Bitcoin Mining Margins in 2026: Behind the Q3 Profit Line

On November 4, Riot Platforms became the first large Bitcoin miner to report its third quarter, and the headline read like a turnaround: a small net profit of about $4.0 million, adjusted EBITDA of roughly $191 million, and a fresh data-center lease worth some $28 billion. Strip away the accounting, though, and the mining business told a plainer story. Riot produced 1,637 bitcoin at a direct cost of $39,265 each, while the coins it mined averaged $70,855 over the quarter. The distance between those two figures, not the profit line, is what a Bitcoin mining margin actually is.

That distance swung violently in 2026, and it is swinging again right now. The third quarter was Bitcoin’s best since 2017, up roughly 43%, which lifted almost every public miner back above its cash break-even. By the time Riot reported, the tailwind had already reversed. Bitcoin was back near $76,000, mining difficulty had climbed to a 2026 high, and hashprice (the single number that captures a miner’s revenue) had slid to its lowest level since the summer. MARA Holdings, Riot’s larger rival, does not report until November 9, and its numbers will carry the same warning label: in 2026, a miner’s profit line is mostly a bet on the Bitcoin price, while its margin is a bet on three very different things.

This piece walks through how price, power and efficiency set a miner’s margin, why the cost-to-mine line matters more than net income, what Riot’s third quarter actually revealed, and why the relief it showed is already fading. The reader who learns to read past the headline will understand more about a mining company in ninety seconds than the net-income figure will ever tell them.

A Margin Is a Bet on Three Numbers

A Bitcoin mining margin is simpler than the quarterly filings make it look. Revenue comes from one activity: winning blocks, collecting the 3.125 BTC subsidy plus a thin layer of transaction fees, and valuing that reward in dollars. Costs come mostly from one input: electricity. Everything else (hardware, hosting, staff, interest) stacks on top. So profitability turns on three variables, and only three: the Bitcoin price, the price a miner pays for power, and how efficiently its machines convert that power into hashes.

Analysts and filings express this through three different margins, and conflating them is the most common mistake a reader can make. The gross or cash margin counts only the electricity and pool fees needed to run a machine today. The operating margin adds the fixed costs of running the business. The all-in margin adds depreciation, the slow write-down of the ASIC fleet a miner paid for up front. Each answers a different question.

The gap between the cash margin and the all-in margin is where most of the drama lives. A rig can be cash-profitable, worth running because power is cheaper than the coins it earns, while the company that owns it still loses money once depreciation is counted. Riot’s own filings are the cleanest example. In the second quarter its cost to mine a bitcoin excluding depreciation was $49,912, but including depreciation it was $90,631, roughly 126% of what each coin was worth. Same quarter, same machines, two opposite verdicts depending on which margin you read.

Hashprice: The Whole Revenue Side in One Number

Because the revenue side depends on four moving parts at once (the block subsidy, fees, the Bitcoin price and the total network hashrate), the industry squeezes it into a single metric: hashprice, a term coined by Luxor’s Hashrate Index. Hashprice is the expected daily revenue per unit of hashing power, quoted in dollars per petahash per second per day. The rough formula is about 144 blocks a day, times the subsidy plus fees, times the Bitcoin price, divided by the network hashrate. If you know a fleet’s hashrate, hashprice tells you its gross revenue before a single cost is subtracted.

In early November hashprice sat near $32.86 per petahash per day, down 8.7% in a single week and roughly 18% below its October 1 level near $40, as the hedging market priced in a winter pause. The metric is powerful precisely because it moves with everything a miner cannot control, and it compresses all of it into one line. When the Bitcoin price falls, hashprice falls. When the network adds hashrate or difficulty rises, hashprice falls. When fees dry up, hashprice falls. In late 2026 all three point the wrong way. Transaction fees have collapsed to under 0.7% of the block reward, a ten-year low, which means the fixed 3.125 BTC subsidy is carrying almost the entire revenue line. That matters for margins because a shrinking fee contribution removes the one cushion that can offset a falling price.

Why the Third Quarter Looked Like a Recovery

To read the Q3 earnings season correctly, start with the backdrop every miner shares. The third quarter was Bitcoin’s best since 2017, rising roughly 43% to close September at $83,554, up from $58,524 at the end of June. That single move did more for miner margins than any efficiency upgrade could. Hashprice climbed back toward $40 per petahash per day by the start of October, and total network mining revenue reached about $1.215 billion in October, up 8.5% on September and the strongest month since November 2025. In September, for the first time in ten months, aggregate miner income exceeded total mining costs by about 6%.

