Bitcoin Mining Margins Explained: Inside the 2026 Profit Squeeze

Bitcoin’s mining difficulty is due for another retarget on July 11, 2026, one more mechanical adjustment in a year that has already produced some of the sharpest difficulty swings since 2022, including three consecutive negative adjustments earlier this year, the first streak like that since July 2022. That volatility is the network’s own thermostat reacting to a blunter problem underneath it: the gap between what it costs to mine a bitcoin and what a bitcoin is actually worth has narrowed so far that, by CoinShares’ count, something like one in five miners in the global fleet is now running at a loss.

Margin, not price and not even hashrate, is the number that actually decides who keeps a mining rig plugged in and who switches it off. This piece works through both halves of that equation: what miners earn per unit of computing power, known as hashprice, and what it actually costs them to keep that power running, from electricity contracts down to depreciation schedules. It also looks at how the two largest US-listed miners, Marathon Digital (MARA) and Riot Platforms (RIOT), are answering a squeeze that has made pure-play Bitcoin mining one of the thinnest-margin businesses in the entire crypto industry.

What Mining Margins Actually Means

In most industries, margin is simple: revenue minus cost, divided by revenue. Bitcoin mining uses the same arithmetic, but both inputs behave in ways an ordinary business owner would find alien. Revenue is not something a miner negotiates or prices; it is set entirely by the protocol (a fixed block subsidy, currently 3.125 BTC roughly every ten minutes, split across however many machines are pointed at the network) and by a global, around-the-clock spot market for BTC that no single miner can influence. Cost, meanwhile, is dominated by one input, electricity, that a miner does control, within the limits of wherever they can get a contract signed.

So when people in mining circles talk about margin, they usually mean one specific comparison: hashprice (dollars of expected revenue per unit of hashing power per day) against a miner’s all-in cost to produce that same unit of hashing power. When hashprice sits above a miner’s cost, every machine running is printing money. When it falls below, that same machine is burning cash every hour it stays on, and the only real decision left is whether to keep it running anyway, hoping price recovers, or switch it off. That binary, repeated across hundreds of thousands of individual machines with different efficiencies and different power contracts, is what actually moves Bitcoin’s network hashrate up and down.

The Margin Equation: Hashprice Minus All-In Cost

Hashprice, a term coined by Luxor’s Hashrate Index, converts all of Bitcoin mining’s moving parts into a single daily number. The formula is roughly: about 144 blocks per day, multiplied by the block subsidy plus transaction fees, multiplied by BTC’s price, divided by the network’s total hashrate. The result is quoted in dollars per petahash per second per day (PH/s/day), and it functions as the closest thing the mining industry has to a single share price for the entire business.

The cost side is messier, because cost means different things depending on which line of a miner’s income statement is being read. Cash cost usually means just electricity plus hosting or pool fees, the money that actually leaves the bank account to keep machines spinning today. All-in cost adds everything else: SG&A, stock-based compensation, interest on mining-fleet debt, and, critically, depreciation and amortization of the ASICs themselves. A miner can be solidly profitable on a cash basis while reporting a large net loss once all-in costs are counted, and in the first quarter of 2026 that was closer to the industry norm than the exception. Riot Platforms’ own first-quarter 10-Q filing disclosed a cash cost of $44,629 per bitcoin, comfortably under that quarter’s roughly $75,964 production value per coin, but an all-in cost of $96,283 once depreciation was folded in, above the value of the bitcoin it actually mined.

Hashprice in Mid-2026: The Revenue Side of the Squeeze

Hashprice has had a rough year. It peaked around $63 per PH/s/day in July 2025, according to CoinShares’ Q1 2026 mining report, then fell to a five-year low near $35 to $37 by November 2025 as BTC’s price cooled off its October all-time high. A brief recovery into the high $30s over the following weeks did not last: by early 2026 hashprice was back down into the high $20s, and it has stayed roughly in a $28 to $30 band through the middle of the year, according to Hashrate Index data. That is a level last seen around the 2020 post-COVID crash, and it sits well below the level set immediately after Bitcoin’s October 2025 price peak.

The table below pulls together the key benchmarks referenced throughout this piece, dated to show how fast the picture has moved.

