Bitcoin Mining Margins in 2026: When the Coin Costs More to Make

On November 3, 2026, Bitcoin changed hands near $76,500, down about 2% on the day and well below the levels above $80,000 it held through much of October, after the fallout from the record Binance exchange breach rattled markets, according to Fortune. For a trader, that is a drawdown. For the companies that actually produce new bitcoin, it is closer to an existential question, because a large share of them now spend more to mine a coin than the coin is worth.

The single number that captures the pressure is hashprice, the daily revenue a miner earns per unit of computing power. It sat around $33 per petahash per day at the start of November, less than a quarter of its 2021 peak and below the roughly $43 it fetched a year earlier, per CoinDesk and CoinLaw data. Against that revenue, the all-in cost to produce one bitcoin at several large public miners ran past $90,000 in the most recent reported quarter, with the weighted industry average cash cost near $75,500, according to the CoinShares Q2 2026 mining report.

This piece explains how a mining margin is built, why 2026 compressed it to the thinnest levels in Bitcoin’s history, and how the biggest operators are responding, often by renting some of their megawatts to artificial-intelligence tenants instead of pointing them at the hash function. The thread running through all of it is simple: a mining margin is a bet on three variables, the Bitcoin price, the cost of electricity, and the efficiency of your machines, and in 2026 all three turned against the median miner at once.

What a Mining Margin Actually Measures

A mining margin is the gap between what a miner earns for producing blocks and what it costs to produce them. The revenue side is straightforward. Each block carries a subsidy, currently 3.125 BTC after the April 2024 halving, plus whatever transaction fees users attached to the transactions in that block. A miner’s slice of that revenue is proportional to its share of the global hashrate. Multiply the bitcoin earned by the spot price and you have revenue in dollars. For why that subsidy number matters so much to the long-run math, see our explainer on Bitcoin halving cycle math.

The cost side is where the differences hide. The largest line is electricity, usually 60% to 80% of a miner’s direct cash cost. On top of that sit hosting or facility fees for operators who rent space, pool fees of roughly 1% to 2%, labor, and maintenance. Then come the costs that do not show up on a monthly power bill but still determine whether a company makes money: depreciation of the mining rigs, which lose value fast, selling and administrative expenses, interest on debt, and taxes.

• Electricity: the dominant line, typically 60% to 80% of a miner’s direct cash cost.
• Hosting and pool fees: facility charges for hosted machines plus roughly 1% to 2% in pool fees.
• Depreciation: the written-down cost of the rigs, spread across every coin they produce.
• Overhead, interest, and tax: selling and administrative expense, debt service, and taxes.

That split produces two very different margins. The gross or cash margin is revenue minus direct cash costs, mostly power. The all-in margin is what remains after depreciation, overhead, interest, and tax. A miner can be comfortably cash-flow positive and still post a large net loss once the cost of the hardware it bought in the last bull market is written down. When a mining executive says the company produces bitcoin for a certain price, the first question to ask is always which cost they mean.

Hashprice, the Only Revenue Number That Matters

If a margin is revenue minus cost, hashprice is the revenue half distilled into a single figure. It measures the dollars of daily income a miner can expect for each unit of hashrate it runs, usually quoted per petahash per second, per day. The formula is the block subsidy plus fees, multiplied by the Bitcoin price, divided by the total network hashrate. When price rises, hashprice rises. When the network adds machines, every existing miner’s hashprice falls. For a deeper walk through the metric, our hashprice explainer breaks down each input.

Jaran Mellerud, founder of the mining advisory firm Hashlabs, does not hedge about its importance. Writing in June 2026, he called it plainly: “Hashprice is the king of Bitcoin mining metrics. If there is only one number you should monitor, it is hashprice.” His framing captures why miners and bitcoin holders look at different dashboards: the BTC price tells you what a bitcoin is worth, while hashprice tells you what a unit of hashrate is worth.

Through 2026, that number told a grim story. Hashprice set a record monthly average low near $30 per petahash per day in June, with daily readings dipping under $28, according to the CoinShares report. It recovered to roughly $38 by September, per CoinLaw, then slid back toward $33 at the start of November as Bitcoin sold off. For context, hashprice averaged around $56 per petahash per day in the third quarter of 2025 and roughly $43 a year ago. It has, in effect, roughly halved over twelve months while the cost of running a machine stayed put.

