SatStack
SatStack Research · 2026-10-20
The cheapest way to own bitcoin through a public equity right now is still a miner — and it is a fair price, not a surprise. Bitcoin closed at $83,186 on October 20. Through MARA, the market sells you a dollar of spot at roughly par: the stock prices at $84,661 per bitcoin it holds, about 1.0× spot. Through Canaan, you pay about $101,178 — 1.2× — and get the best-covered balance sheet in the sector, net cash, and the cheapest hash per dollar of equity. The names the market has re-rated as AI infrastructure — Hut 8, Cipher, Core Scientific, Iris, TeraWulf — fail every mining-value test we run, because you are no longer buying a miner. You are buying a power shell. Capital that wants bitcoin goes to the vaults. Capital that wants AI compute should say so out loud and stop pretending it is buying a miner.
| Ticker | BTC held | Hash (EH/s) | BTC/yr (ann.) | Total debt | Cash | Net debt | Mkt cap |
|---|---|---|---|---|---|---|---|
| ARBK | 2 | 2.4 | 150 | $0.04B | $0.01B | $0.03B | $0.03B |
| BTDR | 61 | 79.9 | 15,720 | $1.84B | $0.46B | $1.38B | $2.22B |
| CAN | 1,868 | 10.1 | 528 | $0.06B | $0.07B | −$0.01B | $0.19B |
| CANG | 1,056 | 27.6 | 2,624 | $0.05B | $0.01B | $0.04B | $0.60B |
| CIFR | 646 | 11.6 | 1,388 | $5.54B | $0.83B | $4.71B | $3.82B |
| CLSK | 13,703 | 38.3 | 7,116 | $1.80B | $0.20B | $1.60B | $2.27B |
| CORZ | 848 | 15.7 | 1,192 | $4.30B | $1.77B | $2.53B | $4.57B |
| DMG | 389 | 1.5 | 248 | $0.01B | $0.00B | $0.01B | $0.06B |
| HIVE | 190 | 25.2 | 4,016 | $0.27B | $0.21B | $0.06B | $0.57B |
| HUT | 17,316 | 28.1 | 3,728 | $7.74B | $0.23B | $7.50B | $8.82B |
| IREN | 0 | 36.5 | 6,075 | $7.59B | $5.90B | $1.70B | $10.93B |
| MARA | 35,577 | 70.3 | 9,688 | $2.42B | $0.42B | $2.00B | $3.01B |
| RIOT | 11,380 | 44.4 | 6,348 | $0.84B | $0.47B | $0.37B | $5.39B |
| WULF | 2 | 5.6 | 716 | $5.21B | $2.62B | $2.59B | $6.25B |
Source: SatStack Bitcoin mining dataset — company filings & IR, 2026-10-20. Market caps re-derived at current prices.
Four dimensions in one breath
We rank every mining-active public miner on four numbers. Two measure the balance sheet, two measure the operating asset.
- BTC liquidity — the BTC stack as a multiple of total debt. How much of the company's borrowings the bitcoin already covers.
- Production per dollar — annualized BTC mined per $1B of market cap. How much output your equity dollar buys.
- Implied BTC price — market cap divided by BTC held. The price per bitcoin you effectively pay through the stock, if the company were its vault.
- Hash price — market cap per EH/s. What a unit of hashrate costs you.
One metric anchors the whole report. The implied BTC price tells you, in dollars, what the market charges you per bitcoin through the ticker — and this report's anchor number is $84,661, the price the market charges through MARA when spot was $83,186. A dollar of spot, sold at $1.02.
Invert the ratio and it becomes a verdict. If each company were worth only its bitcoin, MARA's stock prices the coin at $84,661 — spot, sane. RIOT's prices it at $473,814 — 5.7× spot, a bet on the business, not the coin. WULF's prices it at $3.1B — not a bet on bitcoin at all. The spread is the sector sorting itself: vaults at ~1×, operators at 2–7×, and the AI-repriced names at multiples so large the market has stopped pricing them as miners altogether. Lower is cheaper only among the names the market still prices as bitcoin holders — which is exactly the line this framework draws.
| Dimension | #1 | #2 | #3 |
|---|---|---|---|
| BTC value ÷ total debt (higher = more cover) | CAN 2.64× | DMG 2.32× | CANG 1.78× |
| Production per $1B mcap (BTC/yr, higher = cheaper output) | BTDR 7,081 | HIVE 7,058 | ARBK 6,000 |
| Implied $/BTC via stock (lower = cheaper) | MARA $84,661 | CAN $101,178 | DMG $164,524 |
| $ per EH/s (lower = cheaper) | ARBK $10,416,667 | CAN $18,805,970 | CANG $21,936,186 |
Source: SatStack Bitcoin mining dataset — company filings & IR, 2026-10-20. IREN holds 0 BTC and is excluded from the implied-price ranking.
