h hoge.gg
Subscribe
BTC$67,432.18+2.34%ETH$3,521.44+1.08%SOL$178.62-0.62%BNB$612.30+0.41%XRP$0.6234-0.18%ADA$0.4521+3.12%DOGE$0.1623+1.86%AVAX$38.71-1.24%LINK$17.84+0.92%HOGE$0.00004120+4.21%
BTC$67,432.18+2.34%ETH$3,521.44+1.08%SOL$178.62-0.62%BNB$612.30+0.41%XRP$0.6234-0.18%ADA$0.4521+3.12%DOGE$0.1623+1.86%AVAX$38.71-1.24%LINK$17.84+0.92%HOGE$0.00004120+4.21%
● Mining & Staking

Bitcoin Mining Margins: The $57,000 Stress Test

Bitcoin fell below $57,000 for the first time in two years, erasing the Q3 profits miners had just reported. Here is what a live margin stress test looks like, and which rigs survive it.

On Friday, Bitcoin fell below $57,000 for the first time in more than two years, and roughly $2.4 billion in leveraged crypto positions were liquidated in a single day, according to CoinDesk. It did not fall alone. Nvidia, copper, Asian equities and even gold slid in the same week as investors cut risk and rotated into Treasuries, with a more hawkish Federal Reserve forcing a repricing of every long-duration bet. Bitcoin went out with the growth trade, and the companies that turn electricity into Bitcoin went out with it.

The timing is cruel for miners. Two weeks earlier, the two largest listed operators had reported third-quarter numbers that read, on the surface, like a recovery. MARA posted record net income. Riot unveiled the largest lease in its history. But those results were marked to a 30 September Bitcoin price of $83,554, the close of the best quarter since 2017. The coin behind them is now worth roughly a third less. A mining margin is not a line in a press release; it is a live number that re-prices every block, and today it is moving against almost everyone who still runs a machine.

This is a field report on that live number: what a mining margin actually is, why the Q3 profit headlines lied about it, and what happens to the economics of hashing when the price of the product drops to $57,000 while the network prepares to make mining harder. Call it a stress test. The grid of ASIC machines humming in Texas and Abu Dhabi is being run through one in real time.

A Margin Is Three Moving Parts, and Two Just Broke

Strip away the accounting and a Bitcoin miner’s margin rests on three variables: the price of Bitcoin, the efficiency of the hardware (how many joules it burns per terahash), and the network difficulty (how much hashing the rest of the world is doing). Revenue is set by price and the miner’s share of total hashrate. Cost is set by efficiency and the price of power. The gap between them is the margin.

What makes mining unlike almost any other commodity business is the third variable. Difficulty is a protocol-level governor that re-targets roughly every two weeks to keep blocks arriving about every ten minutes. When miners switch on, difficulty rises and everyone’s share of the fixed 3.125-BTC block reward shrinks. When they switch off, difficulty falls and the survivors earn more per machine. No oil driller or gold producer operates inside a feedback loop that automatically competes its own margin back toward the marginal producer’s cost. Bitcoin miners do, on a two-week clock.

In November 2026 two of those three variables turned hostile at once. Price collapsed, dragging revenue down across the entire fleet. And difficulty, after a brief relief cut, is set to jump again, which will pull revenue down a second time. Efficiency, the one lever a miner actually controls, cannot move fast enough to offset both. That is why the margin broke, and why it is a useful moment to look at how the number is really built.

Hashprice: The Revenue Side Already Collapsed

The cleanest way to see a miner’s revenue is hashprice: the dollars a machine earns per unit of hashing per day, usually quoted as dollars per petahash per second per day ($/PH/day). It bundles the block subsidy, transaction fees, the Bitcoin price and the network difficulty into one figure, so a single number tells you whether the whole industry is making money.

Hashprice has been falling all autumn, and the November crash drove it to the floor. It sat near $40/PH/day on 1 October. By early November it had slipped to roughly $33, its weakest since August. As Bitcoin broke to fresh lows, hashprice fell again to between $28 and $30, the range tracked by the Hashrate Index through 20-21 November. For context, a year earlier hashprice was in the mid-$50s. The revenue a given machine produces has been cut roughly in half in twelve months, and by about a third since the start of October alone.

