The Meter You Cannot Read: Block Proto, Sazmining, and the Arithmetic of Retail Mining

The number that stopped me was not $4,187. It was $153.

I have a habit, formed during a bad year, of reading hardware listings the way I once read Solidity: slowly, line by line, looking for the thing the author hoped I would scroll past. On a Tuesday afternoon I was reading Sazmining's page for Block's Proto immersion board, and the specification table was clean and confident. Two hundred and four terahashes per second. Roughly fifteen joules per terahash. A board for $4,187, or $1,397 down and eighteen payments of $155. And then, underneath, in the smaller type where operational truth usually lives: hosting and maintenance, $153 per month. No end date. No index. No ceiling.

$153 is not a number. It is a sentence that never finishes.

I have spent twenty-nine years watching this industry hand people the same object in different wrapping: a claim on the future, sold at a discount to the belief required to hold it. What Sazmining is offering is not new in kind. It is new in arithmetic, and the arithmetic is what I want to lay out on the table, because once you see it you cannot unsee it, and because the people this product is aimed at are precisely the people least equipped to run the numbers themselves.

The context, plainly

Block — the payments company formerly known as Square — has spent years trying to build bitcoin mining hardware under the Proto name. The product is real and technically interesting: immersion-cooled boards, a modular architecture in which three boards compose a full module, a design philosophy that trades raw efficiency leadership for packaging and deployment simplicity. Proto has also been delayed, repeatedly, and its place inside Block's hardware ambitions has shifted more than once. That history matters, not as gossip, but as a signal about how much institutional patience sits behind a product you are being asked to hold for three years.

Sazmining is a United States-based hosted mining provider. Its business is the retailization of industrial hashrate: you buy hardware, it lives in their facility, they handle power and operations, you receive bitcoin. The new offer is a disaggregation of that model down to a single board, with consumer-style financing attached.

The timing is not accidental. We are in the compressed middle of a post-halving cycle. Network difficulty has been printing records. Hashprice — the revenue a unit of hashrate earns per day — has been squeezed to the point where small operators are switching off machines that were profitable eighteen months ago. When margins compress at the industrial layer, hardware vendors face inventory they cannot move at industrial prices. The classic response is to lower the entry ticket and widen the buyer pool. That is what this is. It is not a technology announcement wearing a financial costume; it is a distribution announcement wearing a technical one.

And note who the buyer is meant to be. Not a professional miner, who negotiates power contracts and buys in container lots and would never accept a quoted rate. The intended customer is an allocator, a retail investor, someone who wants exposure to mining economics without operating anything — a person whose confidence in their own risk literacy is, on average, considerably higher than their actual risk literacy. That gap is the product's margin.

The arithmetic nobody puts in the headline

Start with eighteen months, the length of the financing plan.

You pay $1,397 at signing. You pay $155 for eighteen months, which is $2,790. You pay $153 in hosting for each of those eighteen months, which is $2,754. Total cash out of your pocket in the first year and a half: $6,941.

That is 165.8 percent of the sticker price of the hardware. The "zero percent interest" claim is true in the accounting sense and false in the economic sense; the cost of capital has simply been moved into the hosting line and the hardware premium. My own rough estimate of the effective financing cost, based on the structure alone, lands somewhere around four to six percent annualized. Nobody is being robbed. But nobody is being given a discount either, and the word "zero" is doing a great deal of emotional work in that sentence.

Hold the board for three years and the number becomes $9,695 — 2.3 times the sticker price — because the $153 does not stop when the loan stops.

Here is the line that matters more than any other: $153 divided by 204 terahashes is seventy-five cents per terahash per month, fixed. That fee does not move with the bitcoin price. It does not move with network difficulty. It does not move with your actual delivered hashrate, whatever that turns out to be. It is the one number in the entire arrangement that is guaranteed, and it is guaranteed against you.

So let us be precise about what is being sold. It is not a miner. It is a call option on hashrate revenue, with a monthly rent attached, and the rent is the only part of the contract that behaves like a certainty.