The Cost Stack, From Power Bills to Depreciation

Because revenue per unit of hashrate is identical for everyone, margins are won or lost on the cost side, and the cost side is dominated by the price of electricity. The best-positioned public miners buy power at or below five cents per kilowatt hour. Riot Platforms reported a net power cost of about 3.6 cents per kilowatt hour in the second quarter of 2026 after accounting for the credits it earns by selling capacity back to the Texas grid, while MARA reported an owned-site power cost near four cents, according to figures compiled by CoinLaw. A fraction of a cent per kilowatt hour, compounded across hundreds of megawatts running every hour of the year, is the difference between a profit and a loss.

Depreciation is the quiet killer. A modern mining rig is a depreciating asset with a useful life of only a few years, and the fleets that today’s public miners run were often bought when hashprice was double current levels. That purchase cost has to be spread across the coins the machine produces, and when those coins are worth less than expected, the per-coin depreciation charge balloons. Add selling and administrative expense, stock compensation, and interest on the debt many miners took on to expand, and the all-in cost can sit tens of thousands of dollars above the cash cost.

The practical takeaway is that two companies running the identical hardware can report wildly different margins based purely on where they plugged it in and how they financed it. That dispersion, rather than any single industry average, is the real texture of mining economics in 2026.

The 2026 Squeeze, When the Coin Costs More Than It Sells For

The clearest illustration comes from Riot’s own filings. In the second quarter of 2026, Riot reported a direct cost to mine of $49,912 per bitcoin, covering power and other cash costs. Once rig depreciation was layered in, its all-in cost reached $90,631 per coin. The bitcoin it produced that quarter was worth $71,667 on average. In other words, Riot lost $18,964 on every coin it mined on an all-in basis, and it was the third consecutive quarter in which its all-in costs exceeded the value of what it produced, according to CoinLaw’s reading of the filings.

Riot is far from the worst placed. The CoinShares Q2 2026 report put the weighted average ex-tax cash cost to produce one bitcoin at roughly $75,500 across listed miners, with electricity-only costs ranging from about $26,000 per coin at the most efficient operators to roughly $40,000 at others. On an all-in basis, including depreciation and overhead, the spread was enormous, running from the high $70,000s for the leanest producers to several hundred thousand dollars per coin for stressed operators carrying heavy debt and old machines. The blunt summary, as one industry tally put it, is that a meaningful share of the industry is producing coins for more than the coins are worth.

The reason the April to June quarter looks so brutal is that Bitcoin spent much of it depressed, ending the period around $58,400 before recovering, according to CoinShares. When the price of your only product falls faster than you can cut costs, the margin inverts, and for mining there is no quick lever to pull because the electricity meter runs whether coins are profitable or not.

Efficiency Is Destiny, and the J/TH Dividing Line

Since power is the dominant cost, the efficiency of a machine, measured in joules per terahash, decides how much electricity it burns for a given amount of revenue. The lower the number, the less power a rig consumes to produce each unit of hashrate, and the higher the electricity price it can tolerate before it stops making money. That single metric sorts the industry into survivors and casualties.

The breakeven math is unforgiving. Using hashprice levels from September 2026, CoinLaw calculated the maximum electricity price each class of machine can pay before mining at a loss. A state-of-the-art rig near 16 joules per terahash stays profitable up to about ten cents per kilowatt hour. A mid-generation machine around 25 joules per terahash breaks even near five cents. An aging unit at 45 joules per terahash needs power under four cents just to cover its electricity bill, before a single dollar goes toward hardware or overhead.

Hardware has kept pushing that frontier lower. Bitmain’s Antminer S23 Hydro, a liquid-cooled unit producing 580 terahashes per second at 5,510 watts, runs at 9.5 joules per terahash, the first commercial Bitcoin miner to ship at scale below ten joules, according to a review from Million Miner. That efficiency is a lifeline for operators with cheap power and a death sentence for anyone still running S19-class machines at 30 or more joules per terahash in a region where electricity costs more than a few cents. Efficiency, not scale, is what keeps the lights on.