The vaults: MARA, CAN, CLSK
MARA is the purest bitcoin exposure a public market sells. Its stack of 35,577 BTC covers 1.22× of debt; it prices at 1.0× spot on our implied-BTC measure; and it has not left the top three of that ranking in six consecutive months of testing. The reveal: you are not buying a miner. You are buying spot bitcoin with a mining division attached, at par. This is the BTC-instead-of-a-fund pick — the closest a listed equity comes to holding the coin without paying a premium for a wrapper.
CANAAN (CAN) is the quiet all-rounder. It wins or places in all four dimensions: 2.64× BTC/debt — the best cover in the sector — an implied ~1.2× spot, the second-cheapest hash at ~$19M per EH/s, and a net cash position. The reveal: you are not buying a hardware maker. You are buying a covered vault that happens to mine. The price of admission is size: ~$189M market cap, and a hardware business that comes along for the ride. Small caps are how you pay a small-cap price for a big-cap balance sheet.
CLEANSPARK (CLSK) is the third leg of the group: 13,703 BTC held, third-cheapest bitcoin via stock in every single month of our backtest — but only 0.63× debt cover and the priciest hash of the three, at ~$59M per EH/s. It is the leveraged optionality on this trade: the same vault thesis, with less cover and more operating risk priced in.
Source: SatStack Bitcoin mining dataset — company filings & IR, 2026-10-20.
Source: SatStack Bitcoin mining dataset — company filings & IR, 2026-10-20.
The operators: HIVE, BTDR, CANG
These are the production-efficiency names — the cheapest output per dollar of equity, month after month. HIVE is the credible one: 7,058 BTC per $1B of market cap — second-cheapest production in the universe — cheap hash, and only ~$0.27B of debt. It is the small-cap to actually own on this dimension: minimal leverage, real output, no narrative premium. BITDEER (BTDR) produces even cheaper — 7,081 BTC per $1B, the best in the sector, on the largest fleet in the group at 79.9 EH/s — but it holds only 61 BTC, so it scores zero on the treasury dimensions. You are buying a foundry with a hashrate, not a vault. CANGO (CANG) quietly does everything right: 1.78× debt cover, 4,337 BTC per $1B, cheap hash, and a ~$0.60B cap. If CAN is the all-rounder in the small-cap bucket, CANG is the runner-up that nobody discusses.
The power shells: HUT, CIFR, CORZ, IREN, WULF
These five are the consistent cellar of every mining dimension — not by accident, but by design. The market is paying for a different business, and our ranking exists precisely to say so.
HUT 8 holds 17,316 BTC — more than anyone but MARA — and it is the best exhibit of why the headline number lies. Its stack covers only 0.19× of its $7.74B of debt, and the market prices the stock at ~$509,000 per bitcoin it holds: about six times spot. You are not buying a treasury. You are buying a leverage story whose bitcoin is collateral, not reserve. CIFR's bitcoin covers roughly 1% of its $5.54B of debt — 0.01× — and the stock implies ~$5.9M per BTC held. IREN holds zero bitcoin by construction; it sells everything it mines, so there is no treasury dimension to score. WULF holds 2 BTC and prices hash at $1.12B per EH/s — the most expensive hashrate in the entire universe. CORZ sits at 0.02× debt cover. On mining value, every one of these fails every test. On AI compute, they may be wonderful companies. They are not miners anymore, and the moment you accept that, the whole sector sorts itself.
Source: SatStack Bitcoin mining dataset — company filings & IR, 2026-10-20.
Six months of receipts
We ran this framework every month-end from April through September, using only the data available at the time. Bitcoin went $76.3k → $73.6k → $58.6k → $62.8k → $78.5k → $83.6k, June selloff included. Three things it would have caught, in real time.
| Month (BTC close) | 1. BTC value ÷ debt | 2. Production per $1B | 3. $/BTC via stock | 4. $ per EH/s |
|---|---|---|---|---|
| Apr 2026 ($76.3k) | HUT | HIVE | MARA | ARBK |
| May 2026 ($73.6k) | HUT | ARBK | MARA | ARBK |
| Jun 2026 ($58.6k) | CAN | ARBK | CAN | CANG |
| Jul 2026 ($62.8k) | CAN | ARBK | CAN | CAN |
| Aug 2026 ($78.5k) | CAN | ARBK | MARA | ARBK |
| Sep 2026 ($83.6k) | CAN | ARBK | MARA | ARBK |
Source: SatStack 6-month backtest, Apr–Sep 2026, point-in-time — top-1 ticker per dimension at each month-end.