What makes this squeeze unusual is that it is not simply a story about BTC’s price falling. As HOGE Wire’s hashprice explainer laid out earlier this cycle, hashprice depends on three separate variables, price, block reward, and network hashrate, and the third one has stayed stubbornly high even while the first one whipsawed. BTC recovered to roughly $63,000 to $64,000 by early July 2026, up noticeably from its June low, yet hashprice barely moved, because the network’s total computing power did not fall nearly as much as price did. More machines chasing the same fixed reward means a thinner slice for everyone, regardless of what BTC is worth in dollar terms in any given week.

The Cost Stack: Electricity, Depreciation and What Actually Eats Margin

Electricity is, by a wide margin, the single largest cash cost for anyone mining at scale. Large US-listed miners typically operate on industrial power contracts priced in the low-to-mid single-digit cents per kilowatt-hour, and CoinShares’ research puts the breakeven power price for a mid-generation machine like Bitmain’s Antminer S19 XP at under 6 cents per kilowatt-hour just to stay profitable at a $30 per PH/s/day hashprice. Anyone paying retail grid rates, which in much of the world run several times that, has no realistic path to profitable mining with anything but the newest hardware, if at all.

Below electricity sits a longer list of costs that matter less day to day but dominate the accounting picture:

• Hosting or pool fees, paid to whoever operates the site or the pool a miner points hashrate at.
• SG&A and stock-based compensation, the normal overhead of running a public company.
• Interest on debt raised to buy fleets or build sites.
• Depreciation and amortization of the ASICs themselves, typically spread over a three-to-five-year useful life.

That last item is the one that most often turns a cash-flow-positive quarter into a reported net loss. A fleet bought during the 2024 to 2025 hardware upgrade cycle is still working through several more years of non-cash depreciation charges that show up in GAAP all-in cost figures even though no cash actually leaves the business for them.

There is a second, entirely separate reason public miners’ net income swings so violently: both MARA and Riot, like most listed peers, carry their bitcoin treasuries at fair value under current accounting rules, so a BTC price move flows straight through the income statement as an unrealized gain or loss. MARA’s roughly $1.3 billion net loss in the first quarter of 2026 was driven mostly by a non-cash markdown on its bitcoin holdings after that quarter’s price decline, not by mining operations turning unprofitable outright. Reading a miner’s quarterly net loss as a pure verdict on operating margin, without separating out the bitcoin-price effect, is one of the more common mistakes in coverage of this sector.

The ASIC Treadmill: Why Hardware Generation Is a Margin Variable

If price, block reward, and network hashrate are all outside any individual miner’s control, efficiency is the one lever they actually hold. A machine’s efficiency, measured in joules per terahash (J/TH), determines how much electricity it burns to produce a given amount of hashing power, and therefore how much of a miner’s revenue survives as margin before any other cost is counted. Bitmain’s current flagship, the Antminer S21 XP, runs at roughly 13.5 J/TH (270 TH/s at 3,645 watts), while the more widely deployed S21 Pro sits around 15 J/TH. Both are a significant step up from hardware that was considered cutting-edge just two or three years ago.

That efficiency gap translates directly into how much a miner can afford to pay for power and still break even. The table below is an illustrative calculation, not a directly reported industry figure; it is built from the standard hashprice formula at a $29.60 per PH/s/day hashprice (a recent Hashrate Index reading) and each machine’s published power draw, and it excludes hosting fees, pool fees and non-power overhead, so real-world breakeven power prices tend to run somewhat lower in practice.

The pattern holds up against real disclosures: CoinShares’ own research independently put the breakeven power price for S19 XP-class hardware at under 6 cents per kilowatt-hour at a $30 hashprice, almost exactly what the formula above produces. It also explains why older machines do not simply retire gracefully; they get switched off in waves whenever hashprice dips, then sometimes switched back on when it recovers, depending entirely on the power contract underneath them. JPMorgan’s Nikolaos Panigirtzoglou summarized the mechanism plainly in a client note reported by TFTC in June 2026: ‘When bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower.’

