How to Read a Bitcoin Miner’s Margins: A 2026 Field Guide

A Bitcoin miner, stripped to its mechanics, is a business that buys electricity and sells a commodity it cannot price. In August 2026 that business is hard. Bitcoin trades around $64,000, roughly 48% below its October 2025 record, while the revenue a machine earns per unit of computing power sits near $31.73 per petahash per day, close to a multi-year low. Between 15% and 20% of the global fleet is mining at a loss. When conditions are this tight, the gap between a company that survives and one that quietly turns into a zombie is measured in a single figure: margin.

The trouble is that margin in Bitcoin mining is not one number. It is at least three, and public miners have every incentive to lead with the friendliest one. This guide is about reading past the headline. It walks through how a miner actually earns and spends money, which disclosures matter, where the accounting hides the truth, and how to tell a durable operator from a story stock before you commit a dollar. The figures are current as of August 2026, but the method outlives any single print.

Why Margins Are the Only Number That Matters in 2026

For most of Bitcoin’s history, mining was a high-margin business that forgave sloppy operators. When the price of the coin rose faster than the cost of producing it, almost everyone made money, and the difference between a well-run fleet and a careless one was academic. That era is over. The April 2024 halving cut the block subsidy to 3.125 BTC, network competition kept climbing, and the coin’s price stopped cooperating. CoinShares, in its Q1 2026 mining report, put the weighted-average cash cost to produce one Bitcoin among listed miners at about $79,995 in the fourth quarter of 2025, above where the coin has traded for stretches of 2026. When the average producer’s cash cost sits near or above the market price, margin stops being a footnote and becomes the whole story.

That is why every serious miner now leads its quarterly report with a cost-to-mine figure, and why reading that figure correctly matters more than it used to. Bitcoin’s price is a macro variable that no miner controls; it swings on the Fed, on jobs data, and on flows, and it has spent August 2026 stuck in the mid-$60,000s. What a miner does control is its cost structure, and the disclosures around that structure are where the real information lives. Treat a mining stock’s quarterly report as a stress test, and treat this guide as a tour of the gauges.

Margin, Defined: Revenue Minus the Real Cost of a Coin

Start with the equation everyone agrees on. A miner’s margin is the value of the Bitcoin it produces minus what it costs to produce them. The disagreement, and the opportunity for confusion, is entirely in the second term. There are at least three versions of cost, and they can differ by a factor of two.

The first is direct cost of revenue, dominated by electricity. Subtract it from mining revenue and you get a gross or mining margin, the number that looks best on a slide. The second is cash cost, which adds the other cash a mine burns to keep running: hosting fees, salaries, maintenance, and general overhead, but not the non-cash charge for wearing out the machines. Cash cost tells you whether the operation generates money today. The third is all-in cost, which adds depreciation of the ASIC fleet and data-center buildout, stock-based compensation, and financing costs. All-in cost approximates the GAAP picture and answers a harder question: is the business creating value once you account for the capital it consumes.

A mine can be gross-profitable, cash-profitable, and all-in-unprofitable at the same time, in the same quarter. Retail investors who see the word ‘profitable’ in a headline and stop reading are usually looking at one of the first two. The professionals are looking at the third, and so should you.

To see how far apart these can sit, imagine a mine that earns $60,000 of Bitcoin per coin. If electricity runs $30,000 of that, the gross margin looks like a healthy $30,000. Layer in hosting, staff, and overhead of another $12,000 and the cash margin is still a positive $18,000. Now charge $25,000 of depreciation on the machines and buildings, plus a little interest, and the all-in result flips to a loss. Same coin, same quarter, three defensible numbers, one of them negative. The company will quote whichever supports its story; your job is to know which one you are looking at.

The Revenue Line: Hashprice Is a Miner’s Unit Price

Every margin starts with revenue, and in mining, revenue per unit of work has a name: hashprice. Coined by Luxor’s Hashrate Index, hashprice is the expected daily revenue for one unit of hashrate, quoted in US dollars per petahash per day. It compresses four moving parts into a single price: the block subsidy of 3.125 BTC, transaction fees, the network’s difficulty, and the dollar price of Bitcoin. In plain terms, it is the number of coins one unit of hashrate mines per day multiplied by the coin’s price.