The recovery is real, but it is worth remembering what it recovered from. Late 2025 was, in the words of Wolfie Zhao, head of research at TheMinerMag, a brutal stretch. “Bitcoin mining has entered what is effectively the harshest margin environment of all time,” he wrote in December, as hashprice slipped from a stable third-quarter average near $55 to a structural low around $35. The Q3 2026 rebound pulled the industry off that floor. It did not change the floor.

The Fair-Value Mirage: Why the Profit Line Lies

Here is the accounting wrinkle that makes a miner’s net income almost useless as a margin gauge. Since the Financial Accounting Standards Board adopted rule ASU 2023-08, public miners carry their Bitcoin holdings at fair value and run the quarterly change straight through the income statement. When Bitcoin rises, a miner holding coins books a large unrealized gain that inflates net income. When Bitcoin falls, the same mechanism buries it. MARA’s first-quarter 2026 net loss of about $1.3 billion was driven by a roughly $1 billion mark-to-market swing on its treasury, not by anything that happened in its data centers.

The practical consequence: in a quarter when Bitcoin jumps 43%, net income mostly measures the coin price, not the mining business. That is why the profit line can make the laggard look like the leader. A miner that hoarded tens of thousands of coins will book a paper gain worth hundreds of millions, flattering its headline, even if its machines are less efficient and its power costs higher than a rival that sold its coins to build data centers. When MARA reports on November 9, expect exactly this tailwind. The third-quarter season already showed the pattern at scale: Riot, Core Scientific, Cipher and Hut 8 all posted positive GAAP net income, and in nearly every case the profit came from non-cash marks (bitcoin revaluation or warrant gains) rather than from operations. Core Scientific booked a headline profit of about $1.2 billion while its results for the first nine months of the year were still a net loss. The disciplined way to read any of these reports is to skip net income entirely and look at two numbers: the cost to mine a bitcoin, and the hashprice. Those two are the margin. Everything else is a bet on the chart.

Riot’s Third Quarter, Decoded

Riot’s third-quarter report is a clean test case. Total revenue was $181.4 million. Bitcoin-mining revenue of $116.0 million fell from $160.8 million a year earlier, because the average realized Bitcoin price dropped to $70,855 from $114,361, a decline that a 16% jump in production could only partly offset. Engineering revenue of $47.0 million (custom electrical equipment sold to third-party data centers) and data-center revenue of $18.4 million rounded out the top line, the latter reflecting the first full quarter of the stabilized 25 megawatts delivered to AMD.

Now the margin. Riot’s cost to mine a bitcoin, excluding depreciation, was $39,265, down from $46,324 a year earlier, helped by lower power costs, higher production, and $24.7 million of power-curtailment credits worth $15,067 per bitcoin mined. Its net cost of power came to 3.2 cents per kilowatt-hour, among the lowest in the industry. Against a realized price near $70,855, that direct cost looks comfortable: the cash margin is clearly positive. Yet adjusted EBITDA of $191.3 million translated into net income of only $4.0 million. The gap is depreciation and other non-cash charges, which swallowed nearly the entire operating result, plus a $26.3 million gain on the change in fair value of derivatives that helped nudge the final figure above zero. Counting depreciation, Riot’s all-in cost to mine a coin was $78,805, above the $70,855 it realized per bitcoin; on a fully loaded basis, the mining itself lost money even as the company reported a profit. The all-in margin, in other words, was negative.

One more detail reframes the whole report. Riot ended the quarter holding just 2,418 bitcoin, worth about $202 million, after selling 20,264 coins for roughly $1.5 billion during the year to fund its data-center build. The company that was once a Bitcoin accumulator has become, in effect, a landlord that happens to run some miners. Its $4 million profit is genuine, but it is not a mining-margin story. It is an everything-else story.

