Antminer S23 & S23 Hydro — Price, Profitability and Solo Odds

An Antminer S23 Hyd was listed at $12,299 on 16 September 2026, and at that price it returns its purchase cost in 737 days at $0.04/kWh or 2,015 days at $0.12/kWh. The S23 Hyd 3U at $14,879 does it faster: 446 days to 1,219 days. Every figure on this page comes from difficulty, hashrate and real block rewards read off our own Bitcoin node — not from a vendor calculator.

Key takeaways

• S23 Hyd: $12,299. S23 Hyd 3U: $14,879. OneMiners listed prices, 16 September 2026 — but the same 3U appeared at $79,999 elsewhere the same day.
• Break-even runs 737 to 2,015 days for the Hyd depending on your electricity rate, and 446 to 1,219 days for the 3U. The 3U is the cheapest path to break-even at every rate.
• Bitcoin hashprice was $0.03788 per TH/day on 16 September 2026, of which transaction fees contributed only 0.67% — measured across 144 consecutive blocks.
• The S23 Hydro is still the most efficient Bitcoin miner you can buy: 580 TH/s at 9.5 J/TH, the first model under the 10 J/TH line.
• Solo mining is not income. An S23 Hyd averages 29.9 years per Bitcoin block. On Bitcoin Cash — same chip, same algorithm, same 3.125 subsidy — it averages 37 days, for 16% less expected value.

How much does an Antminer S23 cost in September 2026?

Prices below are a single vendor’s listings on a single day, named and dated. That is deliberate, and the next paragraph explains why.

Why there is no average price here. The same S23 Hyd 3U appeared between $14,879 and $79,999 on the same day across different sellers — a factor of five, with two of those listings sitting on the same vendor’s own site. An average would be an artefact of which sellers you chose to exclude: include the high outlier and you get roughly $39,000, drop it and you get roughly $25,000, take the batch listing alone and you get $14,879. None of those three numbers is more true than the others. Check who you are buying from before you compare prices.

Two practical consequences. First, an S23 quote is only meaningful with a vendor name, a batch and a delivery date attached — a bare number is noise. Second, treat the secondary market with the caution it deserves: hydro-cooled machines are sold used after facility turnover, and a cheap 3U may arrive with thousands of hours of coolant cycles behind it.

What does an Antminer S23 earn? Measured, not estimated

All network figures below were read from SoloFury’s own Bitcoin full node on 16 September 2026 at block 967,317, averaged over the previous 144 blocks:

That fee figure deserves a moment. Transaction fees made up 0.67% of miner revenue — meaning 99.3% of what an S23 earns today is subsidy. Getting this number requires reading 144 consecutive blocks from a node; no public dashboard publishes it. It also sets up the 2028 halving bluntly: fees are nowhere near replacing the subsidy, so the next cut lands almost undiluted on revenue.

As a sanity check, the 912.3 EH/s our node computes sits about 3% above the 885 EH/s reported by a third-party aggregator on the same day — a normal spread, since hashrate is always an estimate derived from a chosen time window.

Antminer S23 break-even by electricity rate

Simple payback on hardware cost, at the listed prices above and the hashprice above. No hosting fees, no pool fees, no resale value, no difficulty growth.

S23 Hyd — 580 TH/s, 5,510 W, $12,299. Grosses $21.97/day.

S23 Hyd 3U — 1.16 PH/s, 11,020 W, $14,879. Grosses $43.94/day.

The 3U wins on capital efficiency at every rate — it costs 21% more than the Hyd and produces twice the hashrate, because the September batch priced it as a volume machine. The gap widens as power gets expensive: at $0.12/kWh the 3U breaks even in a third of the time the Hyd needs.

Read the horizon, not just the number. A 2,015-day break-even is 5.5 years, and it crosses the 2028 halving — which cuts the subsidy in half and, on today’s fee share of 0.67%, cuts revenue by very nearly the same proportion. Any payback window that runs past early 2028 should be modelled twice: once as shown, once with revenue halved from that date. Difficulty growth works against you too, and BTC price appreciation is the only force working for you. Model your exact rate in our solo mining calculator before spending anything.

Network difficulty, hashrate and average block reward measured on SoloFury’s own Bitcoin full node on 2026-09-16 at block 967,317. Not a vendor estimate.

