill, which is the entire point of a generational upgrade. One detail trips people up. Spec sheets sometimes quote the power at the wall and sometimes the power at the hashboard, before power-supply losses. Wall power is the honest figure, because that is what your meter bills. When you compare two miners, make sure both numbers come from the same measurement point, or the comparison is meaningless.
J/TH = power (watts) ÷ hashrate (TH/s)
Take the Antminer S21 Pro. It draws 3,510 watts and produces 234 TH/s. Divide 3,510 by 234 and you get 15.0 J/TH. Run the same math on the air-cooled Antminer S23 at 3,498 watts and 318 TH/s and you land at 11.0 J/TH. The S23 does noticeably more work for almost the same power bill, which is the entire point of a generational upgrade. One detail trips people up. Spec sheets sometimes quote the power at the wall and sometimes the power at the hashboard, before power-supply losses. Wall power is the honest figure, because that is what your meter bills. When you compare two miners, make sure both numbers come from the same measurement point, or the comparison is meaningless.
How do you read J/TH off a spec sheet without getting fooled?
Manufacturers and resellers do not always quote efficiency the same way, and the gaps are where buyers get caught. Three things are worth checking before you trust a number.
First, confirm whether the wattage is measured at the wall or at the hashboard. Hashboard power ignores power-supply losses and flatters the J/TH by several percent. The wall figure is what your meter charges, so ask for it.Second, watch the operating mode. Many miners ship with a normal mode, a low-power eco mode, and sometimes a turbo or overclock mode, each with its own hashrate and its own J/TH. A listing that pairs the turbo hashrate with the eco-mode efficiency is quoting two modes at once, which no single machine delivers.Third, treat the headline hashrate as a target, not a guarantee. Real units run 1 to 2 percent under rated even when healthy, and more when warm. The honest way to compare two machines is to put both on wall power, in the same mode, and recompute J/TH yourself from the raw watts and TH/s.
First, confirm whether the wattage is measured at the wall or at the hashboard. Hashboard power ignores power-supply losses and flatters the J/TH by several percent. The wall figure is what your meter charges, so ask for it.Second, watch the operating mode. Many miners ship with a normal mode, a low-power eco mode, and sometimes a turbo or overclock mode, each with its own hashrate and its own J/TH. A listing that pairs the turbo hashrate with the eco-mode efficiency is quoting two modes at once, which no single machine delivers.Third, treat the headline hashrate as a target, not a guarantee. Real units run 1 to 2 percent under rated even when healthy, and more when warm. The honest way to compare two machines is to put both on wall power, in the same mode, and recompute J/TH yourself from the raw watts and TH/s.
Why does J/TH matter more than hashrate?
Hashrate sells the spec sheet. Efficiency and uptime decide the profit. Two miners can post the same hashrate and earn the same gross revenue, yet one ends the month ahead and the other underwater, purely on the power they burned to get there
.
Put real numbers on it. At $0.08 per kWh, an S21 Pro at 3,510 watts costs about $6.74 a day to run. An S23 at 3,498 watts costs about $6.72, almost identical, but it produces 318 TH/s against the S21 Pro's 234. You are paying the same electricity for 36 percent more hashrate. That is efficiency expressed as money, and over a year it is the difference between a healthy margin and a stalled one. There is a timing angle too. When hashprice is compressed, as it has been through mid 2026 with the metric hovering around $33 per petahash a day, the margin between revenue and power cost is thin for everyone.
In that environment the efficient machine still clears a profit while the inefficient one tips into a loss, because the only lever either operator controls is the power side, and J/TH is that lever. A rising difficulty does the same thing. It squeezes every miner on the network, and the efficient ones are the last to go underwater.This is also why electricity rate and efficiency are joined at the hip. A great J/TH at a terrible power rate still loses money, and a mediocre J/TH at a very cheap rate can still print. The number on the box only becomes profit once you pair it with what you pay per kWh. To see how that plays out across current models, our 2026 ranking of the most profitable Bitcoin miners runs the margin math machine by machine.
.
Put real numbers on it. At $0.08 per kWh, an S21 Pro at 3,510 watts costs about $6.74 a day to run. An S23 at 3,498 watts costs about $6.72, almost identical, but it produces 318 TH/s against the S21 Pro's 234. You are paying the same electricity for 36 percent more hashrate. That is efficiency expressed as money, and over a year it is the difference between a healthy margin and a stalled one. There is a timing angle too. When hashprice is compressed, as it has been through mid 2026 with the metric hovering around $33 per petahash a day, the margin between revenue and power cost is thin for everyone.