Think of hashprice as the wholesale price a miner receives for its product. It is the top line of every margin calculation, and it is brutally exogenous: a single operator has no influence over it. As of the 10 August 2026 Hashrate Index roundup, USD hashprice was $31.73 per petahash per day, down about 1% on the week and worth 0.00049633 BTC. That is roughly a third of the level miners enjoyed during the October 2025 peak. Transaction fees are a rounding error right now, about 0.74% of the block reward, so revenue is almost entirely subsidy multiplied by price.

Two forces move hashprice in opposite directions. A rising Bitcoin price lifts it; rising network difficulty, meaning more competitors chasing the same fixed 3.125-BTC reward, drags it down. Because the coin’s price is the dominant swing factor, a miner’s revenue line is effectively a leveraged bet on Bitcoin, which is why the sell-side spends so much energy on price targets. Understand that dependence, and half the volatility in a miner’s results stops being a surprise.

The Cost Stack: Where a Miner’s Dollars Actually Go

Once you know the revenue line, the game is the cost stack. For a self-mining operator, the components in rough order of size are electricity, hosting or colocation, hardware depreciation, overhead, and financing.

Electricity is the dominant variable cost and the single most important number a miner discloses, usually expressed as an all-in power price in dollars per kilowatt-hour. The spread between operators is enormous. CoinShares pegged IREN’s electricity cost at about $34,325 per Bitcoin in the fourth quarter of 2025, the lowest among listed miners, while less-advantaged fleets paid multiples of that. Hosting and colocation matter for miners that do not own their sites: they pay a third party per kilowatt-hour or per megawatt, trading capital efficiency for a thinner, more predictable margin. Hardware depreciation is the non-cash charge for wearing out ASICs, which have a useful life of only a few years; it is exactly the item that vanishes from cash-cost figures and reappears, often brutally, in all-in cost. Overhead covers salaries, stock-based compensation, and corporate costs, and financing covers the interest on the convertible notes and secured debt many miners used to fund their buildouts.

The mix matters as much as the total. A self-mining operator that owns its substations and land carries heavy depreciation and interest but controls its own power price; a miner that hosts its machines in someone else’s facility swaps that capital burden for a bundled hosting fee, which smooths the margin but caps the upside. When you compare two miners, first work out which model each one runs, because the same headline cost-to-mine figure can hide very different balance sheets underneath.

Fred Thiel, chief executive of MARA Holdings, framed the stakes bluntly to CoinGeek. Mining, he said, “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.” That last clause is the whole thesis of cost-side analysis. When revenue is fixed by the network and you cannot control it, your power price is the one lever that decides whether you live.

Cash Cost vs All-In Cost: The Distinction That Fools Retail

Here is the distinction that separates careful readers from the rest, made concrete with one company’s real numbers.

In the first quarter of 2026, Riot Platforms reported that its cost to mine a single Bitcoin, excluding depreciation, was $44,629, per its quarterly results. Against a production value of roughly $75,964 per coin, that looks like a fat margin: the operation threw off more than $31,000 of cash per Bitcoin. On a cash basis, Riot was comfortably profitable.

Now add depreciation back. Riot’s Form 10-Q shows the all-in cost to mine one Bitcoin, including depreciation, at $96,283, or 126.7% of the coin’s production value. On that basis, the same company, in the same quarter, lost about $20,000 on every coin it mined. Cash-positive and GAAP-negative, simultaneously.

Neither number is a lie. They answer different questions. The cash figure tells you Riot’s mines are not bleeding money to keep the lights on, which is genuinely important, because a cash-positive operation can wait out a downturn. The all-in figure tells you that at current prices, the capital sunk into machines and buildings is not earning its keep, and that if Bitcoin does not rise, the depreciation will eventually show up as real replacement cost. A durable analysis holds both in view. A headline holds one.