The whole S23 family

Bitmain didn’t release one machine — it released a lineup, each model aimed at a different operation. Specifications verified against vendor spec sheets and ASIC Miner Value:

A few things jump out. The air-cooled S23 is the only model a normal person can plug into a wall and run without three-phase power and a plumbing loop — it’s the home-and-small-farm machine. Everything above it is built for facilities. And that 3U at the bottom is genuinely absurd: a single box that matches roughly a dozen Antminer S19 units, or four-plus air-cooled S21 XPs, while sipping under 10 J/TH. We took it apart in detail in our S23 Hyd 3U petahash breakdown.

Why 9.5 J/TH is the whole story

It’s tempting to obsess over headline hashrate, but for any miner who pays for electricity, the number that decides whether you make money is efficiency — joules per terahash. It’s the rent your machine pays the power company for every unit of work.

To see how far the S23 moved the line, line it up against its ancestors. The popular S21 XP did 270 TH/s at 13.5 J/TH. The S23 does 318 TH/s at 11 J/TH air-cooled — roughly 18% more hashrate at 18% better efficiency in the same form factor. The S23 Hydro pushes that to 9.5 J/TH, which Bitmain frames as about 41% better efficiency and 69% more hashrate than the S21 generation. If you want the full arc from the 16nm S9 era to today’s sub-10 J/TH silicon, we walked through it chip by chip in our Bitcoin ASIC chip evolution guide.

Why does this matter now? Because the April 2024 halving cut the subsidy to 3.125 BTC, and the 2028 halving will cut it again to 1.5625. At the hashprice we measured on 16 September 2026 — $37.88 per PH/day — the S23 line clears comfortably. Halve the subsidy with fees still at 0.67% of revenue and that figure drops close to $19 per PH/day at an unchanged BTC price, which is where machines much worse than ~11 J/TH go underwater. Efficiency isn’t a bragging right; it’s a runway. (The full post-halving picture is in our halving aftermath analysis.)

Air vs hydro vs immersion: which S23 is “yours”?

The cooling choice is really an infrastructure choice, and it’s the biggest fork in the road when buying an S23.

• Air (S23, 318 TH/s): Plug-and-play on standard power. Loud (~75 dB, like a vacuum cleaner) and it dumps heat into the room. The only S23 a home miner should consider — and a beast for one.
• Hydro (S23 Hyd, 580 TH/s): Water cooling drops noise to around 50 dB and lets the chip run flat-out without throttling. The catch: you need a coolant loop and 380–415V three-phase power. Facility gear.
• Immersion (S23 IMM, 442 TH/s): For operations already running dielectric-fluid tanks. Excellent thermals, niche audience.

If you’re weighing the cooling question seriously, we did a full teardown of the trade-offs — capital cost, noise, lifespan, density — in our air vs hydro vs immersion comparison. The short version: hydro wins on efficiency, noise and lifespan, but only pays off at scale.

Bitcoin or Bitcoin Cash? The controlled comparison

This is the comparison almost nobody runs properly, because it only works if you hold everything constant. Bitcoin and Bitcoin Cash share the same SHA-256 algorithm, the same 3.125 block subsidy, and the same 10-minute target. Point the same S23 at either one and literally one variable changes: difficulty.

Both sets of figures below come from our own nodes, measured the same day.

Look at what those two ratios do to each other. Bitcoin Cash is 295 times easier to mine, and its coin is 349 times cheaper. The two factors very nearly cancel: the BCH hashprice lands at 84% of Bitcoin’s. That near-cancellation is not a coincidence — it is miners moving hashrate between two chains that pay in different coins until the rates converge.

How long you wait for a block, same hardware

What you actually net, S23 Hyd at $12,299

You pay 16% of expected value to go from 30 years to 37 days. A solo miner is not choosing between “more profitable” and “less profitable” — the expected value is nearly the same. The choice is how much variance you are willing to carry. On Bitcoin an S23 Hyd waits an average of 29.9 years for one block. On Bitcoin Cash, the same machine, the same algorithm, the same 3.125 reward: 37 days. The difference costs you 16% in expected daily value.

