ASIC miner ROI: how to work out real payback before you buy

*Draft of article 113. Date: 2026-08-27. Status: awaiting approval, not published.*

*Every figure here comes from our own material: article 97 (2026-08-14 snapshot), article 81, article 96, article 106, article 56. Nothing else has been added.*

TL;DR

ASIC payback is one fraction: the price of the machine divided by what it clears per day. The fraction is not the problem. The denominator is. It moves every two weeks with each difficulty retarget, and it moves down. So the "pays back in N months" number any online calculator hands you is really the payback period assuming the network stopped growing. It never has.

Second point: net means after electricity, not before. On the 2026-08-14 numbers in article 97, the gap between $0.05 and $0.12 per kWh at 100 TH/s was $2.02 a day, while the entire fee spread across twelve pools, 1% to 4%, came to 9 cents. Your power rate decides payback. The pool fee nudges it.

Below is a step by step method with the hardware price left as a variable, X. We do not invent hardware prices: use the one you are actually being quoted, delivered and cleared through customs.

[IMAGE: photoreal, bright sunny day, wooden table on a country porch, open laptop showing a spreadsheet of numbers, a paper notebook with handwritten figures and a mug of tea beside it, green hills and open sky beyond the railing. Light tones, soft natural daylight]

The formula everything else hangs off

Payback in days = X / (D − E), where

• X is the full cost of entry: the machine, shipping, duties, a PSU if it is not included, and the wiring and socket to carry the load;
• D is daily revenue after the pool fee;
• E is daily electricity cost.

The rest of this article is corrections to D and E. There are four: difficulty growth, downtime and rejected shares, the gap between datasheet and real hardware behaviour, and withdrawal fees plus exchange spread. All four push the result the same way, which is the wrong way.

How to calculate ASIC payback, step by step

1. Fix X. Not the sticker price on the manufacturer's site, the amount that will actually leave your account: machine, shipping, duties, electrical work. If the unit goes to a hosting site, the upfront deposit belongs here too.
2. Take rated TH/s and watts. Models with their hashrate and power draw are listed in our ASIC miner ranking. Efficiency in J/TH is watts divided by TH/s.
3. Work out D. BTC revenue for your hashrate at current difficulty, minus the pool fee. You do not need to do this by hand: the ASIC ROI calculator does it and shows FPPS against PPLNS on the same inputs.
4. Work out E. Watts × 24 / 1000 = kWh per day. Multiply by your real tariff, including everything that lands on the bill, not the discounted daytime rate.
5. Apply a downtime and reject correction. Nobody runs at 100% uptime. Take the uptime you actually expect and multiply D by it. Article 106 covers how to measure your own reject share: compare accepted against rejected for the last day in the pool dashboard, and treat a gap above 5% between live and rated hashrate as a hardware or link fault rather than a pool problem (article 96).
6. Build in falling revenue from difficulty growth. This is the step almost everyone skips. Difficulty retargets every 2016 blocks, roughly every two weeks. Your hashrate stays put, so your share of the network shrinks and BTC revenue with it. Do not compute D × 365. Sum two week chunks where D is slightly smaller each time.
7. Add withdrawal fees and exchange spread if you are measuring payback in fiat. The payout threshold does not take money, it holds it: at 100 TH/s a 0.005 BTC threshold takes about 107 days to fill, a 0.001 BTC threshold about 21 (article 106).
8. Redo the sum after the next retarget. A forecast without recalculation has a shelf life of two weeks. After that it is not a forecast.

Worked example: Antminer S21 XP Hyd

This is the machine behind the article 97 calculation: 473 TH/s, 5676 W, 12 J/TH per the official Bitmain User Guide V4.0.2. It is a hydro unit and among the most efficient on the market. Air cooled ASICs do worse, so most readers will land below these numbers rather than above them.

All inputs from the same 2026-08-14 snapshot: network hashrate 933.99 EH/s, difficulty 127,479,855,693,691, block reward 3.125 BTC, BTC at 64,558 USD.

Daily revenue after fees at 100 TH/s was $3.10 at the cheapest pool on the list and $3.01 at the 4% pools. Scaled to 473 TH/s, a factor of 4.73:

• cheapest pool on the list: $14.66 a day
• 4% pool: $14.24 a day

Electricity: 5676 W is 136.22 kWh a day.

That last column is the answer in terms of X. Divide your entry cost by 1,000 and multiply by the number in the column. A machine costing X dollars pays back in roughly X/7.43 days at $0.05, X/3.34 days at $0.08, and never at $0.12, where it burns $2.11 a day more than it earns and every day of runtime deepens the hole.