Cost to Mine Is Not One Number, and Not Comparable

The headline most readers reach for is “cost to mine one bitcoin,” and it is the most abused figure in the sector. There is no standard definition, so every company draws the line somewhere different, and the numbers are not comparable across filings. Riot’s $39,265 excludes depreciation. MARA reports a narrow purchased-energy figure that was just $38,690 in the second quarter, a number that looks great precisely because it leaves out almost everything except electricity at its owned sites. An independent benchmark tells a harsher story: CoinShares pegged the weighted-average cash cost across public miners at roughly $75,500 per bitcoin in the second quarter, above the average Bitcoin price for the period. And then there are the outliers distorted by the AI transition, where shared infrastructure scrambles the per-coin math entirely.

The lesson is not that anyone is lying. It is that a reader has to ask what each figure counts before comparing it to a rival’s or to the spot price. When a miner leads its release with a low energy-only number, that is a cash margin; it says the rigs are worth running. It says nothing about whether the company earns back the hardware.

The Cost Stack: Where the Money Goes

Behind the single cost-to-mine figure sits a stack. Electricity is the largest variable cost, typically the majority of a miner’s cash spend. Hardware is the largest fixed cost, and it is the reason the cash and all-in margins diverge so sharply: a modern ASIC costs millions per exahash and loses value on a punishing schedule, so depreciation lands on the income statement quarter after quarter whether or not the machines are switched on. Useful-life assumptions across the sector span anywhere from about eighteen months to five years, and that single accounting choice can move a reported all-in cost by tens of thousands of dollars per coin.

On top of power and depreciation come pool fees (generally a few percent of mined revenue), hosting charges for miners that rent space rather than own it, salaries, and interest on the debt most miners took on to build. The efficiency treadmill keeps the hardware line moving. Because each new chip generation is meaningfully more efficient, a fleet that stands still gets competed into the red, which forces continuous reinvestment. Jaran Mellerud, founder of Hashlabs Mining, captured how little slack there is when import tariffs threatened to raise rig prices: “In an industry as cost-sensitive as Bitcoin mining, a 22% price increase on machines can make operations financially unsustainable,” he told Cointelegraph. A business that thin on the hardware line cannot absorb a bad power contract.

The Power Bill Is the Real Margin

If one line decides a miner’s fate, it is the price of electricity. Profitable large-scale operations generally pay somewhere between three and seven cents per kilowatt-hour; Riot’s net 3.2 cents is best in class, and it gets there not by buying cheap power alone but by selling flexibility back to the grid. In Texas, miners enrolled in ERCOT demand-response programs power down during peak demand and are paid for it; the only cost of curtailing is the hashrate they forgo, with no spoiled product or restart penalty, which makes a mining fleet close to an ideal flexible load. Riot booked $24.7 million of such credits in the third quarter, the equivalent of knocking $15,067 off the cost of every coin it mined.

That is also why geography is destiny. Sites in the cheapest jurisdictions pay four to six cents; much of Europe, with industrial power often three to five times higher, is simply uneconomic for Bitcoin mining at current hashprice. The energy intensity carries a footprint beyond the bill, too, from grid strain to the water used to cool large sites and generate the electricity, a cost that rarely appears in any margin table but increasingly shapes where miners are welcome. Fred Thiel, MARA’s chief executive, frames the whole business around this one input. “Bitcoin mining is a zero-sum game,” he told CoinDesk. “As more people add capacity, it gets harder for everybody else. Margins compress, and the floor is your energy cost.” His survival math is blunt: “Our strategy is to be in the lowest quartile in terms of production cost,” he said, “because in a tight market, 75% of the other guys have to shut down before we do.”

The Breakeven Treadmill: Which Rigs Survive

Efficiency is the third leg of the margin, and it decides which machines keep running when hashprice falls. A rig’s cash breakeven is the power price at which the electricity it burns exactly equals the coins it earns; above that price it loses money every day it stays on. The math is simple enough to do on a napkin: breakeven power price in dollars per kilowatt-hour is roughly the hashprice divided by twenty-four times the machine’s efficiency in joules per terahash. Run it at the current $32.86 hashprice and the spread between new and old hardware is stark.

The figures are illustrative cash breakevens (they ignore depreciation), computed from published Bitmain efficiency specs, but the shape is what matters. A current-generation S23 Hydro at 9.5 joules per terahash stays cash-positive up to about fourteen cents, a price almost no serious operation pays, so it prints money on power alone. A four-year-old S19j Pro needs electricity below roughly four and a half cents just to cover its own power bill, which is why the oldest machines are the first to switch off whenever hashprice dips. This is the treadmill: because the marginal, least-efficient fleet sets the floor, every miner is forced to keep buying faster chips just to stay in place, and that reinvestment is exactly the depreciation that turns a cash profit into an all-in loss.

