Every miner eventually faces the same question: is it time to replace the machines? A fleet refresh that PowerCompute announced on September 23 is a useful worked example, partly because of what the company did and partly because of what it chose not to do. It upgraded, but it did not reach for the newest, most efficient hardware on the market. The reason it made that call is the whole lesson in when an upgrade actually pays.
What PowerCompute actually did
The move is a cascade across two sites, not a simple swap. In Oklahoma, the company installed 1,000 Antminer S19 XP units, which pushed the older Antminer S19j Pro machines out of those slots. Rather than scrap the displaced S19j Pro units, it redeployed them to its Mississippi site, where they replaced the genuinely old gear: base Antminer S19 machines averaging about 95 TH/s at roughly 34 J/TH. Another 940 S19 XP units were ordered for Mississippi on top of that.
The headline number was "nearly 39% more hashrate per machine replaced," which compares the new S19 XP at about 132 TH/s against those base S19 units at 95 TH/s. Across the whole operation, total hashrate rises about 11.8% to roughly 862 PH/s, funded with around $400,000 from cash and bitcoin holdings. Two things stand out: the oldest, least efficient machines are the ones that leave, and nothing gets thrown away that still earns.
Why they didn't buy the newest hardware
Here is the part worth sitting with. The S19 XP is a late-2022 machine at 21.5 J/TH. The current flagship, the Antminer S23 Hydro, runs near 9.5 J/TH, more than twice as efficient. So why upgrade to a three-year-old model instead of the best available?
Because the best machine on a spec sheet is not always the best machine for your payback. An upgrade is worth it when the money you save (or the extra bitcoin you earn) covers the purchase price before the machine is obsolete. Efficient prior-generation rigs like the S19 XP typically cost far less per terahash than a brand-new flagship, so the cheaper machine can pay for itself faster, even though the flagship wins on efficiency. The flagship's efficiency edge only pays off decisively in two situations: when your electricity is expensive enough that every extra joule really hurts, or when you will run the machine long enough for the efficiency gap to compound. At a moderate power rate, an efficient used machine bought cheaply is often the better return.
The upgrade math, in plain terms
Three numbers decide it.
The first is efficiency against your power rate. A machine's breakeven point is roughly its efficiency multiplied by your electricity cost. A 34 J/TH machine and a 21.5 J/TH machine paying the same rate have very different breakevens, and the gap widens as the rate rises. That is exactly why PowerCompute retired the 34 J/TH base S19 first: at any given rate, it hits unprofitability soonest.
The second is purchase price per terahash. Divide what a machine costs by its hashrate and you can compare a cheap old rig against an expensive new one on the same footing. A lower price per terahash means a shorter payback, which is the whole reason efficient older gear can beat a flagship on return.
The third is remaining useful life against difficulty. Network difficulty has trended up over time, and rising difficulty quietly shrinks every machine's bitcoin output month after month, upgrade or not. A rig that is four or five years old and near the end of its life is easy to justify replacing. Recent hardware is not, because you would be scrapping earning years you already paid for.
The quiet lesson: cascade, don't scrap
The smartest part of PowerCompute's move is the redeployment. The displaced S19j Pro machines were not junk, they were just no longer the best fit for the Oklahoma site, so they went to Mississippi to knock out something worse. If you run more than one location, or more than a few machines, the upgrade question is not only "new versus old." It is also "where does each machine earn the most," and the oldest, thirstiest unit is usually the one that should actually leave the fleet.
What this means for your rigs
The single biggest input to your own upgrade math is your power rate, and it cuts both ways. A low, fixed rate lets you keep running a paid-off older machine profitably for longer, which means you can wait for the right price on a replacement instead of upgrading on the manufacturer's schedule. A high rate forces the issue sooner, because inefficiency costs you more every day. Either way, the decision is arithmetic, not hype, and it is worth actually running the numbers before buying anything.
If you want to compare an old machine against a new one at your real rate, model both in our profitability calculator, which uses live bitcoin price and difficulty. We sell Whatsminer hardware and Antminer models too, and we host at $0.07/kWh all-in in Southeast Texas, so we are happy to talk through whether an upgrade pencils out for you rather than just sell you the newest box.
Frequently asked questions
When should I upgrade my bitcoin miner?
Upgrade when the savings or extra earnings from a new machine cover its purchase price before it becomes obsolete. The decision rests on three things: the machine's efficiency (J/TH) against your electricity rate, its purchase price per terahash, and how much useful life your current rig has left as network difficulty keeps rising. A four-to-five-year-old machine near end of life is usually worth replacing; recent hardware often is not.
Is it better to buy a new flagship or an efficient older ASIC?
It depends on your power rate and how long you will run the machine. Newer flagships like the Antminer S23 Hydro (about 9.5 J/TH) are far more efficient, but efficient prior-generation rigs like the S19 XP (21.5 J/TH) usually cost much less per terahash, so they can pay back faster at moderate electricity rates. The flagship wins decisively when power is expensive or the runtime is long. Run both cases before deciding.
What is a good efficiency (J/TH) for a miner today?
As of 2026, the newest flagship machines run near 9.5 J/TH, strong current-generation air-cooled units are around 15 to 21 J/TH, and older rigs at 30 J/TH or more are the first to become unprofitable when power costs rise. Lower J/TH means less electricity per unit of mining work. The right target depends on your power rate: cheap power lets higher-J/TH machines stay profitable longer.
Does my electricity rate affect the upgrade decision?
Yes, more than any other factor. A machine's breakeven is roughly its efficiency times your power rate, so a high rate makes inefficient rigs unprofitable sooner and speeds up an upgrade's payback, while a low, fixed rate lets you run older paid-off machines profitably for longer and upgrade on your own schedule. Fathom Labs hosts at $0.07/kWh all-in with no performance fees.
Trying to decide whether to refresh your fleet? Talk to us and we will run the payback math for your exact hardware and rate.