Cost to Mine One Bitcoin: How to Read the Headline Number

‘Cost to mine one Bitcoin’ is the industry’s favorite metric and one of its least standardized. It is not a GAAP measure, there is no regulator-blessed definition, and each company computes it slightly differently. Treat it as a useful disclosure, not a comparable one.

Three things distort it. First, depreciation policy. Two miners with identical electricity costs can report wildly different all-in costs per coin purely because one depreciates its ASICs over two years and the other over four, or because one carries a much larger or newer fleet. MARA’s per-coin cost has often looked like an outlier for exactly this reason, a function of the size and accounting treatment of its fleet rather than necessarily worse operations. CoinShares put TeraWulf’s all-in cost at an eye-watering $471,841 per Bitcoin in late 2025, a figure distorted by its transition rather than a reflection of a broken mine.

Second, what gets netted out. Riot’s cost-to-mine figure is reported after subtracting the power credits it earns for curtailing during grid stress, which is legitimate but means the number blends mining efficiency with a separate, lumpy revenue stream. Third, whether the figure includes depreciation at all. When a miner leads with a cost-to-mine number and buries the phrase ‘excluding depreciation,’ that is a tell. The cash number is real, but presenting it as the cost of the business, with no all-in figure nearby, is marketing.

The practical rule: never compare two miners’ cost-per-coin figures without first checking that both include or both exclude depreciation, and that both handle power credits the same way. More often than not, they do not.

The Hashrate Trap: Energized vs Operating vs Realized

Hashrate is the headline operational metric, and it comes in flavors that are easy to conflate and easy to game.

MARA reported a record 72.2 exahash per second of “energized” hashrate in the first quarter of 2026, up 33% year over year, per its earnings materials. Energized means plugged in and capable of running. It is not the same as hashrate that is actually turned on and hashing, which is why MARA also reports what it produced: 2,247 Bitcoin, about 25 a day, or roughly 5.5% of network rewards. A fleet can be fully energized and partly idle, whether for curtailment, maintenance, or heat, so energized hashrate flatters the apparent scale of an operation relative to what it earns.

Riot draws the distinction more explicitly, separating deployed hashrate from operating hashrate. The gap between the two is a real efficiency signal: it tells you how much of the fleet a miner is actually running. Why does this matter for margin? If you divide revenue by energized hashrate, you get a misleadingly low revenue-per-unit figure and overstate the operator’s headroom. Always sanity-check a hashrate claim against realized production. The Bitcoin a miner actually mined, divided by the network’s total, is the only version of hashrate that pays the bills.

The Bitcoin on the Balance Sheet: Fair Value Runs Through Earnings

There is a second reason a miner’s reported earnings can swing wildly, and it has nothing to do with mining. Under current accounting, public miners carry their Bitcoin holdings at fair value, and changes in that value flow straight through the income statement each quarter.

The effect is dramatic. MARA reported a net loss of about $1.3 billion in the first quarter of 2026, driven not by operations but by unrealized mark-to-market adjustments on its digital-asset holdings as Bitcoin’s price fell, per its results. Riot’s quarterly losses have similarly been dominated by Bitcoin markdowns rather than by mining performance. For a reader, this means a miner’s headline net income is partly a leveraged position on the coin’s price and partly an operating business, blended into one number. To isolate the operating story, look past net income to operating cash flow, to adjusted metrics that strip out the fair-value swing, and to the cost-to-mine disclosures.

The balance sheet also reveals strategy. A miner that holds every coin it mines is making a directional bet on the price. A miner that sells is funding something. MARA sold roughly $1.5 billion of Bitcoin through mid-2026 to fund its shift into AI infrastructure, per CoinDesk, while retaining a large treasury. Whether that is prudent or short-sighted depends on your view of the coin, but it is a decision the balance sheet makes visible, and one worth weighing before you assume a miner’s treasury is a passive store of value.

Power Credits and Curtailment: The Line That Can Flip a Quarter

One line item deserves special attention because it can swing a miner’s cost-to-mine by thousands of dollars per coin: power credits.