That is the number that matters more than any earnings release, because it is the market clearing price for one unit of mining. When hashprice sits at $28, a machine burning 20 joules per terahash in a 10-cent-per-kilowatt-hour facility is losing money on electricity before it pays for a single employee, lease payment or financing cost. The revenue side of the margin equation did not weaken in Q3; it caved in the weeks after Q3 closed, which is exactly the window the earnings reports could not capture.

Marked to a Price That No Longer Exists

Here is where the Q3 headlines mislead. On 10 November, MARA reported third-quarter net income of $632.1 million, or $1.28 per diluted share, up from $123.1 million a year earlier. That looks like a blowout. It is almost entirely an accounting artifact. Under the fair-value accounting rule adopted across the industry (ASU 2023-08), companies now mark their Bitcoin treasuries to market every quarter and run the change through the income statement. MARA’s quarter included an $885.3 million unrealized mark-to-market gain on the Bitcoin it holds, per its shareholder letter filed with the SEC. Strip that paper gain out and the operating business shrank: revenue was $164.2 million, down about 35% year over year.

The paper gain exists because Bitcoin rose during the third quarter. It is being erased because Bitcoin has fallen since. A net-income line that swings about $509 million year over year on the strength of a treasury revaluation tells you about the Bitcoin price, not about whether the company can mine a coin for less than it sells one. At $57,000, the same accounting that manufactured a record Q3 profit is now manufacturing a fourth-quarter loss, and nothing about the mining operation changed in between. The reported profit was marked to a price that no longer exists.

The honest number is cost to mine: what the miner spends to produce one Bitcoin, measured against what it can sell that Bitcoin for today. That figure does not care about the treasury. It is the only margin that survives a price crash with its meaning intact, and it is the one this article keeps returning to.

The HODL Miner Became a Seller

The most telling line in MARA’s quarter was not the profit. It was the selling. For years MARA ran a full-HODL strategy, keeping nearly every coin it mined and buying more, which is precisely why its income statement is now so sensitive to the Bitcoin price. In the third quarter it still held 34,012 BTC. But after the quarter closed, with the price sliding, the company sold roughly 13,597 BTC for about $1.1 billion, used the proceeds and other cash to cut debt by about $1.4 billion, and authorized a $500 million buyback of its own shares, according to its shareholder letter. The archetypal HODL miner started monetizing the treasury at exactly the moment the paper gain it had just booked began to reverse.

It is not alone. Riot sold more than 20,000 BTC over the course of 2026 for roughly $1.5 billion, repaying a Coinbase credit facility and releasing collateral in the process. Taken together, the two largest listed miners accounted for the majority of all Bitcoin disposed of by public miners this year, on the order of $4.5 billion in proceeds. That is the quiet confession behind the recovery headlines: when mining margins compress, the balance sheet, not the drill bit, becomes the source of cash, and the Bitcoin a miner once vowed never to sell becomes working capital.

Miner selling matters for the market too, because it is supply that arrives regardless of price. It is one reason the production-cost floor that stock-to-flow models kept missing has proven so soft this cycle: when the people who make new coins are forced sellers below cost, the theoretical floor stops acting like one.

Riot’s Honest Margin, and the $28 Billion Door

Riot reported its own quarter on 4 November, and it is the cleaner window into the real margin because the company publishes its cost to mine in two forms. Its direct cash cost to produce one Bitcoin, excluding depreciation, was $39,265, down more than 15% year over year as power-curtailment credits ($24.7 million in the quarter, about $15,067 per coin) pushed its net power cost to roughly 3.2 cents per kilowatt-hour. Its all-in cost, including the depreciation of its machines, was $78,805. Against a Q3 realized price of $70,855, that means Riot mined coins for 55% of their value on a cash basis but 111% of their value once you count the hardware wearing out. Cash-positive, GAAP-negative, in the same quarter. Reported net income was just $4.0 million.

