Antminer S9 Profitability in 2026: Can It Still Pay?

At 1,350 W an S9 consumes roughly 32.4 kWh per day. At ten cents per kilowatt hour that is about $3.24 per day in electricity, which comfortably exceeds what 13.5 TH/s earns at 2026 difficulty. At three cents the daily power cost drops near a dollar, and the machine moves from loss-making to marginally positive depending on Bitcoin price.

The practical break-even for an S9 sits somewhere between two and four cents per kilowatt hour, and it moves every difficulty adjustment. If you are paying a residential tariff anywhere in North America, Europe or Australia, an S9 loses money every hour it runs. That is arithmetic, not pessimism.

Three cases genuinely work. Curtailed or stranded power — flare gas, hydro spill, behind-the-meter solar that would otherwise be wasted — where the marginal cost of electricity is near zero. Heat reuse, where the S9 replaces a resistive heater you were going to run anyway and the mining revenue is a rebate on your heating bill. And learning, where you want firmware, pool configuration, PSU swaps and thermal tuning experience before spending real money.

The fourth case is parts. S9 hashboards, control boards, fans and APW3 PSUs still keep older fleets alive, so a cheap unit can be worth more disassembled than hashing. We see operators buy lots specifically as donor stock.

Daily revenue equals your hashrate share of the network multiplied by daily block issuance and fees, minus pool fee. Daily cost equals kilowatts multiplied by 24 multiplied by your rate. Subtract, and you have gross margin per machine per day. Divide the delivered purchase price by that margin and you have payback in days — if the margin is negative, there is no payback at any price.

Run the model at three Bitcoin prices and two difficulty growth assumptions. A legacy machine is highly sensitive to both, so a spread tells you far more than a single point estimate. Rerun it after every difficulty adjustment rather than trusting a figure from last quarter.

Hosting contracts price per kilowatt, not per terahash, so a hosted S9 pays commercial-rate power plus a hosting margin for hashrate that current hardware produces at a fraction of the energy. Most reputable facilities will not accept sub-20 J/TH-class machines any more, and the ones that do are selling you rack space rather than a return.

If you want hashrate without running hardware yourself, host an efficient machine instead. The same monthly hosting spend applied to a modern unit produces multiples of the revenue for the same kilowatt.

Under a few hundred dollars, a low-wattage 110 V home miner gives you real hashrate at a fraction of the noise and heat, and it is a better teaching platform than an S9. In the low four figures, a tested S19j Pro or M30S++ delivers ten to twenty times the hashrate at roughly a quarter of the joules per terahash.

If your power is genuinely free or curtailed, buy legacy units by the pallet, expect failures, and keep spares. Every used and refurbished unit we ship is bench-tested under load with hashboards verified chip by chip, so you know what you are getting before it leaves Hong Kong.