One hundred fifty-three dollars a month. Divide by 204 terahashes. You get $0.75 per TH per month.
Write that number down. Tape it to the monitor. Because it is the only number in this entire offering that does not move.
Bitcoin can print $150,000 or $45,000. Network difficulty can add four percent next epoch or eleven. Your board can run at 99% uptime or sit dark for a week behind a firmware push nobody told you about. The $153 does not care. It clears on the first of the month, in full, indefinitely, with no terminal date printed anywhere in the disclosure.
Sazmining is now selling Block's Proto immersion hashrate boards in single units — $4,187 sticker, $1,397 down, eighteen payments of $155, hosting and maintenance at $153 monthly. The pitch is "0% interest" and "start mining for under fourteen hundred dollars." Both statements are literally true. Both are structurally misleading in the same way a lottery ticket is a low-cost entertainment product.
I have audited a lot of things that looked like this. In 2017 I tore apart proxy contract logic on three mid-tier ICOs before deploying a single dollar, and I found a reentrancy hole that let me exit 48 hours before the exploit. The lesson was never "contracts are dangerous." The lesson was that the price you are quoted and the cash you will actually pay are two different documents, and only one of them is audited by anyone.
Here, nobody audited anything. Every figure in this offering comes from one side of the table.
The structure, stripped of the pitch
Block Proto is a Bitcoin ASIC built around immersion cooling — hashboards submerged in dielectric fluid instead of air, which lets a rack run denser, cooler, quieter. Spec sheet says 204 TH/s at roughly 15 joules per terahash. A complete Proto Immersion Module is three boards. Sazmining is selling them one at a time, financing the balance, and hosting the hardware in its own racks.
That is the entire product. Hardware, credit, and a rack slot, bundled into a single checkout flow.
Some context on the machine itself, because the marketing will not hand it to you. Fifteen joules per terahash in 2025 is not a flex. It is the entry line. Bitmain's S21 XP sits near 13.5 J/TH; the S21 Pro lands right around 15. The industry's new-machine band has compressed into 13–17 J/TH and the leaderboard is decided in decimal places. When a spec sheet gives you an absolute number with no competitive frame, that is a choice, not an omission. Selective specification is the oldest trick in hardware marketing: make the figure look technical enough that nobody asks what it beats.
The three-board architecture matters more than the efficiency number. A single Proto module needs three boards to be a complete system — power delivery, control board, thermal envelope all designed around the full assembly. Buying one board means buying a subsystem, and a subsystem's behavior depends entirely on the rack it lives in. Sazmining's rack. Sazmining's power. Sazmining's firmware cadence. Sazmining's pool routing.
That is packaged as modularity. It reads to me like a supply chain telling you something. When a manufacturer with a history of delayed Proto shipments starts moving a $4,187 product in $1,397 slices through a third-party host, the most economical explanation is that the full-module price point was not clearing demand. Splitting the unit is a pricing-curve repair, not a product innovation. I watched this exact maneuver in DeFi. When emissions stop attracting deposits at the headline APY, the protocol does not improve the yield — it re-denominates it. Smaller tickets, longer lockups, more fee layers. Same capital, more friction, better optics.
Note also who is absent from the disclosure. No third-party hashrate verification. No SLA document. No uptime history. No rate-lock clause. No stated path for what happens when a buyer wants out. Block is a public company, which makes the omission louder rather than quieter — a listed issuer knows precisely what a service-level commitment looks like, and this offering does not contain one. Sazmining is a small operator by comparison, and the hardware sits in its building. That is not a footnote. That is the counterparty.
The cash flow nobody builds
Let me build it, because everything downstream depends on this arithmetic and almost nobody does it before signing.
Sticker: $4,187. Down payment: $1,397 at T+0. Eighteen monthly installments of $155, totaling $2,790. Hosting and maintenance: $153 per month, no stated termination date, described as electricity costs calculated at quoted service rates.
Eighteen months of cash out: $1,397 plus $2,790 plus $2,754 equals $6,941.
That is 165.8% of the sticker price. The 0% interest claim holds in accounting and fails in opportunity cost. Discount that installment stream at anything resembling a real cost of capital and the implied financing rate lands somewhere in the 4–6% annualized band. Not usury. Not free either.
Hold for three years and the figure becomes $9,695. Two-point-three times the sticker. Now the per-unit view the brochure will never print:
Purchase: $4,187 divided by 204 TH/s is $20.52 per TH. Hosting: $153 divided by 204 is $0.75 per TH per month, or $9 per TH per year. Three years of hosting: $27 per TH, which is 132% of what you paid to own the hardware in the first place.