Gross Margin Versus All-In Margin, the Gap That Hides the Pain

The distance between a healthy gross margin and a negative net margin is where most of 2026’s damage lives. Riot’s mining gross margin tells the story on its own: it fell from 50% in the second quarter of 2025 to 59% in the third quarter, then collapsed to 30% by the second quarter of 2026, according to CoinLaw. A 30% gross margin still sounds respectable, yet the same company was losing nearly $19,000 per coin all-in. The gap is depreciation, overhead, and interest.

CleanSpark offers a parallel view. In its fiscal third quarter, the three months ending June 30, 2026, the company reported revenue of $138 million, down 30.5% from a year earlier, a gross margin of 38%, and a net loss of $239.8 million, with adjusted earnings before interest, taxes, depreciation, and amortization of negative $113 million, per its results release. The average revenue per bitcoin mined was roughly $72,000, down about 5% from the prior quarter. A positive gross margin and a nine-figure net loss can coexist, and in 2026 they routinely did.

At the extreme sits Core Scientific, which now treats self-mining as a secondary, loss-tolerant activity while it shifts toward leasing its infrastructure. Industry reporting put its self-mining gross margin at around negative 56%, meaning the operation loses money before any overhead at all, according to a TechFlow analysis of miner filings. That is not an accident or a failure of execution; it is a deliberate choice about where a megawatt of power earns the most.

The Difficulty Treadmill

Every margin conversation eventually runs into the difficulty adjustment, Bitcoin’s built-in mechanism for keeping blocks roughly ten minutes apart no matter how much hashrate joins the network. As of early November, the network hashrate hovered around one zettahash per second, with a seven-day average near 1,069 exahashes and daily readings dipping toward 950 exahashes as some miners throttled after the price drop, according to CoinWarz. Difficulty stood near 148 trillion.

That number is quietly remarkable. A year earlier, in November 2025, difficulty had climbed to a then record near 156 trillion, per CoinDesk. By November 2026 it had eased to about 148 trillion, a rare retreat in a metric that normally only climbs. Mellerud has argued that 2026 could become the first year in Bitcoin’s history to end with a lower hashrate than it began, suppressed by two forces at once: poor mining economics and the conversion of mining capacity to AI. The all-time hashrate high remains the roughly 1,160 exahash per second peak set in October 2025, per CoinShares.

The treadmill matters for margins because difficulty is a dilution machine. When a competitor plugs in a new fleet, it does not create new block rewards; it simply takes a larger slice of the same fixed pie, lowering everyone’s hashprice. That is why margins compress over time even when the Bitcoin price is flat, and why a stall in network growth, painful as it is for the miners switching off, is a quiet gift to the efficient operators that remain.

The Fee Problem, a Security Budget Running on Fumes

In theory, transaction fees are supposed to pick up the slack as the block subsidy shrinks with each halving. In practice, they have been close to irrelevant to mining margins in 2026. Across a representative window of blocks this year, transaction fees supplied only about 0.6% of the total miner reward, with the subsidy providing the rest, according to CoinLaw. When blocks are not full, a miner’s revenue is essentially the subsidy multiplied by the price, and nothing more.

That has a direct consequence for margin volatility. A miner today is almost purely a leveraged bet on the Bitcoin price, with no meaningful fee cushion to soften the blow when price falls. It also raises the longer-term question of how Bitcoin pays for its own security once the subsidy keeps halving toward zero, a debate explored in our look at Runes and Bitcoin’s security budget. For now, the fee market is a rounding error on the margin statement, and miners plan as if it will stay that way.

The AI Escape Hatch

Faced with negative all-in margins, the largest public miners reached the same conclusion: the most valuable thing they own is not a warehouse full of ASICs, it is a grid connection and the right to draw hundreds of megawatts of cheap power. Artificial-intelligence companies, desperate for that exact resource, are willing to sign long, predictable leases for it. So miners started renting their power to AI instead of pointing all of it at the hash function.

Riot struck the signature deal of the year, a 20-year lease for 191 megawatts of capacity at its Rockdale, Texas campus with a leading AI lab identified as Anthropic, expected to generate about $9.1 billion in initial contract revenue, according to CoinDesk. Combined with an existing lease to AMD, Riot now controls about 241 megawatts of contracted capacity and roughly $9.8 billion of long-term revenue, per Data Center Dynamics. Cipher Mining went further still, signing a roughly $5.5 billion, 15-year lease with Amazon Web Services for 300 megawatts, then adding a deal with AI cloud firm Fluidstack worth up to about $7 billion with extensions, a pact in which Google agreed to backstop $1.4 billion of lease obligations and took warrants for about a 5.4% stake, according to Data Center Dynamics and its reporting on the Fluidstack and Google arrangement. Cipher’s AI hosting commitments now total around $8.5 billion.