In that environment the efficient machine still clears a profit while the inefficient one tips into a loss, because the only lever either operator controls is the power side, and J/TH is that lever. A rising difficulty does the same thing. It squeezes every miner on the network, and the efficient ones are the last to go underwater.This is also why electricity rate and efficiency are joined at the hip. A great J/TH at a terrible power rate still loses money, and a mediocre J/TH at a very cheap rate can still print. The number on the box only becomes profit once you pair it with what you pay per kWh. To see how that plays out across current models, our 2026 ranking of the most profitable Bitcoin miners runs the margin math machine by machine.
How do you turn J/TH into a monthly power bill?
Efficiency becomes a number you can budget once you run it through your power rate. The chain is short: watts to kilowatts, kilowatts to a daily cost, daily to monthly.
Take the S21 Pro again at 3,510 watts, which is 3.51 kW. Running flat out, it draws about 84.2 kWh a day, or roughly 2,527 kWh a month. At $0.08 per kWh that is about $202 a month per machine in electricity. Drop the rate to $0.07 and the bill falls to about $177. The difference, $25 a month per unit, is small for one miner and large across a fleet of fifty. Now compare hardware. The air-cooled S23 at 3,498 watts costs almost the same per month to run, about $201 at $0.08, but it produces 318 TH/s against the S21 Pro's 234. Same bill, 36 percent more hashrate.
That is the monthly version of the efficiency argument, and it is why the gap between a 15 J/TH machine and an 11 J/TH machine widens every month it runs. Model your own numbers at the rate you can actually secure before you commit to either the hardware or the hosting.
Take the S21 Pro again at 3,510 watts, which is 3.51 kW. Running flat out, it draws about 84.2 kWh a day, or roughly 2,527 kWh a month. At $0.08 per kWh that is about $202 a month per machine in electricity. Drop the rate to $0.07 and the bill falls to about $177. The difference, $25 a month per unit, is small for one miner and large across a fleet of fifty. Now compare hardware. The air-cooled S23 at 3,498 watts costs almost the same per month to run, about $201 at $0.08, but it produces 318 TH/s against the S21 Pro's 234. Same bill, 36 percent more hashrate.
That is the monthly version of the efficiency argument, and it is why the gap between a 15 J/TH machine and an 11 J/TH machine widens every month it runs. Model your own numbers at the rate you can actually secure before you commit to either the hardware or the hosting.
Ready to Start Mining?
Free worldwide DDP shipping. Professional hosting from $0.055/kWh.
What counts as a good J/TH in 2026?
The bar moves every hardware generation, so a good number is always relative to what is shipping. As of mid 2026, here is roughly where the lines sit.
- Under 13 J/TH is flagship territory. Hydro units like the Antminer S23 Hyd at around 9.5 and the air-cooled S23 near 11.0 lead the field. These clear margin even at standard hosting rates.
- 13 to 16 J/TH is top-tier. The Antminer S21 XP and Whatsminer M70S sit at 13.5, the S21 Pro at 15.0. Strong, current, and widely deployed.
- 16 to 22 J/TH is the middle. Still workable, but it needs power near $0.06 per kWh or below to stay comfortably positive.
- Above 22 J/TH is fading. The standard S19 generation and older Whatsminer M30s land here. At normal co-location rates the margin is thin to negative.
- Above 30 J/TH is effectively dead. S9 and S17 era hardware burns more in electricity than it earns unless power is close to free.
The practical takeaway is blunt. If you are shopping in 2026 and a listing quotes a number above the low twenties, you are looking at obsolete or near-obsolete hardware, no matter how cheap the unit is. The purchase price is a rounding error next to three years of electricity.
How does cooling change the efficiency you actually get?
Cooling is not a side topic in an efficiency discussion. It is one of the main levers, because temperature is what stands between a miner's rated J/TH and its delivered J/TH.
Air-cooled units are the default and the most flexible. They also throttle first. When intake air warms up, the chips heat, the firmware trims clocks, and effective J/TH rises. In a controlled room they hold their rating well. In a hot garage in July they do not.Hydro-cooled units run coolant directly across the hashboards, which holds chip temperatures far lower and steadier. That is why the hydro flagships, the Antminer S23 Hyd near 9.5 J/TH and the S21 XP Hyd at 12.0, sit at the top of every efficiency table.