Read that last line twice. Over a normal ASIC service life, the rent exceeds the purchase price, per terahash. All-in over 36 months: $47.50 per TH. You are not buying a machine. You are buying a call option on hashrate with a monthly premium attached, and the premium compounds while the underlying decays.
What the hashrate actually pays
Using standard industry ranges — call it $0.05 to $0.07 per TH per day at current difficulty with BTC near six figures — 204 TH/s grosses $10 to $14 daily, or roughly $310 to $428 monthly.
Hosting takes $153 of that. That is 36% to 49% of gross, before difficulty drift, before any downtime, before pool fees.
Net lands at $157 to $275 per month under static assumptions. Payback on the $4,187 sticker: 15 to 27 months. That is the number in the pitch deck.
But you did not pay $4,187. You paid $6,941 across eighteen months, and you are still paying $153 a month after that. Measured against actual cash out, payback stretches to 25 to 44 months — and that is still the static case.
Static is a lie in mining. Three things erode it, all one-directional. Network difficulty has historically compounded 3–5% monthly. The next halving, roughly 2028, halves the block subsidy. And hardware depreciates while failure rates climb in the back half of service life. Apply difficulty growth and the halving to the model and most static payback estimates extend 40% to 70%. Twenty-one to forty-eight months. By the far end of that window the board is a depreciating appliance carrying a fixed monthly bill.
The death line
This is the single most useful number in this article. The hosting fee is $153 a month, or $5.10 a day. On 204 TH/s, the board goes cash-negative on operating costs alone when gross revenue falls below $0.025 per TH per day.
Current gross is roughly $0.05 to $0.07. The death line is a 58% to 64% decline in per-TH revenue. A halving alone delivers a 50% cut at flat price. Add four months of difficulty growth on the other side and you are through the floor. At a flat BTC price, the 2028 halving puts a single Proto board underwater on hosting costs alone — not unprofitable, underwater. Still paying rent on a machine that cannot cover its own electricity.
That is the structure in one sentence. Fixed cost. Floating revenue. A host that captures the difference in both directions.
Who captures the immersion dividend
Here is the part that took me longest to see clearly, and it comes out of the cooling design itself.
Immersion's real economics are density, low PUE, and low noise. Those are rack-operator benefits. Denser deployment means fewer square meters per megawatt. Low PUE means less overhead power per unit of compute. Those savings accrue to whoever operates the facility. The buyer pays $153 flat regardless. The host harvests the efficiency dividend; the buyer pays a fixed rate. That is not a partnership. That is a spread, and you are on the wrong side of it.
The engineering is genuinely good. The distribution of the gain is genuinely bad. Both can be true, and in this deal they are.
The seniority problem
Look at the ordering of the cash flow. The host's $153 is contractual, senior, and payable regardless of what the board earns. The buyer's return is residual — whatever survives the fee, the difficulty adjustment, the downtime, and the pool's cut. In structured credit terms, the host holds the senior tranche and the retail buyer holds equity in a single-asset deal with no diversification and no tranche protection.
I have watched this exact ordering destroy people. In 2020 I ran a Python loop across Uniswap and SushiSwap pairs, rebalancing on gas and yield in real time, and it worked — 400% in six months — precisely because I understood emissions are a temporary subsidy and the only durable edge is execution speed. When the subsidy stopped, the farmers who had modeled the yield as a return got wiped. The ones who had modeled it as a decaying fee stream survived. Arbitrage is just patience wearing a speed suit. Hosting fees are the inverse: patience wearing a bill.
There is a second failure mode stacked on top, and it is the one that gets buried in the fine print. Eighteen months of installments against hardware that depreciates on a 36-to-48-month curve creates a window where the outstanding balance exceeds the resale value of the asset. If BTC corrects early, the buyer is servicing debt on a board worth less than the remaining obligation, with no early-payment incentive disclosed and no return path. That is negative equity in a consumer contract. Nobody calls it that in the marketing.
The exit that doesn't exist
Where do you sell a single Proto board?
Antminers have a deep second-hand market with quoted prices, liquidators, and brokers who will bid on a pallet. A single Proto board sitting in someone else's immersion rack has no comparable market, no price discovery, and no physical possession. Your asset is a line item in a hosting agreement with a termination clause you have not read because it was not shown to you.
Liquidity is the only truth that pays the bills — and here it is approximately zero.