HUT's leverage build, as it happened. In April and May, HUT led the debt-cover ranking at 3.10× and 2.98× — its debt was still ~$0.4B. By the time it levered up to ~$7.7B for the AI pivot, it had vanished from the ranking entirely. The framework flags a balance-sheet transformation the quarter the numbers change — not the quarter the rating agencies notice.
The stability of the vault trade. MARA, CAN and CLSK were the three cheapest ways to buy bitcoin through a stock in all six months. Six months, six appearances each. That is not luck; that is a structural price. The claim that MARA trades as a pure BTC proxy holds retroactively, which is exactly what a conviction should do.
CAN's resilience through the drawdown. In June, with bitcoin at ~$58.6k, CAN was the top name on both debt cover and implied BTC price. The all-rounder was still the cheapest, best-covered bitcoin on offer at the exact moment the cycle broke. Cheap hash, net cash and a covered balance sheet are survival features; the June test told you which names had them.
The framework also flagged its own noise. ARBK topped the production-per-dollar ranking in five of six months and never left the top three on cheapest hash — because it is a ~$23M micro-cap whose "wins" are denominator artifacts. HIVE is the credible small-cap on production efficiency. ARBK is a rounding error wearing a leaderboard position.
Where capital goes
If you want bitcoin through a public equity: MARA at 1.0× spot is the cleanest expression in the market. It is not a gift — it is a fair price on the most stable signal we track. CAN is the all-rounder: best debt cover, net cash, cheapest hash, ~1.2× spot — for anyone willing to take small-cap size and a hardware business in exchange for the best-covered balance sheet in the sector. CLSK is the leverage on that trade — the same vault thesis with thinner cover and more operating risk, for capital that wants the beta.
If you want production efficiency rather than a treasury: HIVE is the credible small-cap — cheap output, minimal leverage, no narrative premium. BTDR gives you the cheapest production and the biggest fleet, at the cost of holding almost no bitcoin.
If you want AI compute: buy the AI compute companies, and price them as such. That is a legitimate and possibly excellent thesis. It is not this report's subject, and it is not a mining allocation.
What the framework flags
These are the names the four dimensions keep flagging, stated as the framework sees them:
| Ticker | What the framework flags |
|---|---|
| HUT | Debt cover fell from 3.10× (April) to 0.19× (September) as debt grew to $7.74B. The stock prices at ~$509,587 per BTC held — about six times spot. |
| CIFR | BTC covers 0.01× of $5.54B debt; the stock implies ~$5.9M per BTC. |
| IREN | Holds 0 BTC — nothing to score on three of four dimensions. |
| WULF | The most expensive hashrate in the universe at $1,116,964,286 per EH/s. |
| ARBK | ~$23M micro-cap on FY2025 annuals; its top rankings are denominator artifacts, not a verdict. |
| CORZ | 0.02× debt cover on the mining dimensions. |
Source: SatStack Bitcoin mining dataset & 6-month backtest, Apr–Sep 2026.
Conditions, stated plainly
MARA's 1.0× pricing is a fair price, not an opportunity. The edge is that the market keeps pricing it as an operating company when it is increasingly a vault — and if BTC holds above roughly $70k, the covered names compound and the leverage stories get cheaper to attack. If BTC revisits the ~$58.6k June level, the stories get tested and the balance sheets — CAN, MARA, HIVE — buy the assets. Unless BTC recovers materially, every miner's equity loses twice: once on the coin, once on the re-rating. That is the condition on all of the above, and we do not pretend otherwise.
Method, sourcing, and the fine print
Four standardized dimensions, recomputed monthly from treasury, hashrate, production, debt and cash, with market caps re-derived at current prices rather than stale period-end figures. We cross-check every figure against primary filings — SEC 10-Q/10-K/20-F and company IR pages — and the exercise is not ceremonial. It caught a CIFR debt figure that had been double-counted and corrected it to $5.54B; the verdict did not change — the corrected number still leaves ~1% BTC cover. It also corrected ADR-based market caps for Cango and Canaan, which off-the-shelf pricing gets wrong by an order of magnitude. Sources are cited per row in the underlying dataset.
This report is updated monthly, is not investment advice, and our lawyers made us say so. Figures carry period-basis caveats — monthly versus quarterly reporting, and a mining-cost basis that differs per company and is intentionally not ranked here.
— The SatStack Research Desk