What the economics actually look like

At a hashprice in the range of five to seven cents per terahash per day — a reasonable industry figure for the current regime — 204 terahashes grosses somewhere between $10 and $14 a day, or roughly $300 to $420 a month. The $153 hosting fee consumes between 36 and 51 percent of that gross before a single joule of depreciation is counted.

Net of hosting, you are looking at $150 to $270 a month. Against a $4,187 capital base, that is a static payback of fifteen to twenty-eight months. On paper, that is not absurd. On paper, a lot of things are not absurd.

But the static model is a lie of omission, and it omits three things that always arrive. Difficulty rises — historically three to five percent a month, compounding. The next halving, expected around 2028, cuts the block subsidy in half. And hardware depreciates, fails, and drifts out of the competitive band. When I run those three erosions against the static payback, the honest range moves to twenty-one to forty-eight months. By the upper end of that range, the board is at the end of its useful life and the buyer has spent three and a half years paying a fixed fee to watch a floating number decline.

One more comparison the listing declines to make: $4,187 divided by 204 terahashes is $20.5 per terahash of capital cost. Whether that is fair depends entirely on where the market is clearing new-generation machines per terahash this quarter, and the page does not say. That omission is the most consequential gap in the entire disclosure, because the promise of "lowering the barrier" collapses completely if the barrier was lowered by raising the unit price.

This is the structural heart of the product: the buyer holds all of the downside, and the host holds all of the certainty. The fee claims come first. The returns come last. In mining finance this is called a fee-first structure, and it is the single most under-priced risk in every retail hashrate product I have examined.

The efficiency number is an entry ticket, not a moat

Fifteen joules per terahash is fine. It is not remarkable. The current generation of leading machines sits between roughly 13.5 and 17 joules per terahash, which puts the Proto board squarely in the middle of the pack. Presenting 15 J/TH without a comparative frame is a familiar move: give the reader a number that sounds technical and let them supply the conclusion. It is selective specification, and I have seen it in token whitepapers for a decade.

The modular architecture deserves a harder look too. Three boards compose one complete Proto immersion module. Selling a single board means selling a subsystem: the cooling loop, the power delivery, and the control plane all live in someone else's rack, under someone else's firmware, answering to someone else's uptime policy. That is not modular convenience. That is product fragmentation dressed as flexibility. You own a fragment of a machine and a whole set of obligations.

And notice who actually captures the benefit of immersion. Immersion cooling's real advantages are density, low power usage effectiveness, and quiet operation — all of which accrue to the facility operator, in the form of lower electricity and cooling cost per unit of hashrate. The buyer pays a flat $153. The host keeps the efficiency dividend. That is an asymmetric arrangement, and it is not hidden; it is simply not explained.

The thing nobody is auditing

Here is where I stop being an economist and go back to being what I actually am.

In 2018, while the industry was busy celebrating token launches, I spent six weeks reading forty thousand lines of Solidity in an Ethereum charity token. I found three reentrancy vulnerabilities that could have drained roughly $2.5 million of user funds. The lesson I took from that winter was not "code is dangerous." It was narrower and more useful: risk accumulates where nobody is paid to look. The vulnerabilities were not clever. They were unexamined.

Apply that lens here and the picture sharpens. A retail buyer of this board cannot verify the delivered hashrate. Cannot verify uptime. Cannot verify that the electricity cost behind the "quoted service rate" is what it claims to be. The word quoted is doing enormous work in that phrase — it means the price is set by one party, disclosed by one party, and audited by no one. There is no third-party hashrate attestation in the material, no pool-side direct connection described, no published service-level agreement.

This is the largest unaudited custody arrangement in the asset class, and almost nobody is talking about it. We spent 2022 and 2023 building proof-of-reserves for exchanges, teaching the market to demand cryptographic attestation of custody. We did not extend that instinct to hardware. A miner who "owns" a Proto board owns a serial number and a hope. Ownership and possession have been severed, and the severance is not recorded anywhere a buyer can check.