The catch is infrastructure: hydro needs a coolant loop, a dry cooler, and usually three-phase power, so it is a facility play, not a spare-bedroom one. Immersion sits in a similar bracket, submerging the whole unit in dielectric fluid for the steadiest temperatures of all.The practical read for most buyers is this. If you can host in a facility that already runs hydro or immersion, the best efficiency on the market is open to you. If you are running air-cooled at home, your delivered J/TH will live or die on how well you manage intake temperature, which is the single biggest reason home numbers drift from the spec sheet.
Air-cooled units are the default and the most flexible. They also throttle first. When intake air warms up, the chips heat, the firmware trims clocks, and effective J/TH rises. In a controlled room they hold their rating well. In a hot garage in July they do not.Hydro-cooled units run coolant directly across the hashboards, which holds chip temperatures far lower and steadier. That is why the hydro flagships, the Antminer S23 Hyd near 9.5 J/TH and the S21 XP Hyd at 12.0, sit at the top of every efficiency table.
The catch is infrastructure: hydro needs a coolant loop, a dry cooler, and usually three-phase power, so it is a facility play, not a spare-bedroom one. Immersion sits in a similar bracket, submerging the whole unit in dielectric fluid for the steadiest temperatures of all.The practical read for most buyers is this. If you can host in a facility that already runs hydro or immersion, the best efficiency on the market is open to you. If you are running air-cooled at home, your delivered J/TH will live or die on how well you manage intake temperature, which is the single biggest reason home numbers drift from the spec sheet.
Spec J/TH vs delivered J/TH: why your miner runs hotter than the box says
Here is the part most efficiency guides skip, and it is the part that cost Lucas real money. The J/TH printed on the spec sheet is a best-case lab figure. It is measured on fresh silicon, at an ideal ambient temperature, at stock voltage and firmware. Your machine, in your room, rarely hits that number. Three forces push the delivered figure above the rated one.
Heat is the biggest one:
Air-cooled miners throttle when intake air gets warm. As the chips heat up, the firmware backs off clock speed to protect them, and hashrate drops by 5 to 15 percent. The wattage barely moves, so you are doing less work for the same power, and effective J/TH climbs. Lucas's workshop hit 32 degrees Celsius at the intake by mid afternoon, and that alone explained most of his drift from 15.0 to 16.4
Voltage and firmware are the second:
Input voltage, power-supply losses, and how the miner is tuned all shift the efficiency curve. An undervolted, well-tuned unit can beat its rating. A unit fed unstable voltage through a marginal power supply can fall short of it. Custom firmware moves this in both directions depending on how it is configured.
Aging is the third:
Silicon degrades. Over 18 to 24 months of continuous operation, chips lose a little efficiency, and weaker ones in a hashboard can drop out entirely, which raises the J/TH of whatever capacity remains. Even in good conditions, a healthy fleet tends to run 1 to 2 percent below its rated hashrate over time.Add those together and the lesson is clear. Your real J/TH is the one you measure at the wall, at your site, not the one on the datasheet. The rating tells you the ceiling. Where and how you run the machine decides how close you get to it. A genuine, properly tested unit holds closest to spec, which is why provenance and quality control are not separate from efficiency, they protect the efficiency you paid for.
Does the most efficient miner always win?
Not always, and this is where a lot of buyers overspend. The most efficient miner on the market carries a premium, and that premium only pays back if your power rate is high enough to reward it. There is a crossover.
Below roughly $0.04 per kWh, electricity is so cheap that hardware cost dominates the math. A less efficient, much cheaper unit can out-return a flagship, because the efficiency gap never adds up to enough money to justify the price difference. Above roughly $0.06 per kWh, the relationship flips. Power becomes the dominant cost, the efficiency gap compounds fast, and only top-tier J/TH clears a healthy margin.
When J/TH should not be your deciding number. If you have genuinely cheap power, under $0.04 behind the meter from solar, flared gas, or a private hydro deal, do not reflexively buy the most efficient machine. Run the payback math. A discounted prior-generation unit at 17 or 18 J/TH can beat a flagship on total return, because you recover the lower purchase price long before the efficiency premium catches up. The same caution applies to volatile altcoin miners, where a doubling in coin price over a few weeks dwarfs any J/TH difference between two machines. Efficiency is the right lens for most Bitcoin buyers at normal power rates. It is not a universal law.