This is the same conclusion as the shutdown threshold breakdown: there is no single breakeven price for mining. Each combination of hardware and tariff has its own. The WuBlockchain figure of $46,787 applied to the most efficient class of machines; everything else sits above it.

Why does an online calculator show a shorter payback than reality?

Because it holds difficulty flat. The calculator takes today's daily revenue and divides the hardware price by it, as if the network froze for the whole payback period. Difficulty retargets around 26 times a year, usually upward. Real payback runs longer than the projection, and the faster the network grows, the wider the gap.

How much does difficulty growth stretch payback?

Take a scenario: BTC revenue drops 2% per retarget. Six months in, that is 13 retargets, and the S21 XP Hyd above falls from $14.24 to $10.95 a day. Electricity does not move. At $0.05 the net drops from $7.43 to $4.14, close to half. The denominator shrank, so the payback period grew.

To be clear, that 2% per retarget is an illustrative scenario for showing the mechanics, not a forecast. We publish no recommended difficulty growth figure of our own, because we have no verified long term series to base one on. Read the current pace off the live network difficulty chart and use that number instead. As for how fast the picture moves, Hashrate Index data quoted in article 81 put the seven day moving average network hashrate at 932 EH/s on 2026-08-11, up 6.2% in a week.

Which matters more for payback, the pool fee or the electricity rate?

The rate, by a wide margin. At 100 TH/s the difference between $0.05 and $0.12 per kWh was $2.02 a day, while the difference between a 1% and a 4% pool was 9 cents. More than twenty times apart (article 97). Shopping pools for half a point while your tariff assumption is wrong accomplishes nothing.

That does not make the pool irrelevant. On a thin margin, when a machine hovers near zero, the payout scheme and the threshold decide whether you see your money in three weeks or three months. Current fees, schemes and thresholds across twelve pools are laid out in the pool comparison and in article 97.

Should payback be measured in BTC or in dollars?

Track both, but keep them apart. BTC shows how the hardware is doing: how many coins it produces and how that number erodes as difficulty climbs. Dollars show the hardware and the market together. If BTC revenue holds steady while the dollar figure sags, the cause is price, not your farm (article 96).

Does solo mining change the payback picture?

It changes the distribution, not the total. Expected value is about the same, but the outcome becomes a lottery: nothing for years, then a whole block, or nothing at all. For a small farm's payback maths that is unusable, because payback needs a predictable daily flow. Full breakdown in article 56 on solo versus pool payback.

[IMAGE: photoreal, bright hangar with the doors wide open, rows of ASIC miners on racks, sunlit field and trees visible through the doorway, daylight inside, clean and spacious, a technician in a light shirt checking a tablet. No darkness, no blinking LEDs in gloom]

Pre-purchase checklist

1. Count the entry price in full, shipping, duties and electrical work included.
2. Use your actual tariff from the bill, not the advertised one.
3. Re-measure consumption at the wall once the unit is installed. The datasheet figure was captured under the manufacturer's conditions.
4. Assume realistic uptime, not 100%.
5. Compute in two week chunks with revenue stepping down each time, using the pace from the difficulty chart.
6. Compare the result against secondhand prices for the same model. Sometimes the fastest payback is the purchase you skip.
7. Run the final numbers through the ASIC ROI calculator and check the model in the miner ranking.

FAQ

Which ASIC ROI calculator is the most accurate?

They all run the same formula, so accuracy comes from your inputs. What differs between services is whether they pull current difficulty, let you pick a payout scheme, and account for the pool fee. A projection built on somebody else's electricity rate will be wrong in any calculator. Ours compares FPPS and PPLNS on identical inputs.

Should resale value go into the payback calculation?

Yes, if you plan to sell. The numerator then becomes X minus the price you expect to get within your horizon, not the full X. We will not forecast the secondhand ASIC market, but ignoring it is also wrong: for a machine you will definitely sell in two years, the full X overstates the payback period.

What if the calculation comes out negative?

Do not buy at that tariff. A negative denominator means the machine grows its loss every day it runs, and no pool switch fixes that: the fee spread is measured in cents and the shortfall in dollars. The options are cheaper energy, a more efficient model, or staying out.

How often should payback be recalculated after purchase?

After every difficulty retarget, so roughly every two weeks. While you are there, compare the worker's live hashrate in the pool dashboard against the rated figure. A gap above 5% points to hardware or connectivity and feeds straight into a longer payback (article 96).

---

*Author: POOL BTC Team. Figures from articles 81, 96, 97 and 106, 2026-08-14 snapshot. Estimates only, not financial advice.*