Antminer S23 Hyd 3U

The Antminer S23 Hyd 3U is the first single Bitcoin miner to exceed one petahash per second: 1,160 TH/s at 9.5 J/TH — roughly 11 kW — in a 3U rack-mount hydro chassis, shipping since Q1 2026. It compresses what an entire aisle of S9-era machines once produced into one server slot, and it moves the solo mathematics into territory no single unit has ever occupied: a 1-in-16 yearly Bitcoin ticket, and an expected Bitcoin Cash block every three weeks, from one box. This guide covers the milestone, the infrastructure truth, and that math — for the S23 family itself, see our dedicated S23 review.

Key takeaways

• The petahash threshold is crossed: 1,160 TH/s from one unit at 9.5 J/TH — the sub-10-joule era’s flagship, and a ~3.6× density jump over the air-cooled S23 in a fraction of the rack space.
• It is facility hardware, full stop: ~11 kW continuous, three-phase power, a hydro loop, 3U racking. The honest home-miner relationship with this machine is hosting contracts or admiring from a distance.
• The solo math is historic: one unit expects a BTC block in ~16 years (a genuine 1-in-16 per year), a BCH block every ~22 days, and multiple daily blocks at the ladder’s bottom — the first single machine to make Bitcoin solo mining a plausible rather than poetic position.
• Firmware freedom lags the silicon: third-party firmware (Braiins OS+, VNish) is not yet supported on the S23 generation as of early 2026 — a real consideration for tuning-dependent operators.
• Every deployment raises the water level: petahash-per-slot density accelerates network hashrate growth — the difficulty everyone else’s odds are computed against.

The threshold: how one box became a small farm

Numbers this large need a ruler. In the S9 era, 1,160 TH/s meant roughly eighty machines — an aisle of shelving, tens of kilowatts of fans, a wall of noise. The S23 Hyd 3U delivers it from three rack units, silently by mining standards, because hydro cooling removes the acoustic and thermal ceilings that cap air-cooled density. It’s the culmination of the trajectory our chip-evolution history traces — 5nm-class silicon, sub-10 J/TH efficiency, and liquid thermal engineering compounding into a machine that redefines the unit of account: fleet operators now think in petahash per rack rather than terahash per shelf. Shipping began Q1 2026 after the World Digital Mining Summit unveiling, and institutional orders are already reshaping deployment maps. One consequence lands on every reader of this site regardless of what they run: machines like this are why network hashrate crossed a zetahash and keeps climbing — the denominator under everyone’s odds.

The infrastructure truth

We’d rather disappoint you in paragraph six than after a purchase: this is not home hardware, and no garage heroics change the three-phase and hydro-loop math. The realistic paths to its hashrate class for an individual are hosting (colocation facilities run the infrastructure; you own the machine and the decisions — including which pool it points at) or renting (marketplaces sell the same hashes by the hour with zero capex — the rental math here). Both preserve the part of this story that actually belongs to individuals: the solo mathematics below.

The solo math nobody has written: one petahash, five lotteries

Here is why this machine earns an article on a solo-mining site (live difficulties, July 2026):

Pause on the first row, because it is genuinely new: no single machine in Bitcoin’s history has held a 1-in-16 yearly shot at the grand prize. Over a five-year deployment that compounds to roughly a 27% chance of at least one full Bitcoin block — solo mining BTC stops being poetry and becomes a position a rational operator might actually hold, especially as the hybrid logic of pointing owned or hosted petahash at the lottery matures. The second row is arguably more disruptive: an expected BCH block every 22 days means a Hyd 3U on solo BCH behaves like an income stream with three-week variance — a category of machine-economics that simply didn’t exist before Q1 2026. And the bottom rows dissolve the lottery frame entirely. As ever, these are Poisson expectations with real variance (the math here), and the chain trade-offs are priced in the ladder comparison — but the headline stands: one box now spans the whole ladder, from plausible BTC winner to block-a-shift workhorse.

Verdict: how to think about the petahash era

If you operate facilities: the Hyd 3U is the density king — 9.5 J/TH, petahash per 3U, and hydro heat begging for reuse — with the firmware-freedom gap as the one genuine caveat against the mature S21 ecosystem. If you’re everyone else: don’t buy the box; steal its lesson. The petahash era means difficulty will keep climbing on the back of machines like this, which makes the two individual strategies sharper, not weaker — choose your rung deliberately (the same ladder logic applies whether you hash with 1 TH or 1 PH, and the small chains are where individual hashrate still expects wins), and rent the milestone when you want to feel it (petahash-hours are for sale by the hour, no plumbing required). The machines get bigger; the math stays honest; and the coinbase still pays whoever’s share clears the bar — that part, no threshold ever changes.