Below roughly $0.04 per kWh, electricity is so cheap that hardware cost dominates the math. A less efficient, much cheaper unit can out-return a flagship, because the efficiency gap never adds up to enough money to justify the price difference. Above roughly $0.06 per kWh, the relationship flips. Power becomes the dominant cost, the efficiency gap compounds fast, and only top-tier J/TH clears a healthy margin.
When J/TH should not be your deciding number. If you have genuinely cheap power, under $0.04 behind the meter from solar, flared gas, or a private hydro deal, do not reflexively buy the most efficient machine. Run the payback math. A discounted prior-generation unit at 17 or 18 J/TH can beat a flagship on total return, because you recover the lower purchase price long before the efficiency premium catches up. The same caution applies to volatile altcoin miners, where a doubling in coin price over a few weeks dwarfs any J/TH difference between two machines. Efficiency is the right lens for most Bitcoin buyers at normal power rates. It is not a universal law.
How does where you run it change your real J/TH?
If delivered J/TH depends on heat, voltage, and maintenance, then the running environment is not a detail, it is part of the spec. This is the honest case for hosting, and it is a numbers argument, not a sales pitch.
A climate-controlled facility holds intake temperatures stable, so air-cooled units stop throttling and hold their rated hashrate through the afternoon. Clean, stable three-phase power keeps the efficiency curve where the manufacturer intended. Regular maintenance and dust control slow the aging curve. Put the same S21 Pro that drifted to 16.4 in Lucas's workshop into a controlled hall and the meter reads close to 15.1 again. The machine never changed. The conditions did. That is the logic behind professional hosting at $0.07 to $0.08 per kWh: you are not just renting cheap power, you are buying back the efficiency you already paid for at purchase.
For most operators running air-cooled current-generation hardware, the delivered J/TH inside a real facility is the closest you will get to the spec sheet outside a lab. To match a model to your situation, start with the current-generation Bitcoin miners and check each one's rated efficiency against the power rate you can actually secure. When Lucas moved six of those S21 Pros into a hosted hall, the meter settled at 15.1, not 16.4. He swapped no hardware and reflashed no firmware. The number on the box finally matched because the room finally matched. That is the whole lesson in J/TH, compressed into one fleet. It is not a fixed property of the miner. It is a property of the miner and the place you run it, and the operators who treat it that way are the ones still in profit when difficulty climbs.
A climate-controlled facility holds intake temperatures stable, so air-cooled units stop throttling and hold their rated hashrate through the afternoon. Clean, stable three-phase power keeps the efficiency curve where the manufacturer intended. Regular maintenance and dust control slow the aging curve. Put the same S21 Pro that drifted to 16.4 in Lucas's workshop into a controlled hall and the meter reads close to 15.1 again. The machine never changed. The conditions did. That is the logic behind professional hosting at $0.07 to $0.08 per kWh: you are not just renting cheap power, you are buying back the efficiency you already paid for at purchase.
For most operators running air-cooled current-generation hardware, the delivered J/TH inside a real facility is the closest you will get to the spec sheet outside a lab. To match a model to your situation, start with the current-generation Bitcoin miners and check each one's rated efficiency against the power rate you can actually secure. When Lucas moved six of those S21 Pros into a hosted hall, the meter settled at 15.1, not 16.4. He swapped no hardware and reflashed no firmware. The number on the box finally matched because the room finally matched. That is the whole lesson in J/TH, compressed into one fleet. It is not a fixed property of the miner. It is a property of the miner and the place you run it, and the operators who treat it that way are the ones still in profit when difficulty climbs.
Frequently asked questions
What is a good J/TH for an ASIC miner in 2026?
Anything under 13 J/TH is current flagship efficiency, and under 16 J/TH is top-tier. Machines between 16 and 22 J/TH still work but need power near $0.06 per kWh or lower. Above 22 J/TH, most hardware struggles to stay profitable at standard hosting rates.Is lower J/TH better?Yes. J/TH measures energy spent per unit of work, so a lower number means the miner produces more hashrate for every watt it draws. Lower J/TH directly means lower running cost for the same mining output.
What is the difference between J/TH and TH/s?
TH/s, terahashes per second, is how much work a miner does. J/TH, joules per terahash, is how much electricity it spends per unit of that work. Hashrate drives revenue, efficiency drives cost, and you need both to judge a machine.
How do I convert watts to J/TH?