Now re-run that at today’s price. A realized price near $57,000 does not touch Riot’s $39,265 cash cost, so the cash margin survives, but the all-in cost of $78,805 now sits about 38% above the price of the product. And Riot is one of the lowest-cost operators in the listed group. For the median public miner, whose power is not 3 cents and whose fleet is older, the all-in math at $57,000 is bleak.

Riot’s answer, like the rest of the sector’s, is to stop being only a miner. On the same day it reported, it disclosed a 15-year triple-net lease for the entire 910-megawatt Corsicana campus with a leading frontier AI lab, expected to generate roughly $28 billion in revenue over the initial term, per the company’s Q3 press release. Combined with its 191-megawatt Rockdale lease and an AMD deal, Riot has now contracted about 1,151 megawatts across three leases, roughly $38 billion of future data-center revenue and about $2.2 billion in average annual net operating income at full deployment. Chief executive Jason Les called Corsicana the largest lease in Riot’s history. The margin on a megawatt rented to an AI tenant is vastly higher, and vastly steadier, than the margin on a megawatt pointed at a SHA-256 hash. That door, however, is not open to everyone, a point this piece returns to below.

Q3 2026 scoreboardMARARiot
Reported net income$632.1M$4.0M
Main driver of the profit line$885.3M unrealized BTC gainmining plus first AI-lease revenue
Revenue$164.2M (-35% YoY)$181.4M (+1% YoY)
BTC produced2,2471,637
Cost to mine per BTC$39,674 (owned-energy basis)$39,265 cash / $78,805 all-in
BTC held at quarter-end34,0122,418
BTC sold (post-quarter / 2026)~13,597 for ~$1.1B~20,264 in 2026 for ~$1.5B
Power / AI contracts4.8 GW target, no anchor tenant yet1,151 MW leased, ~$38B contracted
Sources: MARA SEC shareholder letter; Riot Platforms Q3 2026 press release.

Cost to Mine Is Not One Number

Before stress-testing the margin, a warning about the cost figures themselves: there is no single, comparable cost to mine a Bitcoin. Each company defines it differently, so pulling one number from each press release and ranking them is a trap.

MARA quotes a cost per Bitcoin of $39,674 on an owned-energy basis, close to Riot’s $39,265. But MARA’s figure reflects power it buys at roughly four cents at its own sites and excludes much of the overhead that Riot’s broader cash cost captures, so the two are not like-for-like. Then there is depreciation, the single largest swing factor. It is a non-cash charge, and crucially it is not a cost you can avoid by curtailing: switch a machine off to dodge a high power hour and you still carry the capital cost of the hardware wearing out and falling behind. That is why the gap between a cash-profitable quarter and a GAAP loss is almost entirely the depreciation line, and why useful-life assumptions (some miners write ASICs off over two years, others four or more) quietly reshape who looks profitable.

For an industry-wide anchor, CoinShares pegged the listed miners’ weighted-average cash cost at roughly $75,500 per Bitcoin in its 2026 mining report, with all-in costs well above that. On that basis, the aggregate sector was already producing coins for more than they were worth before the November crash. The crash did not create the problem; it removed the last cushion.

The $57,000 Stress Test

Stack the cost measures against the price of the product and the stress test writes itself. Every honest measure of what it costs to make a Bitcoin now sits above, or uncomfortably near, the price you can sell one for. The only figure comfortably below spot is Riot’s narrow cash cost, and Riot is a best-in-class operator. For the sector’s weighted average, and for JPMorgan’s estimate of the network’s all-in production cost, the price is underwater.

The margin stack, per BTCFigureSource
Riot cash cost to mine (ex-depreciation), Q3$39,265Riot Q3
MARA cost to mine (owned-energy), Q3$39,674MARA SEC letter
Bitcoin spot, 21 November 2026~$57,000CoinDesk
Riot realized price, Q3$70,855Riot Q3
CoinShares listed-miner weighted-avg cash cost~$75,500CoinShares
Riot all-in cost incl. depreciation, Q3$78,805Riot Q3
JPMorgan network production cost (Nov 2026)~$86,000JPMorgan
At $57,000, spot sits below every cost measure in the stack except one operator’s narrowest cash figure.

