← All guides

Miner payback with difficulty drift: the break-even math that does not assume today lasts forever

Costs & math · 10 min read

General information only, not financial advice. Mining returns are not guaranteed and vary with difficulty, price, and power cost. Every figure below is either a dated published input named in the text or a labelled scenario. Replace the scenario inputs with your own quotes and rates before spending anything.

The kill criterion, before any maths

955 days. That is the static payback this article computes for an efficient current machine on cheap power, with honest costs included. The drift version of the same calculation never arrives. Both numbers come from the same inputs, and the difference between them is the whole decision.

So the kill criterion goes first, while walking away is still free. Do not buy the machine if its payback, calculated with difficulty drift rather than at today's rate, lands past the next halving, expected around 2028, when the block reward falls from 3.125 BTC to 1.5625 BTC and revenue per unit of hashrate drops again. Do not buy it if the monthly net turns negative before the purchase price is recovered, because from that month the machine costs money to keep switched on. And do not buy it if the plan only works at a Bitcoin price higher than todays. A plan that needs the price to rescue it is a price bet wearing a mining costume.

The inputs, dated and sourced

Revenue starts from hashprice, what one unit of hashrate earns per day. A November profitability report checked against Hashrate Index and network data put hashprice at $33.18 per PH per day on 2 November 2026, equal to $0.0332 per TH per day. Difficulty in the same report stood at 148.36 T, after rises of 5.92 percent on 16 October and 5.54 percent on 29 October. Hashprice had averaged about $38 through October and $39.33 in September. The snapshot input, in other words, was already falling when it was measured.

The worked machine is the Antminer S21 XP, rated by the manufacturer at 270 TH/s and 3,645 W, which is 13.5 J/TH, on 220 to 277 V input. The contrast machine is the S19j Pro at 104 TH/s and 3,068 W, or 29.5 J/TH. Manufacturer figures are typical values, and efficiency alone never establishes profit.

The remaining inputs are labelled scenario choices, stated so any reader can swap them. Hardware price $3,800 as a worked purchase input, to be replaced with a real delivered quote including shipping and tax. Power at $0.05 per kWh. Pool fee 2 percent. Uptime 97 percent, because fans fail, breakers trip, and internet connections drop. Repair reserve $0.30 per day, because a machine expected to run for years will need parts inside that window.

The static payback, and why it flatters

Gross revenue for the S21 XP at the dated hashprice is 270 times $0.0332, which is $8.96 per day. Power draw is 3.645 kW around the clock, 87.48 kWh per day, costing $4.37 at the scenario rate. Apply the pool fee and the uptime factor to revenue, the uptime factor to power, subtract the repair reserve, and the net is $3.98 per day. Divide the $3,800 purchase input by $3.98 and the static payback is 955 days, roughly 31 months.

Raise the power rate to $0.07 per kWh and the same stack nets $2.28 per day, stretching the static payback to 1,666 days. The published break even check agrees with the direction: the report rule divides hashprice per TH by J/TH times 0.024, which puts power only break even for a 13.5 J/TH machine at about $0.10 per kWh. Power only. The full stack breaks even lower, because the pool, the downtime, and the repairs also eat.

The older machine fails earlier and more plainly. The S19j Pro grosses 104 times $0.0332, which is $3.45 per day, while its power alone costs $3.68 per day at $0.05 per kWh. It is underwater before pool fees, before downtime, before anything breaks. There is no payback period to drift. There is a daily loss to stop.

The same payback with difficulty drift, month by month

Difficulty adjusts every 2,016 blocks, roughly every two weeks, to hold the average block time at 10 minutes. When total network hashrate rises, difficulty rises with it, and a fixed machine earns less. October delivered two rises above 5 percent inside a fortnight. A drift scenario of 3 percent per month is therefore not pessimism. It is slower than the month the inputs were measured in.

Hold the Bitcoin price flat, so price neither rescues nor sinks the plan, and let revenue fall 3 percent per month as difficulty compounds. Costs stay as stacked above. The monthly picture changes completely.

MonthNet for the monthCumulative netStill owed on $3,800
1$119$119$3,681
6$83$606$3,194
12$47$973$2,827
18$16$1,143$2,657
22Turns negativePeaks near $1,150About $2,650, never recovered

Read the last row twice. Under this scenario the machine recovers about 30 percent of its purchase price, then crosses into monthly loss at month 22, still inside the window before the expected 2028 halving cuts revenue again. At a 5 percent monthly drift the monthly net turns negative at month 14, with about $715 recovered. A faster Bitcoin price rise could offset the drift, and a price fall makes every row worse. Neither possibility is a plan. The sourced pattern is that calculators show a snapshot, and the snapshot is the most flattering day the machine will ever have.

The decision rule

Run any purchase through three checks, in this order. First, the full stack static test: revenue at a dated hashprice, minus power, pool fees, downtime, and a repair reserve. A machine that fails this test, like the S19j Pro at $0.05 power, is a pass today. Second, the drift test: rebuild the payback month by month at 3 percent monthly difficulty growth with the price held flat. The purchase survives only if the price is recovered with room to spare. Third, the halving test: any payback path that needs income after the expected 2028 halving is counting revenue at a rate the protocol has already scheduled to fall.

A machine that passes all three at your real power rate and your real delivered price is a defensible purchase. A machine that passes only the static test is a snapshot, and snapshots are how mining hardware gets sold. Run the drift version before the money moves, while the kill criterion can still do its job.

— Sterling Cole, Money Writer — Earning Desk