Not every pivot goes smoothly. CoreWeave agreed in July 2025 to acquire Core Scientific in a $9 billion all-stock deal, but Core Scientific’s own shareholders rejected it at a special meeting on October 30, 2025, with roughly 203 million votes against and only about 21 million in favor, terminating the merger, according to The Block. Core Scientific now pursues its own data-center strategy as an independent company on the Nasdaq. CleanSpark framed the industry-wide shift carefully. “We continue to successfully execute on our strategic evolution to a diversified digital infrastructure platform,” chairman and chief executive Matt Schultz said in the company’s results release. Read through the corporate language, and the message is that mining is becoming one use of the power, not the only use.

The Public Miners’ Scoreboard

The headline losses are large, but they need a caveat. In the second quarter of 2026, MARA reported a net loss of $611.3 million on revenue of $174.9 million, down 27% year over year, while CleanSpark’s net loss for the quarter ending June 30 came to $239.8 million. MARA and CleanSpark posted a combined $851 million in quarterly losses, according to reporting compiled by Yahoo Finance. MARA is due to report its third-quarter results after the close on November 3, with analysts expecting another loss on revenue near $183 million, per MarketBeat.

The caveat is that much of the GAAP loss is non-cash. Under fair-value accounting rules, miners that hold bitcoin on their balance sheets mark those holdings to market every quarter, so a falling price produces a large paper loss even if the company never sold a coin. CleanSpark, for instance, still held about $814.9 million of bitcoin and $202.6 million of cash at quarter end, per its release. The cash picture is less dire than the net-loss line suggests, which is exactly why the gross margin and the per-coin cash cost, not the bottom line, are the numbers insiders watch.

The Case for Mining Through the Trough

If the all-in math is negative, why keep mining at all? The classic argument is the mine-versus-buy comparison. Speaking to The Block in February 2025, Zach Bradford, CleanSpark’s chief executive at the time, put it bluntly: “Why buy bitcoin at current prices when we can mine it for $34,000?” That logic was airtight when the marginal cost sat far below the spot price. In 2026 it inverted for many operators, as cash costs climbed toward and past the price of a coin, which is precisely why the industry’s tone shifted from accumulation to survival.

Several levers still justify running through the trough. The first is operating leverage: because power costs are largely fixed, a recovery in the Bitcoin price flows almost entirely to the bottom line, so miners that stay online capture the upside fast. The second is flexible load. In Texas, Riot and others earn credits by curtailing consumption and selling power back to the grid during peak demand, which is how Riot pushed its net power cost down to about 3.6 cents per kilowatt hour, per CoinLaw. A mining facility that can switch off in seconds is a valuable grid asset, and that value shows up as a lower effective electricity cost. The third is the treasury itself: miners that hold their coins are making a bet that the price will eventually make today’s costs look cheap.

Geography, Power, and the SEC Lens

Because electricity is the margin, geography is strategy. Since China’s 2021 mining ban, the United States has become the center of gravity for public miners, with Texas a particular magnet thanks to its deregulated grid, abundant wind and solar, and demand-response programs that pay large loads to curtail. The winning sites share a profile: cheap, often stranded or intermittent power, flexible interconnection terms, and room to grow. The losing sites are those paying retail-adjacent rates on old machines.

For US-listed miners, the margin story is also a disclosure story, and here the Securities and Exchange Commission sets the frame. Public miners file quarterly 10-Q and current 8-K reports, but the “cost to mine” figures they headline are non-GAAP metrics with no standardized definition, which is how one company can advertise a cash cost in the $30,000s while quietly carrying an all-in cost above $90,000. Investors have to read past the press release into the filings to find the comparable number. The SEC’s adoption of fair-value accounting for crypto holdings has at least made the balance-sheet treatment more consistent, forcing miners to mark bitcoin to market and disclose the swings, even as the operating cost metrics remain a patchwork. The lesson for readers is to treat any single advertised mining cost with skepticism until the basis is clear.