The Network Already Clawed It Back

The cruelest feature of Bitcoin mining is that recovery contains the seeds of its own reversal, and the timing of the Q3 reports makes the point for us. As the price rose through the autumn, idle rigs switched back on and new capacity energized, pushing the network’s hashrate back above one zettahash per second (a seven-day average near 1,070 exahash). Difficulty, which retargets roughly every two weeks to keep blocks ten minutes apart, responded exactly as designed: it rose 5.54% on October 29 to 148.36 trillion, the eleventh upward adjustment of 2026 and about 16% above its mid-September level, leaving it only around 5% below the record of roughly 156 trillion set in November 2025. More machines chasing the same fixed 3.125 BTC per block means each machine earns less. That is the difficulty thermostat in reverse: a rising price lifts margins, the extra margin pulls in hashrate, and the added hashrate competes the margin straight back out.

Then the price leg broke too. By November 6, Bitcoin had slid back toward $76,000, down roughly 6% on the day and giving up much of the third-quarter gain, as risk appetite cooled and the market was still digesting the fallout from the late-October breach at Binance. Hashprice, caught between a higher difficulty and a lower price, fell 8.7% in a single week to $32.86 and roughly 18% from its October 1 level. The two blades of the scissors, rising cost of production and falling revenue, closed at once. JPMorgan analysts had by late September pegged the industry’s all-in production cost near $85,000 a coin and called it a soft floor; with Bitcoin back near $76,000, the average miner was once again under water on a fully loaded basis, no matter how flattering the just-reported quarter looked. The relief the Q3 numbers captured was, in a real sense, already gone by the time anyone read them.

The AI Escape Hatch, and Who Gets In

The surest way to fix a thin mining margin, it turns out, is to stop mining. The third-quarter reports made that strategy explicit. Core Scientific, until recently one of the largest Bitcoin miners in North America, told investors it will leave Bitcoin mining altogether by the end of 2026 to become a pure-play AI colocation company. “We expect to complete our exit from Bitcoin mining by year end,” chief executive Adam Sullivan said as the company reported third-quarter revenue of $232.9 million, more than 90% of it from colocation. Its remaining self-mining ran at a small gross loss, and the headline $1.2 billion profit was a warrant revaluation, not cash from operations.

Core Scientific is the clearest case, not the only one. IREN’s Bitcoin-mining revenue fell 88% to about $28 million as it redeployed power to a $9.7 billion AI-cloud contract with Microsoft. TeraWulf signed a roughly $19 billion, 401-megawatt, twenty-year lease with Anthropic at a Kentucky campus, with Google backstopping the rent. Cipher, now styled Cipher Digital, has booked around $16.6 billion of contracted AI revenue across three campuses, including a 300-megawatt lease with Amazon Web Services. Hut 8 carries a multibillion-dollar AI backlog, and Riot’s own roughly $38 billion of contracted data-center revenue now dwarfs the mining line that still pays most of its bills.

The logic is margin, pure and simple. A megawatt leased to an AI tenant on a fifteen or twenty year contract earns far more, and far more predictably, than a megawatt pointed at the Bitcoin network, where revenue rides the hashprice rollercoaster. CoinShares estimates that an AI or HPC megawatt throws off roughly three times the annual profit of a megawatt spent mining, which is why the capital is flowing one way. Matt Kimmell, an analyst at CoinShares, told Bloomberg the shift may “prove to be the end of an era for some large US miners, not necessarily in terms of survival, but in how they operate.” The catch is that not every miner has a tenant. KBW analyst Stephen Glagola warned of a lease catalyst drought stretching into early 2027 and cut price targets across the group, noting that several miners now trade near or below the value of the leases they have already signed. The margin squeeze, in other words, is sorting the sector into AI landlords and stranded miners.

What MARA’s Report Will, and Will Not, Show

That sorting is the subtext of MARA’s November 9 report, the most-watched of the season because MARA remains the giant most committed to mining itself. Three lines will matter more than the headline. The first is the fair-value tailwind: MARA holds the sector’s largest Bitcoin treasury, roughly 35,600 coins at mid-year, so a 43% quarterly price rise could flip its result to a large paper profit after billion-dollar losses earlier in 2026. That number will say almost nothing about its mining margin. The second is cost to mine. MARA leads with a narrow energy-only figure ($38,690 in the second quarter) that leaves out most of the cost stack; the honest read is an all-in cost set against the realized price, the same test Riot failed on a fully loaded basis.

