Choose an ASIC miner by working out which resource limits you. If your power capacity is fixed, the machine with the lowest J/TH gives the most hashrate from it. If capital is the limit and your electricity is cheap, a less efficient machine at a lower price per TH can be the rational choice. Everything else in a comparison (cooling type, voltage, hashrate version, age) either rules a model out or adjusts that trade-off.
Key points
- Rule models out first on electrical fit and cooling. A machine that does not match your supply is not a candidate at any price.
- Efficiency (J/TH) decides electricity cost per unit of hashrate and how long a machine stays above break-even. Price per TH decides how much capital you commit.
- At equal hashrate, an efficiency premium is recovered only through electricity savings, so it pays back faster the more you pay per kWh.
- Compare all-in cost per TH, including the infrastructure each kW requires, not only the machine price.
- Hashrate versions of one model can share the same efficiency. Compare them on price per TH.
For the purchase process itself (seller checks, payment, delivery), see How to Buy a Bitcoin Miner.
Efficiency against purchase price
Efficiency is J/TH = power (W) ÷ hashrate (TH/s). Price per TH is unit price ÷ hashrate.
Take two illustrative machines with the same hashrate. Prices are invented for the arithmetic and are not market prices.
| Machine A | Machine B | |
|---|---|---|
| Hashrate | 200 TH/s | 200 TH/s |
| Power | 3,500 W | 2,700 W |
| Efficiency | 17.5 J/TH | 13.5 J/TH |
| Illustrative price | $2,400 | $4,000 |
| Price per TH | $12.00 | $20.00 |
| Energy per day (kW × 24) | 84.0 kWh | 64.8 kWh |
Both earn the same revenue, because revenue follows hashrate. B costs $1,600 more and saves 84.0 − 64.8 = 19.2 kWh per day. The time to recover the premium is premium ÷ (kWh saved per day × price per kWh).
| Electricity price | Saving per day | Days to recover $1,600 |
|---|---|---|
| $0.03 | $0.576 | 2,778 |
| $0.05 | $0.960 | 1,667 |
| $0.08 | $1.536 | 1,042 |
| $0.12 | $2.304 | 694 |
At $0.03 per kWh the premium takes more than seven years to recover, during which several newer generations are likely to arrive. At $0.12 it takes under two years. On savings alone, cheap power weakens the case for paying up for efficiency.
Break-even resilience
Savings are not the whole argument. Break-even electricity price = daily revenue ÷ daily kWh. Assume, for illustration only, revenue of $0.05 per TH per day, so $10.00 per day for either machine.
| Illustrative revenue per TH per day | Machine A break-even | Machine B break-even |
|---|---|---|
| $0.05 | $0.119 per kWh | $0.154 per kWh |
| $0.03 | $0.071 per kWh | $0.093 per kWh |
If revenue per TH falls from $0.05 to $0.03 and you pay $0.08 per kWh, Machine A costs more to run than it earns while Machine B still covers its electricity. Revenue per TH tends to fall as network difficulty rises, so the efficient machine has a longer working life. Use the calculator with current network values. These figures are not a forecast.
When power is the limit
With 100 kW available for miners, you can run 28 of Machine A (98.0 kW, 5,600 TH/s, $67,200) or 37 of Machine B (99.9 kW, 7,400 TH/s, $148,000). B delivers 32% more hashrate from the same supply for 2.2 times the capital. When you cannot add power, hashrate can only grow through lower J/TH.
Hashrate versions of one model
Manufacturers sell one model in several hashrate versions because chips vary in production. Bitmain's published specification for the Antminer S21 Pro lists 220, 234 and 245 TH/s versions at 3,300, 3,510 and 3,675 W. All three are 15 J/TH. The larger version uses proportionally more power, so the choice is about price per TH and how many units fit your circuits.
The version label does not always fix the efficiency. MicroBT's shop lists the WhatsMiner M50S++ at 160 TH/s in both a 21 J/TH version (3,360 W) and a 22 J/TH version (3,520 W). Confirm the hashrate and the power or J/TH of the exact unit on the quotation, and use compare to line versions up.