The distinction that matters here is cash cost versus all-in cost. Three weeks ago, with Bitcoin near $76,000, the sector was mostly all-in underwater but cash-positive: miners were losing money on a full-accounting basis while still generating cash to keep the lights on. At $57,000, the picture darkens. The least efficient fleets, running older machines at ordinary grid power, are now approaching the point where the power bill alone exceeds the value of the coins the machines produce. That is the cash-cost line, and crossing it is what forces a miner to switch off rather than merely report a loss. The $57,000 stress test is a test of how much of the global fleet is near that line.

The Breakeven Treadmill: Which Rigs Survive at $28 Hashprice

Whether a specific machine survives comes down to two numbers: its efficiency, in joules per terahash, and the price it pays for power. At a given hashprice you can solve for the breakeven power price, the cost of electricity at which a rig earns exactly what it spends. The rough formula is breakeven dollars per kilowatt-hour equals hashprice divided by (24 times the machine’s joules per terahash). Below that power price the machine mines at a profit; above it, every block is a loss.

Run that at an early-October hashprice of $40 and again at today’s roughly $28.58, and the squeeze is stark. A top-tier hydro-cooled Antminer S23 at 9.5 J/TH needed sub-17.5-cent power to break even in October; today it needs sub-12.5-cent power. A legacy S19j Pro at 29.5 J/TH needed sub-5.7-cent power then; now it needs power under 4 cents, a level available only to miners sitting on stranded hydro, flared gas or behind-the-meter curtailment deals. The efficiency specs below come from manufacturer data sheets, including Bitmain’s published figures.

ASIC modelEfficiency (J/TH)Breakeven power at $40 hashpriceBreakeven power at $28.58 hashprice
Antminer S23 Hydro9.5~$0.175/kWh~$0.125/kWh
Antminer S21 XP13.5~$0.123/kWh~$0.088/kWh
Antminer S21 Pro15.0~$0.111/kWh~$0.079/kWh
WhatsMiner M60S18.5~$0.090/kWh~$0.064/kWh
Antminer S19 XP21.5~$0.078/kWh~$0.055/kWh
Antminer S19j Pro29.5~$0.057/kWh~$0.040/kWh
Illustrative breakeven power price by machine, computed as hashprice / (24 x J/TH). Excludes depreciation, overhead and financing.

Two lessons fall out of that table. First, the treadmill never stops: at $28 hashprice, even a current-generation S21 Pro needs power below 8 cents, which rules out most ordinary industrial grids. Second, the spread between the best and worst machines is enormous, so there is no single mining margin, only a distribution. The newest hydro rigs in cheap-power jurisdictions are still comfortably profitable on a cash basis at $57,000; the three-year-old air-cooled fleet in a 10-cent market is bleeding, and that fleet is the first to go dark.

The Network’s Counterpunch: Down 9 Percent, Then Up 8.5

When machines go dark, the network responds, and that response is the story of the past two weeks. On 14 November, at block 975,744 mined by Foundry USA, Bitcoin difficulty fell 9.13%, from 148.36 trillion to 134.82 trillion, the third-largest downward adjustment of 2026, reported Bitcoin.com. Blocks had slowed to about 10.7 minutes as higher-cost miners powered down, so the protocol made mining easier to bring block times back toward ten minutes. For the miners still running, that cut was a raise: a bigger slice of the same reward.

The relief is already ending. Hashrate has started climbing back as idled machines return and freshly shipped rigs energize, and the next adjustment, expected around 27 November, is projected to rise between 8.4% and 8.5%, per KuCoin and other trackers. That would claw back most of the 14 November cut and push difficulty back toward 146 trillion. So the sequence miners face is down 9, then up 8.5, inside a single fortnight, while the price of what they produce fell by double digits. The governor gives with one hand and takes with the other.

