How long does an ASIC miner last, and what ends its life

Ask how long a miner lasts and you will get answers between three and seven years, all of them defensible, because the question hides two separate ones. How long will the hardware physically keep working, and how long will running it still make money. On almost every site, the second number arrives first.

Physical life

The silicon itself is durable. What wears out around it is mechanical and thermal: fans, power supplies, solder joints stressed by heating and cooling cycles, and connectors that loosen. Kept in clean, cool, dry air at a steady load, a machine running for five or more years is unremarkable. Kept in a dusty shed at forty degrees, cycled on and off daily and overclocked, the same machine can be a wreck in eighteen months.

So physical life is mostly a site quality question rather than a model quality question. The conditions that extend it are the same ones in maintenance and spares: filtration, hot and cold separation, stable power and a maintenance calendar.

Economic life, which is what actually matters

A machine stops earning when the network's revenue per unit of hashrate falls below what it costs to run. Every year, more efficient machines join the network, difficulty rises, and the revenue per terahash trends downward. Your machine did not change. Its position against the field did.

That is why efficiency, not hashrate, sets useful life. A machine at a low joules per terahash figure survives further into a declining market than a machine that produces the same terahashes less efficiently, and it survives longest of all on cheap power. The arithmetic is in the breakeven power price: economic life ends on the day the network's earnings cross your machine's breakeven, and where that day falls depends as much on your electricity contract as on the hardware.

The two things that move the date

A halving cuts revenue in one step and can retire a whole efficiency class overnight, which is the planning problem in what the halving does to a fleet. A rising coin price does the opposite and can bring retired machines back into service, which is why old hardware never quite disappears from the market. Neither is predictable, so plan on the arithmetic and treat the price as upside.

Extending life on purpose

• Underclock. The single most effective lever. Better efficiency for less hashrate pushes the breakeven further out, often by quarters. See tuning a fleet.
• Move it to cheaper power. A machine that is finished at twelve cents may be fine at four. Hosting or a different site can be worth more than the resale value.
• Improve the site, not the machine. Cleaner air and better separation buy years of physical life at very low cost.
• Run it seasonally. Some operators keep older machines for the cold months, when cooling is free and heat is useful, and park them in summer.
• Sell it while it still has a buyer. Not an extension, but usually the most valuable option, and the one people take too late.

What used hardware really tells you

A used machine's age matters less than its efficiency class and the conditions it lived in. A three year old machine from a clean, well run hydro site can be a better purchase than an eighteen month old machine from a dusty container that cycled daily. Since you usually cannot verify the history, the practical protections are the test record, the condition notes and a warranty you can actually claim on, which is the whole argument in new or used.

Residual value

Machines hold value in a rising market and lose it quickly in a falling one, and the drop is steepest for the least efficient models. That is asymmetric in a useful way: the time to sell older hardware is when everything feels good, which is exactly when nobody wants to. Keep residual value out of your payback model, as in payback on a mining rig, and treat any recovery as a bonus rather than a plan.