That matters more than the efficiency gap, more than the difficulty curve, more than the halving. A depreciating, illiquid, fixed-cost asset is fine if you can hand it to the next buyer. This one you cannot hand to anyone. The friction is not accidental. An asset that is expensive to exit is an asset that stays on the books, and staying on the books is the entire business model of a host.
Now the argument against myself
The bear case above is easy. The interesting question is who is actually on the other side of this trade.
The buyer is almost certainly not a miner. A professional buys machines outright, negotiates power at four to six cents, owns the rack, and would never accept an unauditable fixed fee with a unilateral pricing clause. The target here is an allocator — someone with crypto exposure who wants mining economics without building a shed in Texas. A retail investor. Possibly an equity holder in Block who likes the story and wants a piece of the vertical.
That distinction reframes everything. Evaluate this as an investment and the expected-value gap is brutal: marketing leads with $1,397 while the eighteen-month commitment is $6,941 — a five-fold perception gap engineered into the pricing architecture. Evaluate it as a paid educational experience, where you learn what difficulty feels like by watching your own hashrate decay in real time, and the pricing is defensible. The product is not fraudulent. It is behavioral pricing design, and behavioral pricing design works because the buyer never builds the spreadsheet.
Then there is the regulatory layer, the most under-discussed piece and potentially the most consequential. Run the Howey factors. Money invested: yes. Common enterprise: arguably, since the host operates the hashrate and buyers share pool outcomes. Expectation of profit: yes. Efforts of a third party: emphatically, because the host controls power, maintenance, firmware, and uptime. A financing-plus-full-custody bundle looks a great deal like an investment contract, and the 0% interest label looks like pricing designed to sit outside consumer-credit rules by burying the cost inside the hardware number. The $20.52 per TH entry price is where that buried cost lives.
I have seen what happens when the counterparty layer gets ignored. In 2022 I shorted LUNA into the collapse at 5x margin and made $90,000 in 72 hours because the peg math was observably broken. The lesson I kept was not about the short. It was that I was routing through perpetual DEXs and centralized venues while the entire system was failing, and a winning trade can still be lost to a counterparty that stops answering. Your hardware living in somebody else's rack is an unsecured claim on a small operator's balance sheet, and nothing in this offering addresses that operator's solvency, insurance, or continuity.
So my read cuts two ways. The bearish version: this is a risk-transfer product, with difficulty, price, depreciation, and exit liquidity pushed wholesale onto the least sophisticated participant while the host locks a fixed claim on top. The bullish version: risk-transfer is how markets mature, and there is a real second-order trade behind this one. If retail mining financialization scales — and the equipment-plus-financing-plus-service structure maps almost perfectly onto equipment leasing and then onto ABS — verifiability becomes a product. Third-party hashrate attestation. SLA audits. Host credit ratings. Nobody is selling those yet. That is a 12-to-24-month window, and it is the trade I would actually put on.
The third trade, the one I like most: if BTC draws down and retail buyers capitulate into a market with no bids for single Proto boards, you get distressed hardware at a genuine discount. That only works if you can host it yourself. Which tells you exactly where the margin sits in this business and who is holding it.
What to watch, and where I draw the line
Three numbers, and none of them is the sticker price.
First, the hosting rate page. Any adjustment above 5% confirms the fixed cost is not fixed, which converts the structure from a lease into an open-ended liability. Second, whether an SLA with a published uptime figure and a compensation clause ever appears. Absence across two quarters is the tell that opacity is the product, not an oversight. Third, the per-TH market price for new machines. At $20.52 per TH to enter plus $9 per TH per year in rent, the buyer needs a durable efficiency advantage to justify the premium, and 15 J/TH does not provide one.
My line in the sand is $0.025 per TH per day. Above it, the board services its own rent and the structure is merely expensive. Below it, every additional month is a cash call on a decaying asset with no exit.
Survival isn't about conviction; it's about position sizing. If someone insists on doing this, the only defensible size is capital they have already written to zero — not because the hardware fails, but because the exit does.
Hedge the ego, not just the portfolio. The ego wants the $1,397 entry to be the price. The portfolio knows the price is $9,695 and a rack you will never inspect.
The chart is a map; the trader is the terrain. This chart shows a fourteen-hundred-dollar entry and a ninety-seven-hundred-dollar reality, and the terrain is a building you will never visit, run by a company whose solvency you cannot audit, charging a fee you cannot negotiate, on a machine you cannot sell.
The question worth sitting with is not whether retail mining democratization is real. It is who gets paid when the hashrate stops covering the rent — and whether the answer is different this cycle than it was in 2018, when the same arithmetic broke a different cohort of buyers who also believed the entry price was the whole price.