This is the mechanism JPMorgan analyst Nikolaos Panigirtzoglou has described in plain terms: when Bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower, as reported by TFTC. The 14 November cut was that quote playing out exactly. The looming 27 November increase is the counter-move: the moment relief appears, enough capacity returns to compete it away. A mining margin lives inside that loop, and the loop is why no recovery in this business is ever quite as durable as it looks.

Below Production Cost, for the Third Time This Year

Zoom out from any single company and the clearest summary of the stress test is JPMorgan’s. In a note dated 20 November, the bank put the network’s estimated production cost at about $86,000 per Bitcoin, more than $26,000 above the market price, as covered by Crypto Briefing. That estimate has drifted higher through the year as hashrate and difficulty climbed, even as the price fell, which is why the gap is now the widest it has been. It is the third stretch in 2026 that Bitcoin has traded below its own production cost.

A commodity trading below its cost of production is not a stable state, and the difficulty governor is the reason it tends to resolve. As the highest-cost miners capitulate and switch off, difficulty falls, the production cost falls with it, and the surviving miners’ margins recover until the price either rises to meet cost or cost drops to meet price. That is the self-correcting floor, and it is real, but it is slow and brutal: it works by bankrupting the marginal producer, not by sparing it. The question a $57,000 price poses is whether this cycle’s cull is a quick washout of a few weak operators or a deeper shakeout like 2018 and 2022, when names such as Giga Watt, Compute North and Core Scientific went through bankruptcy. The difference this time is the exit the earlier cohorts never had.

The AI Escape Hatch Is Not a Fire Exit for Everyone

The release valve on this shakeout is artificial intelligence. A Bitcoin miner is, underneath the ASICs, a company that has secured power, land and a data-center shell, which is exactly what AI developers are desperate to rent. The economics are not close. As Needham analyst John Todaro put it to CoinGeek, the revenue per megawatt and EBITDA margins are far higher for high-performance computing and AI colocation than for mining. Riot’s Corsicana lease, at roughly a 33% unlevered yield on cost with a near-99% operating margin, is a different business from hashing, bolted onto the same substations.

The sector has pivoted hard. Core Scientific, IREN, TeraWulf and Cipher have all signed or expanded multibillion-dollar AI colocation deals, and CoinShares research head James Butterfill has estimated that some listed miners could draw as much as 70% of their revenue from AI by the end of this year, up from roughly 30%, per The Block. The irony is sharp: the thing saving Bitcoin miners is the decision to do less Bitcoin mining. The same compute buildout is drawing scrutiny over who actually pays for all this capacity, a question explored in our look at how verifiable compute found its customer in 2026, and it was not AI.

But a fire exit only helps if you can reach it. Winning a hyperscale AI lease requires a large, contiguous, well-connected power site, an investment-grade-ish balance sheet to fund the fit-out, and the patience to wait out multi-year construction. MARA, the largest miner by hashrate, is targeting 4.8 gigawatts of power yet still has no anchor AI tenant at scale. Smaller miners with scattered sites and stretched balance sheets have nothing to pivot. For them the AI door is painted on the wall. The pivot is bifurcating the industry into AI landlords and the miners who will simply be competed out, and the November price is accelerating the split.

Why Miners Fell Harder Than Bitcoin

A mining margin is operating leverage on the Bitcoin price: mostly fixed costs (power contracts, leases, depreciation, debt service) against revenue that swings with the coin. When the price rises, margins expand faster than the price; when it falls, they contract faster. That is why mining equities behave like a leveraged bet on Bitcoin, which is itself a leveraged bet on global liquidity. In a week when the Fed signaled a more hawkish path and investors fled long-duration risk, Bitcoin fell with Nvidia and the AI trade, and the miners fell with Bitcoin, only more.

The macro backdrop is not a side note; it is the first of the three variables. The same rate expectations that are compressing crypto are the ones to watch into December, when the next policy decision could either deepen the drawdown or mark its bottom, a setup we unpack in crypto’s December countdown. For miners, a December that steadies the Bitcoin price would relieve the margin faster than any difficulty cut, because price is the variable with the most leverage over the number.