Pool Fees and Payout Schemes: The Small Print of Realized Margin

Almost nobody mines solo. With Bitcoin’s network hashrate sitting around 780 to 845 exahashes per second in early July 2026, according to CoinWarz, a single mid-sized rig has essentially no chance of finding a block on its own inside any reasonable timeframe, so the overwhelming majority of hashrate points at a small number of shared pools instead, smoothing out payouts in exchange for a fee.

Foundry’s own explainer lays out the main payout structures pools use:

• PPS (Pay Per Share): pays a fixed amount per submitted share regardless of whether the pool actually finds a block; the operator absorbs all the variance and charges the highest fee for it.
• FPPS (Full Pay Per Share): the default for most large farms; works like PPS but folds in an estimate of transaction fees.
• PPS+: pays the block-subsidy portion via PPS and the transaction-fee portion via PPLNS.
• PPLNS (Pay Per Last N Shares): pushes variance back onto the miner, who is paid only when the pool finds a block, in exchange for the lowest fee.

Fees across these models typically run from 0% to about 4%. At a healthy hashprice, a percentage point of pool fee barely registers. At $29 to $30 per PH/s/day, it is a meaningful share of an already thin margin, which is part of why the largest self-hosted miners increasingly either run their own pool (MARA operates one) or push toward Stratum V2’s Job Declaration feature, which lets a miner build its own block template and use the pool purely for payout smoothing rather than paying for full block construction. The framing is usually about decentralization and censorship resistance, but it has a margin dimension too: every function a miner can bring in-house is one less fee shaving revenue off the top.

Halving Math: Why the Squeeze Repeats Every Four Years

Bitcoin’s block subsidy has been 3.125 BTC since the April 2024 halving and is scheduled to fall to 1.5625 BTC around April 2028. That is a scheduled, overnight 50% cut to the industry’s single largest revenue line, with no corresponding cut to electricity bills, ASIC loan payments, or lease obligations. HOGE Wire’s halving cycle math explainer covers the price side of that cycle in detail; on the mining side, the effect is more mechanical and less debatable: the reward pool that hundreds of exahashes are competing over gets cut in half, and the only ways to absorb that are a higher BTC price, higher transaction-fee revenue, more efficient hardware, cheaper power, or some combination of all four.

Transaction fees have so far been a minor offset, typically under 5% of total miner revenue in most recent months, occasionally spiking higher during periods of network congestion but never coming close to replacing what a halved subsidy takes away. That is the crux of a warning Marathon CEO Fred Thiel gave in comments reported by CoinGeek: the expected transition from block-reward-driven revenue to fee-driven revenue ‘hasn’t happened,’ and unless Bitcoin’s fiat value grows at 50% or more annually, ‘the math gets very tough after 2028.’

Every previous halving has produced some version of this squeeze, and every previous halving has eventually resolved it through a mix of price appreciation and older hardware exiting the network. What is different in 2026 is that the industry is heading into its next halving already running historically thin margins, rather than fat ones, which leaves considerably less cushion if 2028 does not bring the kind of price rally earlier cycles did.

Who Is Underwater Right Now: Breakeven Hashprice and the 15% to 20% Figure

CoinShares’ Q1 2026 mining report put a number on the damage: roughly 15% to 20% of the global mining fleet was operating at a loss at prevailing prices, concentrated heavily among older, less efficient machines and operators without the cheapest power contracts. JPMorgan’s framing, reported by TFTC, put it in blunter terms still: BTC traded roughly 19% below the bank’s estimated $78,000 all-in production cost for about five consecutive months into mid-2026.

Fred Thiel has described the underlying dynamic about as directly as a public-company CEO is likely to. Bitcoin mining, in his words reported by CoinGeek, ‘is a zero-sum game. As more people add capacity, it gets harder for everybody else. Margins compress, and the floor is your energy cost.’ Unlike a normal industry, where a new competitor entering can, at least in theory, grow the overall market, a new miner plugging in additional hashrate does not grow Bitcoin’s fixed block-reward pool by a single satoshi; it just claims a larger slice of the same pie, thinning everyone else’s share along with its own eventual margin.

That zero-sum structure is also what makes the network mostly self-correcting. As underwater miners power down, total hashrate falls, difficulty adjusts downward at the next retarget, and the miners left standing suddenly control a larger share of the same fixed reward, which partially repairs margin without BTC’s price moving at all. That is exactly what has been happening through 2026: CoinShares documented three consecutive negative difficulty adjustments earlier this year, the first streak like that since July 2022, and CoinWarz data shows another retarget due on July 11, 2026, following what was already the second-largest single difficulty drop of the year.