Margins, Leverage, and the Rest of the Market

Mining margins do not sit in isolation; they amplify whatever Bitcoin itself is doing. A miner with fixed power costs and a fleet of depreciating machines is a leveraged position on the Bitcoin price, which is why miner equities tend to swing harder than the coin in both directions. When the Binance breach pulled Bitcoin below $77,000 in early November, the knock-on volatility reached everything from on-chain trading strategies to the mining stocks, which fell faster than the coin itself.

That leverage cuts both ways, and it explains the industry’s split personality in 2026. The miners pivoting hardest to AI are, in effect, trying to swap a volatile, price-linked margin for a stable, contracted one. The miners doubling down on efficiency are betting that a stall in network hashrate will lift hashprice for whoever remains. Both are rational responses to the same squeeze, and both concede the central point: the era of easy mining margins, when any functioning machine made money, is over.

What to Watch Into 2027

Four variables will set mining margins over the next year. The first is the Bitcoin price, still the single largest lever, with the Federal Reserve’s December meeting and the broader macro picture driving sentiment. The second is network hashrate: if Mellerud is right that growth has stalled, surviving miners could see hashprice stabilize even without a price rally. The third is the AI lease ramp, where the question shifts from signing deals to actually delivering megawatts and collecting rent, since contracts worth billions mean nothing until the data halls energize. The fourth is the efficiency floor, as sub-ten-joule machines like the S23 Hydro reset the breakeven power price lower and strand another layer of older hardware.

Further out, the next halving in 2028 will cut the subsidy again, to 1.5625 BTC per block, which will double the pressure on both margins and the fee market unless the Bitcoin price more than compensates. For now, the picture is clear enough. A mining margin in 2026 is a function of price, power, and efficiency, and the companies that will still be mining in 2027 are the ones that either own the cheapest electrons on the grid or have found someone willing to pay more for them than the hash function ever could.

Frequently Asked Questions

Is Bitcoin mining profitable in 2026?

It depends entirely on the miner. On a cash basis, operators with power costs around four cents per kilowatt hour and modern machines remained gross-margin positive through 2026. On an all-in basis, including depreciation and overhead, many large public miners produced bitcoin for more than it was worth during the second quarter, with Riot, for example, losing roughly $19,000 per coin. Efficiency and electricity price, not scale, decide who makes money.

How much does it cost to mine one Bitcoin in 2026?

The weighted average cash cost across listed miners was about $75,500 per coin in the second quarter of 2026, according to CoinShares. Electricity-only costs ranged from roughly $26,000 at the most efficient operators to about $40,000 at others, while all-in costs including depreciation and interest stretched from the high $70,000s to well over $300,000 per coin for stressed operators. Always check whether a quoted cost is cash or all-in.

What is hashprice and why does it matter for margins?

Hashprice is the daily revenue a miner earns per unit of hashrate, usually quoted per petahash per day. It equals the block subsidy plus fees, times the Bitcoin price, divided by the network hashrate. It is the revenue side of the margin in a single number, which is why analysts call it the master metric. In early November 2026 it sat around $33 per petahash per day, down from roughly $43 a year earlier.

Why are Bitcoin miners switching to AI and data centers?

Because the most valuable asset a miner owns is cheap, large-scale grid power, and AI companies will sign long, predictable leases for it. Riot agreed a 20-year, roughly $9.1 billion lease with Anthropic, and Cipher Mining assembled about $8.5 billion in AI hosting commitments with AWS, Fluidstack, and Google. Leasing power to AI swaps a volatile, price-linked mining margin for stable contracted revenue.

What is the most efficient Bitcoin miner in 2026?

Bitmain’s Antminer S23 Hydro is the most efficient unit shipping at scale, producing 580 terahashes per second at 5,510 watts for an efficiency of 9.5 joules per terahash, the first commercial Bitcoin miner to go below ten joules per terahash. That efficiency lets it stay profitable at electricity prices up to about ten cents per kilowatt hour, roughly triple what an older 45-joule machine can tolerate.

By the HOGE Wire mining desk, covering the economics of proof-of-work and the companies betting on it. Figures current as of November 3, 2026; nothing here is investment advice.