Which side you want depends on something no calculator can tell you: whether you are funding this from other income and can wait, or whether you need the machine to produce a result you can see. If you want the deeper mechanics of the BCH side, our ASERT difficulty explainer and the BCH solo mining guide are the next reads.

Bitcoin Cash uses ASERT, which retargets difficulty on every block, so its difficulty moves more than Bitcoin’s. The figures above are averaged over the last 144 blocks, measured on our own nodes at the timestamp shown.

Firmware, tuning, and squeezing out more

An S23 ships ready to mine, but the open-firmware scene can sharpen it. Aftermarket operating systems like Braiins OS+ and LuxOS add per-chip auto-tuning, watt-anchored power targets, and a low-power efficiency mode — the toolkit that has historically unlocked 10–15% better efficiency on older Antminers. If you’ve tuned an S21 before, the workflow will feel familiar; we cover the principles in our undervolting guide, and the same logic of trading a sliver of hashrate for a chunk of efficiency applies straight to the S23.

What an S23 means for solo mining

Here’s where most hardware reviews stop and we keep going — because raw hashrate only matters once you decide what to do with it. At the difficulty measured on 16 September 2026, here is what each S23 faces on Bitcoin:

The median column is the one people skip and shouldn’t. Because block-finding is a Poisson process, the median wait is about 69% of the average — half of all S23 Hyd operators would find their first block inside 20.7 years, and half would wait longer. The average is dragged upward by the long tail.

The sentence that matters more than any table on this page. At $0.07/kWh an S23 Hyd has an expected value of $21.97 a day. In a pool you actually collect that $21.97, every day. Solo, you never collect it: you collect zero until you find a block, and at 580 TH/s the average wait is 29.9 years. Then, in one shot, you take 3.146 BTC — about $240,000 at today’s price. The expected value is identical. The distribution is not. Choose on the distribution, because that is the only thing that differs.

Those are averages, not schedules, and people genuinely win early. On April 2, 2026, a solo miner running roughly 230 TH/s (about a single Antminer S21) solved block 943,411 for 3.139 BTC — around $210,000 at the time — against daily odds near 1-in-28,000, confirmed on mempool.space and reported by CoinDesk. An S23 is faster than that S21. In the twelve months to June 2026, solo pools found roughly 20 BTC blocks — about one every 18 days — including wins on hardware as small as a NerdQAxe or Bitaxe pulling a few TH/s. The math says “decades”; variance says “maybe Tuesday.” If the lottery framing is new to you, our explainer on what lottery mining is, the deeper odds breakdown, and the full Poisson variance math are the right next reads.

This is exactly what SoloFury is built for. Our non-custodial pool lets you solo-mine BTC, BCH, BC2, BCH2 and XEC with one machine, with the block reward paid straight to your own wallet and a 1% fee. Check which chain is easiest to hit right now on the live Network Radar, then point your S23 at it in about two minutes via the Start guide.

Verdict: who should buy the S23?

Buy the air-cooled S23 if you want the most powerful single-box air miner money can buy in 2026, you have sub-$0.08/kWh power, and you can tolerate the noise and heat — it’s a phenomenal solo-mining engine and a credible home/small-farm workhorse. Buy the Hydro, 2U or 3U if you run a facility with three-phase power and a cooling loop and you’re optimizing for the lowest J/TH on the market to survive the next halving cycle. Of those, the 3U is the one to buy on capital efficiency — at the September batch price it breaks even faster than any other model at every electricity rate.

Look elsewhere if you want a quiet desk miner — there, a Bitaxe or the new Nexus S1 makes far more sense than a 75 dB industrial machine. And if your power costs more than $0.12/kWh, run the calculator twice before spending a cent — at those rates even the most efficient miner on earth needs 5.5 years to return its price, and that window crosses a halving.

Plan your S23 with SoloFury’s free tools

No signup, no email, just the numbers:

• Solo mining calculator — odds, break-even and expected return for the S23’s exact hashrate and your power price.
• Profitability tool — compare expected return across BTC, BCH, BC2, BCH2 and XEC for one machine.
• Network Radar — live difficulty, retarget countdown and hashrate for every SHA-256 coin, so you always aim at the easiest chain.
• Start mining — point your S23 at SoloFury in about two minutes, payouts straight to your wallet.