MARA vs Riot: Two Approaches to the Same Margin Problem

The two largest US-listed miners, Marathon Digital (MARA) and Riot Platforms (RIOT), spent the first quarter of 2026 answering the margin question in overlapping but distinct ways, a contrast HOGE Wire covered in more depth in its Marathon vs Riot comparison. Both reported sizable GAAP net losses driven substantially by non-cash bitcoin fair-value markdowns rather than collapsing operating margins outright, and both leaned on their balance sheets, not just their hashrate, to fund a pivot toward AI and HPC infrastructure.

Riot’s own quarterly disclosure is the cleanest public illustration of the cash-cost-versus-all-in-cost gap described earlier in this piece: a $44,629 cash cost per bitcoin against $75,964 of production value, a healthy cash margin, but a $96,283 all-in cost once depreciation was included, equal to about 127% of that quarter’s production value, according to the company’s own 10-Q filing. Marathon’s route to shoring up its position was more balance-sheet-driven: the company sold over 20,800 BTC during the quarter, ending Q1 2026 with roughly 35,300 BTC still held, using proceeds to pay down debt and fund its buildout rather than relying on mining margin alone to carry the transition. CoinShares’ broader research also flagged Marathon’s per-BTC cost figures as an outlier among public miners, largely a function of depreciation policy applied to its unusually large fleet rather than a sign of worse underlying operations.

Neither company has fully proven out its AI pivot yet. Riot’s new data-center segment produced a modest but real $33.2 million of the quarter’s revenue, its first material non-mining line, backed by an expanded AMD hosting agreement and a molten-salt nuclear power deal aimed at supporting several more gigawatts of future capacity. Marathon’s roughly 2.2 gigawatts of power, after its Long Ridge Energy gas-plant purchase, is larger in absolute terms, but as of this writing it has yet to announce an anchor hyperscaler lease at the scale of Riot’s, Core Scientific’s, or Hut 8’s, which remains the single biggest open question in its own margin story.

The AI/HPC Pivot as a Margin Survival Strategy

The logic behind the industry-wide pivot toward AI and HPC hosting is straightforward once mining margin is understood the way this piece has framed it: GPU colocation revenue is contracted, typically a fixed monthly payment from a hyperscaler or AI-cloud tenant, rather than exposed to a volatile, protocol-set commodity price that resets every 210,000 blocks. Needham & Co analyst John Todaro put the comparison plainly in remarks reported by CoinGeek: ‘the revenue per megawatt and EBITDA margins are far higher for HPC and AI colocation than for mining.’

The scale of the shift by mid-2026 is substantial. CoinShares tallied more than $70 billion in cumulative AI and HPC contracts announced across the public mining sector, anchored by deals like Core Scientific’s roughly $10.2 billion, 12-year colocation agreement, Hut 8’s roughly $7 billion, 15-year, 245-megawatt lease with Fluidstack, and TeraWulf’s roughly $12.8 billion in total contracted HPC revenue. CoinShares’ James Butterfill, quoted by The Block, estimated listed miners could derive ‘as much as 70% of their revenues from AI by the end of this year, up from roughly 30% today.’

None of this comes free. Funding gigawatt-scale data-center buildouts has pushed several former pure-play miners deep into debt: IREN carries roughly $3.7 billion in convertible notes, TeraWulf around $5.7 billion in total debt, Cipher about $1.7 billion in senior secured notes at a 7.125% coupon. The bet only pays off if these companies can actually sign anchor tenants at scale and deliver power on schedule; a miner that raises the capital and builds the shell without landing that anchor lease ends up carrying data-center-grade debt against mining-grade margins, arguably a worse position than the one it started in.

Geography and Power Contracts: Why Location Is Margin

Because electricity is the dominant marginal cost and its price varies by an order of magnitude depending on where a machine is plugged in, geography is arguably the single biggest structural margin decision a mining company makes, ahead of even which ASIC generation it buys. Stranded or curtailed power, flared associated gas, and oversupplied renewable grids can bring contracted rates down to a few cents per kilowatt-hour or less, while retail grid power in much of the world runs several multiples higher, enough on its own to erase margin regardless of hardware efficiency.