In my research group's 2026 review of AI-crypto integrations, we found that seventy percent lacked transparent ownership models — systems that claimed to be trustless while quietly concentrating control in an unexamined operator. The disease is not specific to AI. It is what happens whenever convenience is allowed to substitute for verifiability. Trust is not a transaction; it is a resonance, and resonance requires a signal you can actually receive. Right now the buyer is receiving a marketing page.

The tail: negative equity and the missing exit

There is a scenario the sales page does not address and that every buyer should write down by hand.

Bitcoin falls. Difficulty keeps climbing anyway, because difficulty lags price and the network is stubborn. Gross revenue drops below $153 a month. The buyer is now paying to keep a losing machine alive, and the installment obligation continues regardless. Meanwhile the resale value of a single Proto board — a board with no established secondary market, no mature price discovery, no liquid bid — falls below the outstanding balance. That is negative equity on a depreciating industrial asset, held by a retail participant with no hedging tools and no way to stop paying.

Compare this to an Antminer. Love it or hate it, an S21 has a global second-hand market, a rough spot price, and a phone full of buyers. A Proto board has none of that yet. The absence of an exit is not a footnote; it is the product's defining feature. Buy in, lock in. The installment schedule and the fact that the hardware is already racked in someone else's facility together create exit friction that looks, from the outside, indistinguishable from intentional retention design.

And beneath all of it sits a legal question the industry keeps declining to answer. Run the Howey factors: money invested, yes; a common enterprise, arguably, since the host operates the hashrate and buyers share pool outcomes; expectation of profit, obviously; and profits derived from the efforts of others — which is precisely what full-service hosting is. The fourth prong is doing a great deal of quiet work. Financing plus full operational control is not a hardware sale with a service attached; it is closer to an investment contract with a machine as collateral. That the marketing leans so hard on the phrase "hardware purchase" is not an accident of drafting. It is a legal posture.

The contrarian cut

The easy reading is that Sazmining is selling hope to people who cannot do the math. I do not think that is right, and I think the easy reading lets everyone off the hook too cheaply.

The harder reading is that they are selling exactly what they advertised: a board, a fee, a service. The disappointment, when it comes, will be a failure of imagination on the buyer's side rather than a breach on the seller's. And that matters, because it means the ethical line is not drawn where we usually draw it. It is not drawn at the price. It is drawn at the exit.

A bad deal you can leave is a market. A bad deal you cannot leave is a trap. The difference between the two is not the interest rate or the joules per terahash; it is whether there exists a door, and whether anyone told you where it is. On that test, this product is not yet a trap, but it is standing very close to the line, because no door has been described.

I will also say the unpopular thing. Retail mining is not irrational if what you are buying is proximity — a way to touch the thing you believe in, to feel the network as a physical hum rather than an abstraction on a screen. That desire is legitimate. To own nothing is to feel everything, deeply, and there is a real human hunger behind it that no spreadsheet will ever satisfy. But a feeling should be priced as a feeling. When it is priced as an investment, the buyer eventually discovers the difference, and the discovery is expensive — and it costs the industry something too, because every disappointed retail miner becomes a voice arguing that this whole thing was always a casino.

Takeaway

Over the next eighteen months, mining will keep sliding toward finance: hashrate-backed instruments, equipment leases wrapped as yield products, retail tranches of industrial capacity. The regulatory frame will follow the money from commodity law toward securities law, and the compliance cost of the entire sector will step up when it does. The winners in that transition will not be the ones selling the most hashrate. They will be the ones selling verifiability — independent metering, published service levels, audited uptime, a pool connection the buyer can check without asking permission.

So the question I would put to Block and Sazmining is not about joules or interest rates. It is simpler. If we can prove a reserve on a chain, why can we not prove a machine in a rack? The soul does not mint; it manifests. And what manifests here, so far, is a page of small type and a number that never stops.