It is worth noting how differently the two halves of the mining-and-staking world weather this. A proof-of-stake validator earns a yield denominated in its own token and carries almost no marginal cost, so a drawdown dents the fiat value of the reward but not the ability to keep earning it, a dynamic we track in which liquid-staking protocols survive the drawdown. A miner, by contrast, has a power bill due every month in dollars. When revenue falls below that bill, the machine stops. Staking bends in a bear market; mining breaks.

What Actually Survives

Pull the threads together and the survivors of a $57,000 stress test share three traits: genuinely cheap power, a balance sheet that can fund both machine refreshes and AI construction, and a site profile that lets them switch from hashing to colocation. MARA chief executive Fred Thiel framed the end state bluntly to CoinGeek: mining is a zero-sum game where margins compress as capacity is added and the floor is your energy cost, and that by 2028 a miner will either be a power generator, be owned by one, or be partnered with one. The days of plugging a shipping container of ASICs into the grid and clipping a margin are ending.

Two longer clocks sit behind the current squeeze. The first is the 2028 halving, which will cut the block subsidy again and force the margin math to work on half the issuance, unless transaction fees rise to fill the gap. So far they have not; fees have run under 1% of the block reward for much of 2026, and the cheapest-fee activity keeps migrating to other layers, as our piece on how L2 fees became almost free details. The second clock is regulatory, though here the news is calmer: in the United States the SEC has signaled that proof-of-work mining itself is not a securities activity, which removes one tail risk even as power-market and tax questions multiply at the state level.

None of that changes the number on the screen today. At $57,000, with hashprice near $28 and difficulty about to climb, the median public miner is producing a product that costs more to make than it fetches, and is covering the gap by selling treasury coins and renting its power to AI. That is not a recovery. It is a stress test, and the fleet is being graded in real time, block by block.

Frequently Asked Questions

How do you calculate a Bitcoin mining margin?

A mining margin is revenue minus cost per Bitcoin produced. Revenue is captured by hashprice, the daily dollars a machine earns per petahash of computing, which bundles the block subsidy, fees, the Bitcoin price and network difficulty. Cost is driven by the machine’s efficiency in joules per terahash and the price of electricity, with depreciation added for the all-in figure. The margin is positive on a cash basis when the Bitcoin price exceeds the power cost of producing a coin, and positive on an all-in basis only when it also covers depreciation and overhead.

Is Bitcoin mining profitable at $57,000?

It depends entirely on the machine and the power price. At a hashprice near $28 per petahash per day, only efficient modern rigs on cheap power (below roughly 8 cents per kilowatt-hour) are cash-profitable, while older fleets on ordinary grid power are losing money. On an all-in basis that includes depreciation, most listed miners are underwater at $57,000, since JPMorgan estimates the network’s production cost at about $86,000.

What is hashprice and why does it matter?

Hashprice is the revenue a miner earns per unit of computing power per day, usually quoted in dollars per petahash per second per day. It combines the Bitcoin price, the block subsidy, transaction fees and network difficulty into one number, so it is the single best gauge of mining revenue. In late November 2026 hashprice fell to roughly $28 to $30, its weakest since August and about half its level a year earlier.

Why did MARA report a record profit while mining margins fell?

MARA’s $632.1 million third-quarter net income was driven mainly by an $885.3 million unrealized mark-to-market gain on its Bitcoin treasury under fair-value accounting, not by mining profitability. Its operating revenue actually fell about 35% year over year. Because the gain reflects the Bitcoin price during the quarter, it reverses as the price falls, which is why the reported profit says little about the real cost-to-mine margin.

What happens to Bitcoin miners when the price falls below production cost?

The highest-cost miners switch off because their power bill exceeds the value of the coins they produce. As machines go dark, network difficulty adjusts downward, which lowers the production cost and raises the margin for the miners still running. This self-correcting floor is real but slow, and it works by pushing marginal and over-leveraged operators toward curtailment, losses or bankruptcy.

By the HOGE Wire mining desk. Figures current as of 21 November 2026; nothing here is investment advice.

Share 𝕏 Post Telegram