That is why West Texas, inside the ERCOT grid, has become the default address for large-scale US mining, including Riot’s Corsicana and Rockdale sites and several of Marathon’s facilities: abundant wind and solar generation regularly oversupplies the grid, pushing wholesale prices toward zero or negative at times, and ERCOT pays large flexible loads, mining fleets among them, to curtail consumption during periods of peak system stress. For a miner, that curtailment payment functions less like a power saving and more like a second, smaller revenue line layered on top of hashprice, one more lever that has nothing to do with BTC’s price at all.

The concentration this creates is also a risk in its own right. A grid-stress event, a change in state-level incentive policy, or local political pushback over water and power use in a single region can move the needle for a meaningful share of the entire US mining industry at once, which is part of why several large miners have been diversifying into new sites, and increasingly new countries, rather than continuing to stack capacity in a single grid.

Hedging Margin: Derivatives, Forward Curves and What the Market Expects

A market this exposed to two volatile, largely uncontrollable inputs eventually grows financial instruments to hedge them, and hashprice is no exception. Luxor’s hashprice derivatives desk offers forwards running out as far as twelve months, six-month daily-settlement contracts, and hashrate futures, letting a miner lock in a fixed hashprice for a stretch of time rather than stay fully exposed to whatever the spot market does, conceptually similar to an airline hedging jet fuel.

What the forward curve has been pricing is itself informative. A May 2026 snapshot of Luxor’s forward market priced hashprice at roughly $28.94 per PH/s/day (quoted as 0.00047 BTC per PH/s/day) running through November 2026, implying a miner’s cost of capital in the 6% to 13% annualized range baked into those contracts. In plain terms, the market was not pricing in a snapback anywhere close to 2025’s levels; it was pricing in more of the same. That tempers even CoinShares’ Butterfill’s own more optimistic scenario, that a BTC recovery to $100,000 could lift hashprice back to roughly $37 per PH/s/day, since forward pricing suggests professional counterparties were not betting heavily on that outcome arriving within the contract window.

This market is still small and comparatively illiquid next to established commodity hedging in oil or power, so most miners, particularly smaller and mid-sized ones without a treasury desk, remain largely unhedged and fully exposed to whatever spot hashprice does week to week. That is a meaningful part of why the 15% to 20% unprofitable-fleet figure skews toward smaller operators and older hardware rather than the largest public names, which have both better access to hedging tools and the balance-sheet flexibility to ride out a bad quarter.

The Regulatory Backdrop: Why Mining Is Not a Security, and Why That Matters for Margin

One tail-risk that is not currently eating into mining margins, at least in the United States, is securities-registration overhang. The SEC’s Division of Corporation Finance stated in a March 2025 staff statement, covered by The Block, that proof-of-work mining, whether solo or through a pool, does not implicate US securities law, reasoning that rewards flow from a miner’s own computational contribution rather than from a third party’s managerial effort, the standard the Howey test requires for something to count as a security. A broader interpretive release in March 2026 extended a similar no-securities view to staking, wrapping, and no-consideration airdrops, part of the same deregulatory push HOGE Wire tracked in its explainer on how SEC crypto enforcement actually works.

For a margin-squeezed industry, that guidance matters less as an active tailwind than as an absent headwind: it removes the possibility of registration costs, enforcement actions, or compliance overhead becoming yet another line item stacked on top of already-thin operating margins. It is worth stressing that this guidance sits at the staff level, is non-binding on courts, and could in principle be revisited by a future commission, so miners are treating it as a currently favorable environment rather than a permanent settlement. Crypto derivatives, including hashrate and hashprice futures, remain under the CFTC’s jurisdiction rather than the SEC’s, a separate and generally more settled regulatory lane.

It is also worth noting how differently this margin conversation reads on the proof-of-stake side of crypto. A liquid-staking validator’s economics turn almost entirely on commission rates and slashing risk, not on an electricity bill or a depreciating fleet of physical hardware; HOGE Wire’s explainer on how stETH and rETH actually work is a useful point of comparison for anyone trying to understand why Bitcoin mining behaves so differently, economically, from validating a proof-of-stake chain.