The third is the anchor tenant, or the lack of one. MARA has bet on vertical integration, owning its power rather than renting capacity to a hyperscaler: U.S. energy regulators cleared its roughly $1.5 billion purchase of the 505-megawatt Long Ridge gas plant in early November, and it has a roughly one-gigawatt AI partnership with Starwood, but it has not yet signed a marquee AI lease at scale. Fred Thiel’s wager is to be the lowest-cost producer and control the electrons; the question his report must answer is whether that wager is paying for itself while rivals bank twenty-year rent checks.

The 2028 Problem and the SEC

Even a flawless pivot leaves the structural ceiling in place. In the spring of 2028 the block subsidy halves again, from 3.125 to 1.5625 BTC, cutting in half the input that supplies more than 99% of miner revenue. With fees stuck under 0.7% of the reward, there is no cushion waiting to fill the gap; unless the Bitcoin price roughly doubles to compensate, the revenue line simply halves, and a margin that already looks tight gets tighter. Thiel has been blunt about where that leads: “By 2028, you’ll either be a power generator, be owned by one, or be partnered with one.” The AI pivot is, among other things, a hedge against that cliff.

Whether transaction fees can ever carry the load is the same question Bitcoin’s developers keep circling as they weigh the network’s next soft fork, and it has a mirror on the other side of the consensus divide, where a proof-of-stake validator’s yield is the price of security much as a miner’s margin is, a yield that held up even through this year’s market stress. For US-listed miners, at least, the regulator is a source of stability rather than risk. In March 2025 the SEC’s Division of Corporation Finance stated that proof-of-work mining on public networks is not a securities offering, and that pooling rewards does not change the analysis. The pressure on mining margins is economic, not legal. The real SEC question for these companies now is disclosure, how they mark their Bitcoin and how honestly they report cost to mine, which is exactly the line this quarter’s numbers asked readers to scrutinize.

Frequently Asked Questions

What is the cost to mine one Bitcoin in 2026?

There is no single figure, because every miner defines it differently. Efficient machines on cheap power can post an energy-only cost under $40,000, but the public-miner weighted-average cash cost was about $75,500 in the second quarter of 2026, and JPMorgan put the all-in production cost near $85,000 by late September. With Bitcoin near $76,000 in early November, the average miner was below its all-in cost of production.

What is hashprice and why does it matter for mining margins?

Hashprice is a miner’s expected daily revenue per unit of hashing power, quoted in dollars per petahash per second per day. It compresses the block subsidy, transaction fees, the Bitcoin price and the total network hashrate into one number. In early November 2026 it sat near $32.86, down 8.7% in a single week. When hashprice falls below a fleet’s cost per unit of hashing, that fleet loses money.

Are Bitcoin miners profitable in 2026?

It depends on which margin and which miner. On a cash basis, efficient operations paying three to four cents per kilowatt-hour stayed profitable; on an all-in basis including depreciation, many did not. Riot’s third-quarter all-in cost of $78,805 exceeded the $70,855 it realized per coin. The profits miners reported for the quarter were mostly bitcoin fair-value gains, not mining margin.

Why are Bitcoin miners switching to AI?

Because mining margins are thin and volatile. CoinShares estimates a megawatt leased to an AI tenant earns roughly three times the annual profit of a megawatt spent mining, and the revenue is far more predictable. Core Scientific plans to exit Bitcoin mining entirely by the end of 2026, and IREN, TeraWulf and Cipher are winding their mining down to lease power to AI customers.

What happens to Bitcoin mining margins after the 2028 halving?

In the spring of 2028 the block subsidy halves from 3.125 to 1.5625 BTC, cutting in half the input that supplies more than 99% of miner revenue. With transaction fees under 0.7% of the block reward, nothing is waiting to fill the gap. Unless the Bitcoin price roughly doubles, or a miner has diversified into AI hosting, margins compress further.

By Yuki Tanaka, HOGE Wire mining desk. Figures are current as of November 6, 2026 and move with the market.