at block 961,632
at block 961,640
hashrate, early Sept
in a single week
Proof-of-work mining is a cyclical business, and most of its surprises arrive from the edges rather than the centre. While the industry watched the hashrate race in Bitcoin and Kaspa, a peripheral corner of the market produced the year's most consequential hardware story: a chain split that carried Bitcoin’s ledger onto BLAKE2b, and in doing so turned a shelf of obsolete Siacoin ASICs into the only machines capable of securing it.
The detail that matters is not that BTCB2 resembles Bitcoin. It is that BTCB2 is Bitcoin’s ledger — split away on 8 August 2026 at block 961,632, carrying the same balances and the same supply schedule, and then, three weeks later at block 961,640, replacing SHA‑256d proof-of-work with BLAKE2b. Block reward stayed at 3.125. Target block time stayed at roughly ten minutes. Only the hash function changed.
That single substitution did something no amount of market commentary could: it made the world’s entire installed base of Bitcoin mining hardware useless on the new chain, and it handed the work to an orphaned hardware family instead.
A chain does not care how much hashrate exists. It cares how much hashrate speaks its language — and on 30 August, the language changed.
Why an S21 cannot follow the ledger it helped build
An ASIC is not a computer. It is a hash function cast into silicon. An Antminer S19 or S21 contains circuits that compute SHA‑256d and nothing else — there is no firmware update, no efficiency penalty, no partial compatibility. When the proof-of-work target changes, that silicon does not slow down. It stops being relevant.
How it unfolded
Nobody has been building these chips
The reason this market moved so violently is not demand. It is supply. Bitcoin’s hashrate can respond to a price signal because a global supply chain is already producing SHA‑256 ASICs at volume. BLAKE2b has no such supply chain. Goldshell and iBeLink built these machines for Siacoin, a network whose economics had long since stopped justifying new production.
So the eligible hardware pool is effectively fixed at whatever already exists — much of it depreciated, much of it switched off in a garage or a warehouse. When a fixed pool of machines is pointed at a Bitcoin-scale reward schedule, the per-machine share of that reward is unlike anything in mainstream mining. A single 19 TH/s unit represents roughly a third of one percent of a 5.1 PH/s network. A flagship Bitcoin miner is about two hundred-thousandths of one percent of Bitcoin.
That asymmetry is the whole story, and it is also the whole risk: it exists precisely because the network is small, young and thinly traded. Both halves of that sentence are true at the same time.
The six machines that can mine BTCB2
OneMiners lists six BLAKE2b units in its BTCB2 collection, from a compact entry machine to a flagship industrial unit. Together they represent roughly 77 TH/s of the lineup’s combined capability. Specifications below are manufacturer figures; prices were read live from oneminers.com on the date of publication. This is the eligible fleet, not a buying recommendation — for a ranked verdict on value, noise and payback, see the buyer’s guide.
Efficiency, and what a terahash costs to buy
Two comparisons do most of the work when choosing between these machines. The first is thermal efficiency — joules consumed per terahash produced, where lower is better. The second is capital efficiency — what you pay, up front, for each terahash of capability.
What the plot actually says
The BM‑S3 occupies the strongest engineering position in the lineup by a clear margin — it is simultaneously the fastest machine and the only one under 200 J/TH. But capital cost tells a different story: at $421 per terahash it is the most expensive way to buy hashrate in the group, while the least efficient machine, the BM‑S1 Max, is the cheapest at $383.
That inversion is characteristic of a young network. On Bitcoin, where electricity routinely consumes most of gross revenue, efficiency dominates every purchasing decision. On a chain where the reward per unit of hashrate is currently this high, power is a much smaller share of the equation, and price per terahash carries more weight than joules per terahash. That relationship will not hold permanently — it inverts the moment enough hashrate arrives. The buyer’s guide carries that argument down to a per-machine verdict.
Full specification table
| Model | Hashrate | Power | J / TH | List price | $ / TH | Daily revenue* |
|---|---|---|---|---|---|---|
| Goldshell SC Lite | 4.4 TH/s | 950 W | 215.9 | $2,999 | $682 | $19.68 |
| Goldshell SC5 Pro | 11 TH/s | 2,820 W | 256.4 | $4,599 | $418 | $48.50 |
| iBeLink BM‑S1 Max | 12 TH/s | 3,150 W | 262.5 | $4,599 | $383 | $53.51 |
| Goldshell SC5 Pro II | 14 TH/s | 3,300 W | 235.7 | $5,599 | $400 | $61.77 |
| Goldshell SC6 SE | 17 TH/s | 3,330 W | 195.9 | $6,999 | $412 | $74.93 |
| iBeLink BM‑S3 | 19 TH/s | 3,100 W | 163.2 | $7,999 | $421 | $107.44 |
| Hashrate and power are manufacturer-rated. J/TH and $/TH are calculated from those figures. List prices read live from oneminers.com on 21 September 2026. *Daily revenue figures are gross, before electricity, and were published by OneMiners on 16 September 2026 — they move with coin price and network difficulty and should be re-checked before any purchasing decision. | ||||||
Where this hardware actually runs
A BLAKE2b unit is smaller than a modern Bitcoin miner, but it is still a 3 kW industrial appliance that runs continuously and is loud enough to require hearing protection at close range. The constraint on this hardware has never been the chips — it is stable power, adequate cooling and someone on site when a fan fails at 3 a.m.
OneMiners operates a hosting network across 20 sites. The relevant number for an operator is not a fleet hashrate figure — as a hosting operator we do not claim one — but contracted capacity, electricity cost and uptime.
What resolves this, and when
BTCB2 solved a specific problem for a specific group of people: owners of BLAKE2b ASICs who had watched the economics of Siacoin erode beneath them. Demand for both used and new Goldshell and iBeLink hardware has risen sharply as a result, and a hardware category that was effectively written off is being sourced, tested and redeployed.
The honest framing is that this is a supply anomaly, not a permanent state. The current per-machine share of network reward exists because only a few petahash of eligible hardware exist in the world. That gap closes in one of two ways: inventory comes off shelves and out of storage — which is already happening at over 30% a week — or manufacturers restart BLAKE2b production. Either path compresses the advantage.
What stays true through the compression is the underlying arithmetic: price per terahash determines how quickly hardware pays for itself, efficiency determines how long it stays viable afterwards, and electricity cost determines the floor under both. Those three variables outlive any individual chain.
BTCB2 is a young network with thin liquidity, limited exchange and wallet support, and an unproven security budget. That is not a footnote to the opportunity — it is the same fact, described from the other side.
Industrial power, cooling and monitoring for Goldshell SC and iBeLink BM‑S units — with on-site repair and dead-on-arrival protection.
See the BTCB2 miners →Explore hostingInformational only. Not financial advice. Mining outcomes depend on coin price, network difficulty, electricity cost and regulation, all of which change. BTCB2 is a young chain with thin liquidity and limited exchange support, so figures can move very quickly in either direction. Do your own due diligence.
Disclosure: OneMiners sells and hosts the hardware described in this article.
Sources: Chain-split and activation data, network hashrate and daily revenue figures — OneMiners, “What Is BTCB2 (Bitcoin BLAKE2b)?” and “BTCB2 Mining Profitability 2026” (16 September 2026). Hardware specifications and list prices read live from oneminers.com product pages, 21 September 2026. J/TH and $/TH calculated by the OneMiners research desk from those figures.