Outlook: Margins Before and After the 2028 Halving

The near-term path for margin depends overwhelmingly on BTC’s price, since network hashrate has proven far stickier than most analysts expected through 2025 and 2026. CoinShares’ Butterfill offered one useful anchor: a recovery to $100,000 would plausibly lift hashprice back to around $37 per PH/s/day, an improvement, but still well under 2025’s peak, which says something about how much more computing power is now permanently competing for the same reward pool compared with a year ago.

The medium-term story is the AI and HPC pivot, and whether it genuinely decouples a growing share of miner revenue from hashprice, or simply saddles a subset of the industry with data-center-scale debt against contracts that take years to fully ramp. By the time the next halving arrives around April 2028, cutting the block subsidy again to 1.5625 BTC, any miner still overwhelmingly dependent on block rewards alone will face the same trilemma the industry is already navigating: get more efficient, get cheaper power, or get acquired. Thiel put the choice as bluntly as any public-company CEO has, in comments reported by CoinGeek: ‘By 2028, you’ll either be a power generator, be owned by one, or be partnered with one. The days of being a miner plugged into the grid are numbered.’

What ties all of this together is that margin, not price alone and not headline hashrate records, is the number that will determine whether Bitcoin mining consolidates into a small number of vertically integrated energy-and-compute conglomerates over the next few years, or remains the more fragmented, geographically dispersed industry it has been for most of its history. Every lever covered in this piece, hardware efficiency, power contracts, pool fees, hedging, and now AI diversification, is ultimately a way of defending that one number against a protocol that periodically, predictably, cuts the reward in half regardless of what any individual miner would prefer.

Frequently Asked Questions

What is a good profit margin in Bitcoin mining?

There is no single industry-standard percentage, because miners do not report a uniform margin metric. The more useful comparison is hashprice against a miner’s own all-in cost per PH/s/day. With hashprice running around $29 to $30 per PH/s/day in mid-2026 and the industry-wide weighted-average cash cost near $80,000 per BTC, per CoinShares, an operation generally needs newest-generation hardware (around 13.5 to 15 J/TH) and power under roughly 6 cents per kilowatt-hour to hold a real margin rather than just breaking even.

Why are Bitcoin mining margins so thin in 2026?

Two forces hit at once. BTC’s price pulled back sharply from its October 2025 all-time high, while network hashrate and difficulty stayed structurally elevated as more efficient machines kept coming online even through the price decline. That means the same fixed block reward is now split across far more computing power than during 2025’s rally, pushing hashprice down to a five-year low near $28 to $30 per PH/s/day, according to CoinShares.

How much does it cost to mine one Bitcoin?

It depends on which cost is being measured. CoinShares put the weighted-average cash cost across publicly listed miners at about $79,995 per BTC in the fourth quarter of 2025, while JPMorgan estimated a roughly $78,000 all-in cost industry-wide in mid-2026. Individual miners vary widely: Riot Platforms disclosed a first-quarter 2026 cash cost of $44,629 per bitcoin excluding depreciation, but $96,283 once depreciation was included, more than that quarter’s production value per coin.

What happens to mining margins after the 2028 halving?

Bitcoin’s block subsidy is scheduled to fall from 3.125 BTC to 1.5625 BTC around April 2028, an overnight 50% cut to the industry’s largest revenue line with no matching cut to electricity or hardware costs. Unless BTC’s price or transaction-fee revenue rises enough to offset that, margins compress further. Marathon CEO Fred Thiel has said the math ‘gets very tough after 2028’ without sustained annual price growth of 50% or more.

Why are Bitcoin miners building AI data centers instead of just mining?

Because AI and HPC colocation revenue is contracted, typically a fixed monthly payment from a hyperscaler tenant, rather than exposed to a volatile, protocol-set commodity price. Reported EBITDA margins on that business also tend to run higher than on pure mining. CoinShares estimates listed miners could derive as much as 70% of revenue from AI by the end of 2026, up from roughly 30% in mid-2026, though the pivot requires heavy borrowing and landing anchor tenants at real scale.

Written by the HOGE Wire mining and staking desk.