A press release crossed the wire this week advertising a bitcoin miner that draws 5,200 watts and produces 9,600 TH/s. Run the division yourself. That is 0.54 joules per terahash. The best shipping ASIC on earth — Bitmain's S21 XP — sits near 13.5 J/TH. MicroBT's M60S sits near 18.5. The claim is not incremental improvement; it is roughly 25 times better than the physical ceiling of the industry, announced not at a semiconductor conference but through a paid CryptoPotato placement from a company called ASICID Inc. When a single data point lands that far outside the distribution, the anomaly is the story. Tracing the hash that broke the ledger starts with one impossible number.
Context
ASIC efficiency improves on a curve every mining engineer knows by heart: roughly 5–10% per year, gated by process nodes and the slow death of Moore's Law. The industry crawled from 7nm to 5nm to 3nm, and each step bought fractions, not multiples. Post-halving — after April 2024 cut the block subsidy — the margin for error compressed further. Hashprice fell, older rigs went dark, and survivors consolidated around whoever could shave a joule. This is the market ASICID is selling into: one where a 2% efficiency edge wins procurement contracts worth nine figures.
ASICID claims three products — HomeRack, IDMINER 2, and IDMINER 1 — plus a Scrypt line for Litecoin and Dogecoin, all "mass-produced and shipping." The release discloses no chip fabrication partner, no process node, no third-party teardown, no live footage. Based on my audit experience — from the 2017 ICO cycle through the 2022 Terra forensics — this is the exact signature of a claim that expects to be read, not verified. Note the naming: a machine spec'd at 9,600 TH/s and 4×1,300W is labeled "HomeRack." That hashrate corresponds to roughly 48 S21 units, about 170 kilowatts and several hundred kilograms. No home rack swallows that. The label is a marketing fiction bolted onto an engineering one.
Core
Three numbers, one conclusion.
First, the SHA-256 line. HomeRack: 9,600 TH/s at 5,200W → 0.54 J/TH. IDMINER 2: 2,400 TH/s at 1,300W → 0.54 J/TH. IDMINER 1: 1,150 TH/s at 700W → 0.61 J/TH. The internal consistency is suspicious in the wrong direction: a real product line shows efficiency degrading as you scale down, not holding a flat wall across three SKUs. Flat numbers are a spreadsheet, not a silicon tape-out.
Second, the Scrypt line. HomeRack claims 3,200 GH/s at 5,200W → 1.63 W/GH. Antminer L7 delivers 9.5 GH/s at 3,425W → ~360 W/GH. To match ASICID's claimed output you would need roughly 337 L7 units. Those units draw about 1.15 megawatts. ASICID claims the same work in 5,200 watts — an underestimate of roughly 220x. No fabrication node in existence closes a 220x gap; that is not engineering, that is arithmetic that was never run.
Third, the revenue math. HomeRack advertises up to $25,590 per month. At $0.06/kWh, 5,200W costs about $225/month in electricity. That implies a ~99% net margin. Real ASICs at that power price net single digits to ~30%, with payback windows of 12–40 months. Reverse-engineer the $25,590 and you land on a bitcoin price near $138,000; the IDMINER 1 implies roughly $126,000. All three models are internally consistent because they were computed from one optimistic assumption set — and, critically, the revenue scales with hashrate while ignoring power draw entirely. The manufacturer did not underestimate electricity by accident. It underestimated it by roughly 32x to manufacture the illusion of profit.
Then the control point. The units ship preconfigured, locked to ASICID's own "zero-fee pool." That is the only ecosystem integration disclosed. No third-party pool lists its hashrate. No independent miner has posted a teardown. An open pool is a public ledger; a captive pool is a settlement engine you cannot audit. Auditing the invisible supply chain means asking who holds the keys — and here the vendor holds the firmware, the pool, and the revenue ledger simultaneously. The footnotes concede the return figures are "estimates based on network conditions at publication," which is legal cover, not a promise.
Contrarian
The reflex is to call this an outright scam. I will not — correlation is not causation, and "too good to be true" is a prior, not a proof. There is a genuinely boring explanation that fits every data point: a marketing agency generated plausible-sounding specs from a template, attached a Hong Kong headquarters and US operations to appear legitimate, and never built anything. That is not fraud requiring a factory; it is fraud requiring a PDF. The burden of proof sits entirely with ASICID, and it has offered none.
But the more uncomfortable angle cuts the other way. If the specs were real, they would be an industry-level black swan — the kind of breakthrough that surfaces first in patents, foundry capacity bookings, and pool hashrate spikes, never in a paid placement with a "highest returns" disclaimer. You cannot compute ROI on a miner with no price, and a seller who hides the denominator is not selling hardware — they are selling a story about hardware.
Takeaway
Watch three signals over the next 30 days: a disclosed unit price, a live pool stat showing the claimed hashrate, or a named foundry partner. Zero of three, and this was never a product — just entropy in the order book, priced at $25,590 a month. The arbitrage window closes fast. So does the refund window.
