Used ASIC miners cost far less per TH and use far more electricity per TH. That makes them a reasonable purchase where power is cheap and you can absorb failures, and a poor one where power is expensive or you need the machine to stay profitable for years. New machines cost more up front, carry the manufacturer's warranty and keep earning above their electricity cost for longer as network difficulty rises.
Key points
- Compare on break-even electricity price, not on purchase price. A cheap machine that cannot cover its power bill has no payback.
- A used machine's margin is thin, so it is the first to become unprofitable when revenue per TH falls.
- Judge a used lot on per-unit test results by serial number, hashboard count, fan and PSU condition, repair history and firmware, not on a lot average or photographs.
- Assume the manufacturer's warranty has expired on a used unit unless the documents show otherwise. A seller warranty is a separate promise. Get its terms in writing.
- Used suits cheap power, short planning horizons and operators who can repair hardware.
The trade-off in numbers
Two illustrative machines. Prices and revenue are invented for the arithmetic and are not market figures.
| Used, older generation | New, current generation | |
|---|---|---|
| Hashrate | 100 TH/s | 200 TH/s |
| Power | 3,000 W | 3,500 W |
| Efficiency (W ÷ TH/s) | 30.0 J/TH | 17.5 J/TH |
| Illustrative price | $500 | $3,000 |
| Price per TH | $5.00 | $15.00 |
| Energy per day (kW × 24) | 72 kWh | 84 kWh |
| Energy per day per TH | 0.72 kWh | 0.42 kWh |
The used unit costs one third as much per TH and uses 71% more electricity per TH (0.72 ÷ 0.42 = 1.71).
Now assume illustrative revenue of $0.05 per TH per day, which gives $5.00 per day for the used unit and $10.00 for the new one. Daily margin = revenue − (kWh per day × price per kWh).
| Electricity price | Used, daily margin | New, daily margin | Used, margin per $1,000 invested | New, margin per $1,000 invested |
|---|---|---|---|---|
| $0.03 | $5.00 − $2.16 = $2.84 | $10.00 − $2.52 = $7.48 | $5.68 | $2.49 |
| $0.05 | $5.00 − $3.60 = $1.40 | $10.00 − $4.20 = $5.80 | $2.80 | $1.93 |
| $0.07 | $5.00 − $5.04 = −$0.04 | $10.00 − $5.88 = $4.12 | negative | $1.37 |
At $0.03 per kWh the used machine returns more than twice as much per dollar invested. At $0.07 it loses money while the new machine still clears $4.12 per day. The break-even electricity price (daily revenue ÷ daily kWh) shows why: $5.00 ÷ 72 = $0.069 per kWh for the used unit against $10.00 ÷ 84 = $0.119 for the new one.
What happens when revenue falls
These margins are a snapshot. Revenue per TH moves with network difficulty and the Bitcoin price. Suppose it falls 30% to an illustrative $0.035 per TH per day.
| Electricity price | Used, daily margin | New, daily margin |
|---|---|---|
| $0.03 | $3.50 − $2.16 = $1.34 | $7.00 − $2.52 = $4.48 |
| $0.05 | $3.50 − $3.60 = −$0.10 | $7.00 − $4.20 = $2.80 |
The used machine at $0.05 per kWh went from a $1.40 margin to a loss. The new machine kept about half its margin. This is the real cost of buying an older generation: the period in which it earns anything can be short, and it ends without warning. Run both candidates in the calculator with current network values, then test a lower revenue figure before deciding. None of this is a forecast.
If your power is limited
With 100 kW for miners you could run 33 of the used units (99 kW, 3,300 TH/s, $16,500) or 28 of the new ones (98 kW, 5,600 TH/s, $84,000). The new fleet produces 70% more hashrate from the same supply. Where power capacity is the scarce resource, used machines waste it.
Remaining life
A miner has two lives. Its economic life ends when revenue no longer covers electricity at your tariff. Its physical life depends on heat, dust, humidity and how it was run.
Fans and power supplies are wear parts and are replaceable. Hashboards are where a used machine's risk sits. They fail from thermal cycling, corrosion and dust-related overheating, and board-level repair requires equipment and skill. A used unit that spent years in a hot, dusty or humid facility is closer to those failures than its tested hashrate reveals. ASIC Miner Lifespan covers the mechanisms.
Warranty position
Manufacturer warranties are time-limited and counted from shipment. Bitmain's published terms for the Antminer S21 state 365 days from the shipping date. Bitdeer states a 365-day warranty starting from the shipment date for its SEALMINER models, and MicroBT states a one-year warranty period for the WhatsMiner models in its shop.
A used machine from an earlier generation is therefore likely to be past that period. Whether any remaining manufacturer cover passes to a second owner depends on the manufacturer's terms, so read them and ask for the original shipping date. Anything the seller offers on top is the seller's own warranty. Ask what it covers, how long it lasts, who pays freight for returns and what the dead-on-arrival window is.
What to check on a used lot
Condition labels such as certified refurbished, grade A, grade B and for parts or repair have no industry-wide definition. Ask the seller to define each grade in writing.
| Check | What to ask for | Why |
|---|---|---|
| Tested hashrate | Result per unit by serial number, test duration and ambient temperature | A lot average hides weak units. A short test hides thermal faults |
| Pool-side hashrate | A screenshot from a mining pool over a day | The pool measures real shares. A firmware dashboard can mislead |
| Hashboard count | Every board detected with the expected chip count | A machine running two boards out of three delivers about two thirds of its hashrate |
| Fans | All fans reporting speed, no bearing noise, matching type | Fans are cheap, but a failed fan may have overheated the boards |
| Power supply | Included or not, model, original or replaced | A missing or mismatched PSU is an extra cost and a fault source |
| Repair history | Which boards were repaired or swapped, and by whom | Reworked boards and mixed board versions are a frequent source of repeat faults |
| Firmware | Stock or third-party, and whether it can be returned to stock | Third-party firmware may charge a fee, alter tuning or block updates |
| Access | Web interface password reset and no remote lock | A locked unit may need a replacement control board |
| Physical state | Close photos of connectors, heatsinks and boards | White or green deposits indicate corrosion |
| Operating history | Climate, hosting type, any immersion use | Immersion-run units may lack fans and carry fluid residue |
For a lot of any size, pay for inspection of a sample, or all units, before releasing the balance. On arrival, test every unit yourself within the reporting window. Keep spare cooling fans, power supplies and hashboards in mind when you price the lot, because some units will need them.
When each makes sense
| Situation | Leaning | Reason |
|---|---|---|
| Electricity well below the used unit's break-even price | Used | Low capital, margin survives |
| Electricity near the used unit's break-even price | New | Used margin disappears with a small fall in revenue |
| Fixed power capacity, capital available | New | More hashrate per kW |
| Short site lease or temporary power | Used | Less capital at risk over a short horizon |
| No repair capability, small fleet | New | One failure is a large share of your hashrate |
| In-house repair bench and spares | Used | Failures are a cost you can manage |
| Financing or investor reporting | New | Warranty and predictable performance |