In deregulated grids like the one in Texas, large flexible loads can earn money by agreeing to curtail consumption during demand spikes. A Bitcoin mine is an almost perfect participant, because the only cost of switching off is the hashrate it forgoes; there is no half-finished product to ruin and no restart penalty. Riot has built this into its model. It reported that power curtailment credits rose 169% year over year in the first quarter of 2026, a swing large enough to materially lower its reported cost to mine a coin, per its quarterly results.

For an investor, credits are a genuine and underappreciated margin lever, but they carry two caveats. They are seasonal and weather-dependent, concentrated in the summer months when grids strain, so a strong credit quarter does not annualize. And when a miner nets them against cost-to-mine, the resulting figure mixes operational efficiency with grid-services revenue. Read the credit line separately, then decide how much of it you believe repeats.

Build the Table Yourself: Efficiency, Power Price, and Breakeven

You do not need a miner’s investor deck to estimate whether its fleet makes money. With two inputs, machine efficiency and power price, you can build the table yourself.

Efficiency is measured in joules per terahash (J/TH); lower is better. The current generation sets the frontier. Bitmain’s Antminer S21 XP runs at about 13.5 J/TH; the S21 Pro at 15 J/TH; MicroBT’s WhatsMiner M60S at about 18.5 J/TH; and the previous-generation S19 XP at 21.5 J/TH. A machine’s efficiency, combined with hashprice, dictates how much revenue it earns per unit of electricity. Hashrate Index publishes exactly this breakdown. In the week to 10 August 2026, at a hashprice near $31.73, the most efficient fleets earned about $112 per megawatt-hour of electricity consumed, while the oldest machines still running earned about $42, per its roundup.

Flip that around and you get breakeven power price: the electricity cost above which a machine loses money. The math is simple. Breakeven dollars per kilowatt-hour equals hashprice divided by 24 times efficiency in J/TH. At today’s hashprice, a top-tier 13.5 J/TH machine breaks even near ten cents per kilowatt-hour, while a legacy 30-plus J/TH rig needs power below four and a half cents just to avoid losing money on electricity alone, before any other cost.

This is the single most useful back-of-envelope in mining. If you know a miner’s fleet mix and its disclosed power price, you can place it on this table and know, roughly, whether it is above or below water, without taking management’s word for it.

Reading the AI Pivot Without Getting Sold a Story

The biggest story in mining margins in 2026 is not mining. It is the rush to repurpose power and land into AI and high-performance-computing data centers, where the economics are far better. CoinShares counts more than $70 billion in cumulative AI and HPC contracts announced across listed miners, and its head of research, James Butterfill, told The Block that listed miners could earn as much as 70% of their revenue from AI by the end of the year, up from roughly 30% today.

The appeal is straightforward margin. As Needham analyst John Todaro put it to CoinGeek, “the revenue per megawatt and EBITDA margins are far higher for HPC and AI colocation than for mining.” A megawatt leased to an AI tenant on a multi-year contract earns more, and more predictably, than the same megawatt pointed at the Bitcoin network. Riot has started reporting it: its first data-center revenue, $33.2 million in the first quarter of 2026, prompted chief executive Jason Les to call the quarter “a definitive inflection point for Riot, as we officially transitioned into an active, revenue-generating data center operator.”

Here is where a margin reader has to be skeptical, because the pivot is also where the vapor is. A press release announcing a ten-billion-dollar contract is not ten billion dollars of revenue; it is a headline number spread over ten or fifteen years, contingent on buildout, financing, and a tenant honoring the lease. When you read an AI-pivot claim, ask four questions. Is the contract signed, or a letter of intent? Is there an anchor tenant named, ideally a creditworthy hyperscaler? Is there a prepayment that de-risks the buildout? And is the capital to build it already raised, or does the miner need to sell Bitcoin or issue equity to fund it, as many have? Miners are, as CoinDesk put it, becoming AI companies and selling their Bitcoin to fund the transition. The ones that can point to signed, funded, anchored contracts are describing a business. The ones with only announcements are describing a hope.

