Payback on a mining rig: how to work it out without fooling yourself

Payback is the simplest question in mining and the one most often answered wrong. It asks how long the machine takes to return what it cost. The arithmetic is trivial. The difficulty is entirely in the inputs, and the standard online calculator gets at least two of them wrong by default.

What payback actually measures

Take the landed cost of the machine and divide it by the daily profit it produces. Landed cost, not the quoted hardware price: freight, duty, brokerage, power supply and your share of spares all belong in the numerator. Daily profit, not daily revenue: gross revenue minus electricity, pool fee and your operating cost per unit. The result is a number of days.

That is it. What follows is the discipline of not lying to yourself about either half of the fraction.

The three inputs people get wrong

• The hardware price. Using the ex works quote instead of landed cost understates the numerator, often by a fifth or more. Our piece on what a miner really costs covers what belongs in that figure.
• Static difficulty. Most calculators hold difficulty constant. Difficulty has trended up for the whole history of the network. A model that assumes today's hashprice for eighteen months is not a model, it is a snapshot repeated.
• Uptime of 100 percent. Nobody runs at 100 percent. Commissioning takes days, firmware updates take hours, and a curtailment clause takes whole afternoons. Ninety five percent is a good year on a well run site.

A worked example, honestly

Two hundred units, 230 TH/s each, 3,300 W each. Landed cost 3,400 dollars per unit, so 680,000 dollars for the batch. Power price seven cents. At a hashprice of 0.045 dollars per terahash per day, one unit grosses 10.35 dollars and burns 3.3 kW times 24 hours times 0.07, which is 5.54. Contribution is 4.81 dollars per unit per day, 962 for the fleet.

Now apply the corrections. Take 95 percent uptime and a two percent pool fee, and the fleet contributes roughly 880 dollars a day. Simple payback is 680,000 divided by 880, about 773 days, a little over two years. Then assume difficulty grinds upward and hashprice erodes ten percent a year, and the honest figure lands well past two and a half years. That is the number to take to a lender, not the nine months the calculator produced.

The residual value question

Payback treats the machine as worthless at the end, which it is not. An efficient machine that has paid itself back still has resale value, and a fleet planned in generations can sell the older half into a rising market rather than running it into the ground. That is the argument in phasing a buildout. Residual value is real, but do not put it in the payback calculation. Keep payback pessimistic and let the resale be upside.

Where honest payback changes the decision

Once you run it properly, two things usually happen. The efficient machine wins by more than expected, because its advantage compounds over every day of the payback period rather than showing up once. And the power price turns out to matter more than the hardware discount you spent three weeks negotiating. A cent per kilowatt hour on a 3.3 kW machine is roughly 290 dollars a year. Getting two cents off your power contract is worth more than getting five percent off the hardware, every time.