Cooling type
| Air | Hydro | Immersion | |
|---|---|---|---|
| Heat leaves by | Air through the chassis | Coolant through cold plates | Dielectric fluid around the boards |
| You must provide | Airflow in and out, filtration | Coolant loop, pump, dry cooler | Tank, fluid, pump, heat exchanger |
| Noise at the machine | High, from fans | Low, noise moves to the dry cooler | Low |
| Power per unit | Lower | Higher | Varies |
Hydro and immersion models are built for that cooling method and cannot be swapped into an air-cooled layout. Choose the cooling method for the site, then choose among models of that type. ASIC Cooling Methods Compared goes deeper.
Electrical fit
Check the published input range against the voltage your site actually delivers.
| Published input range | Example from manufacturer data |
|---|---|
| 220 to 277 V single-phase | Bitmain Antminer S21, S21 Pro, S21 XP (air) |
| 220 to 240 V | MicroBT WhatsMiner M60S (air) |
| 380 to 415 V three-phase | Bitmain Antminer S21 XP Hyd. |
| 380 to 480 V | MicroBT WhatsMiner M63S (hydro) |
| 380 to 480 V three-phase | Bitdeer SEALMINER A4 Pro Hydro |
Why it matters: a 480 V three-phase system provides 480 ÷ √3 = 277 V from phase to neutral. That is the top of the 220 to 277 V range and outside a 220 to 240 V range. A 415 V system provides 415 ÷ √3 = 240 V, inside both. Likewise a 380 to 415 V hydro unit is not rated for a 480 V supply, while a 380 to 480 V unit is.
Supply design, breaker sizing and distribution must be done by a qualified electrician to local code. See Bitcoin Miner Power Requirements.
Generation and age
Each chip generation lowers J/TH. Manufacturer-published figures for five air-cooled Bitmain models show the scale:
| Model (version) | Hashrate | Power | Efficiency | kWh per day per 100 TH/s |
|---|---|---|---|---|
| Antminer S19j Pro (100T) | 100 TH/s | 2,950 W | 29.5 J/TH | 70.8 |
| Antminer S19 XP (134T) | 134 TH/s | 2,881 W | 21.5 J/TH | 51.6 |
| Antminer S21 | 200 TH/s | 3,500 W | 17.5 J/TH | 42.0 |
| Antminer S21 XP | 270 TH/s | 3,645 W | 13.5 J/TH | 32.4 |
| Antminer S23 | 305 TH/s | 3,355 W | 11 J/TH | 26.4 |
The last column is J/TH × 100 × 24 ÷ 1,000. The oldest machine in the table uses 2.7 times the electricity of the newest for the same hashrate (70.8 ÷ 26.4). Older generations cost less per TH for exactly that reason. Browse by efficiency band to see what sits in each range.
New or used
A used machine lowers capital cost and usually raises J/TH, and its remaining life and warranty position are less certain. It suits low electricity prices and operators who can repair hardware. New vs Used ASIC Miners sets out the checks.
Total cost per TH including infrastructure
Every kW of miners needs transformers, switchgear, distribution, cooling and space. That cost scales with power, so an inefficient machine carries more of it per TH.
All-in cost per TH = machine price per TH + (infrastructure cost per kW × J/TH ÷ 1,000)
With an illustrative infrastructure cost of $400 per kW:
| Machine A | Machine B | |
|---|---|---|
| Machine price per TH | $12.00 | $20.00 |
| Infrastructure per TH | 400 × 17.5 ÷ 1,000 = $7.00 | 400 × 13.5 ÷ 1,000 = $5.40 |
| All-in per TH | $19.00 | $25.40 |
The gap narrows from $8.00 to $6.40 per TH. On a site where infrastructure is expensive, it narrows further. Add freight, import duty and tax, which your quotation and customs broker will give you.
A selection sequence
- List models whose input voltage, phase and cooling type match your site.
- For each, compute the break-even electricity price in the calculator and drop any with little margin over your tariff.
- Compute all-in cost per TH for the rest.
- If power is your limit, rank by J/TH. If capital is your limit, rank by all-in cost per TH.
- Check availability status, hashrate version and warranty terms on the quotation before committing.