Divide the miner's power draw in watts by its hashrate in terahashes per second. A 3,510 watt miner producing 234 TH/s runs at 3,510 divided by 234, which is 15.0 J/TH. Use wall power, not hashboard power, for an honest figure.
What is the most efficient ASIC miner in 2026?
The hydro-cooled Antminer S23 Hyd leads at roughly 9.5 J/TH, with the air-cooled S23 close behind near 11.0 J/TH. Among widely available air units, the Antminer S21 XP and Whatsminer M70S sit at 13.5 J/TH.
Does a miner's J/TH get worse over time?
Slightly, yes. Over 18 to 24 months of continuous running, chips degrade and weaker ones can drop out, which lifts effective J/TH. Stable temperatures and good maintenance slow this, but no machine holds its launch-day efficiency forever.
Why is my miner pulling more watts than its rated efficiency?
Usually heat. Warm intake air makes the chips throttle, so hashrate falls while wattage stays high, which pushes your real J/TH above the rating. Unstable voltage, a weak power supply, or aggressive firmware can do the same.
Does hydro cooling improve J/TH?
Yes, meaningfully. Liquid cooling holds chip temperatures lower and steadier than air, which lets manufacturers push higher clocks at better efficiency and removes the throttling that inflates real-world J/TH. It is why hydro units lead the efficiency rankings.
How does J/TH affect mining profitability?
Profit is revenue minus power cost, and J/TH sets the power cost side. Multiply it by your electricity rate to get the energy cost per terahash. At a fixed power rate, a lower J/TH machine always keeps more of the block reward.
Is a more efficient miner worth the higher price?
It depends on your power rate. Above about $0.06 per kWh, the efficiency premium usually pays back. Below about $0.04 per kWh, a cheaper, less efficient unit can return more overall, because the power saving never recovers the price gap. Run the payback math at your real rate.
What is a good J/TH for an ASIC miner in 2026?
Anything under 13 J/TH is current flagship efficiency, and under 16 J/TH is top-tier. Machines between 16 and 22 J/TH still work but need power near $0.06 per kWh or lower. Above 22 J/TH, most hardware struggles to stay profitable at standard hosting rates.Is lower J/TH better?Yes. J/TH measures energy spent per unit of work, so a lower number means the miner produces more hashrate for every watt it draws. Lower J/TH directly means lower running cost for the same mining output.
What is the difference between J/TH and TH/s?
TH/s, terahashes per second, is how much work a miner does. J/TH, joules per terahash, is how much electricity it spends per unit of that work. Hashrate drives revenue, efficiency drives cost, and you need both to judge a machine.
How do I convert watts to J/TH?
Divide the miner's power draw in watts by its hashrate in terahashes per second. A 3,510 watt miner producing 234 TH/s runs at 3,510 divided by 234, which is 15.0 J/TH. Use wall power, not hashboard power, for an honest figure.
What is the most efficient ASIC miner in 2026?
The hydro-cooled Antminer S23 Hyd leads at roughly 9.5 J/TH, with the air-cooled S23 close behind near 11.0 J/TH. Among widely available air units, the Antminer S21 XP and Whatsminer M70S sit at 13.5 J/TH.
Does a miner's J/TH get worse over time?
Slightly, yes. Over 18 to 24 months of continuous running, chips degrade and weaker ones can drop out, which lifts effective J/TH. Stable temperatures and good maintenance slow this, but no machine holds its launch-day efficiency forever.
Why is my miner pulling more watts than its rated efficiency?
Usually heat. Warm intake air makes the chips throttle, so hashrate falls while wattage stays high, which pushes your real J/TH above the rating. Unstable voltage, a weak power supply, or aggressive firmware can do the same.
Does hydro cooling improve J/TH?
Yes, meaningfully. Liquid cooling holds chip temperatures lower and steadier than air, which lets manufacturers push higher clocks at better efficiency and removes the throttling that inflates real-world J/TH. It is why hydro units lead the efficiency rankings.
How does J/TH affect mining profitability?
Profit is revenue minus power cost, and J/TH sets the power cost side. Multiply it by your electricity rate to get the energy cost per terahash. At a fixed power rate, a lower J/TH machine always keeps more of the block reward.
Is a more efficient miner worth the higher price?
It depends on your power rate. Above about $0.06 per kWh, the efficiency premium usually pays back. Below about $0.04 per kWh, a cheaper, less efficient unit can return more overall, because the power saving never recovers the price gap. Run the payback math at your real rate.