The Self-Correcting Floor: Why Margins Do Not Stay at Zero

If mining margins are this bad, why does the industry not simply collapse? The answer is a mechanism built into Bitcoin itself, and understanding it keeps a margin reader from panicking at every downturn.

Bitcoin’s difficulty adjusts roughly every two weeks to keep blocks arriving about every ten minutes. When the coin’s price falls below the marginal producer’s cost, the highest-cost machines power down, hashrate drops, and at the next adjustment difficulty falls, which raises hashprice for everyone still running. JPMorgan’s Nikolaos Panigirtzoglou described the loop precisely in a client note reported by TFTC: “When bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower.” JPMorgan put all-in production cost near $78,000 with the coin trading below that for months, exactly the condition that triggers the cull.

None of this is new. The downturns of 2018 and 2022 ran the same script, forcing a wave of overleveraged, high-cost miners into bankruptcy while lower-cost operators bought their machines cheaply and emerged larger. Every cycle, the squeeze transfers hashrate from the weak to the strong. What is different in 2026 is the exit that did not exist before: instead of only failing or being acquired, a struggling miner can now try to convert its power into an AI data center, which is why this shakeout looks less like a clean cull and more like a sorting into two very different businesses.

You can watch it happen in the data. Network difficulty was set to fall about 3% at its next adjustment, estimated around 23 August 2026, per Hashrate Index, a sign that marginal capacity is switching off and the floor is doing its work. This is why the weakest hands, not the industry, get culled in a squeeze; the question for any single company is which side of the marginal-cost line it sits on, a dynamic familiar from other corners of crypto, where some operations survive stress and others do not. The self-correcting floor protects the network. It does not protect the highest-cost miner, and it will not protect a shareholder who bought one.

A Margin-Reading Checklist: Green Flags and Red Flags

Pull it together into something you can use. When you open a miner’s quarterly report or investor deck, here is what separates signal from spin. Start with the green flags, the marks of an operator built to survive a bad year.

• A disclosed all-in power price, ideally under six cents per kilowatt-hour.
• An all-in cost to mine that includes depreciation, presented next to the cash figure, not instead of it.
• A modern, efficient fleet, mostly sub-16 J/TH, that sits comfortably on the right side of the breakeven table.
• AI or HPC contracts that are signed, funded, and anchored by a named, creditworthy tenant with a prepayment.
• Operating cash flow that is positive on its own, before any Bitcoin fair-value gains.
• Honest hashrate reporting that reconciles deployed capacity with coins actually produced.

Now the red flags, the tells that a company is leaning on its friendliest metric.

• A cost-to-mine number quoted only excluding depreciation, with the all-in figure hard to find.
• Reliance on energized hashrate to imply a scale the production numbers do not support.
• AI announcements that are letters of intent, unfunded, or missing a named tenant.
• Liquidity that depends on selling Bitcoin or issuing dilutive equity every quarter to stay afloat.
• Power credits presented as recurring when they are seasonal.
• A fleet skewed toward machines above 25 J/TH at a power price above five cents.

None of these is disqualifying on its own, and context matters; a miner mid-transition may carry a distorted all-in cost for good reasons. But the pattern is what counts. A company that discloses cleanly, funds its buildout, and sits on the right side of the cost curve is telling you it can survive a bad year. A company that leans on the friendliest metric in every paragraph is telling you something too.

What the SEC Regulates, and What It Leaves to You

A word on the rules, because they shape what you can and cannot rely on. In March 2025 the SEC’s Division of Corporation Finance issued a staff statement clarifying that proof-of-work mining on a public network is not, by itself, the offer or sale of a security, and that mining in a pool does not change that. The activity of mining, in other words, sits largely outside the securities regime.

The companies are not. MARA, Riot, and their listed peers are SEC registrants, which is precisely why you have 10-Qs, 8-Ks, and audited figures to read in the first place. The cost-to-mine disclosures this guide dissects exist because these firms answer to public-company reporting rules, even though the specific ‘cost to mine one Bitcoin’ metric is a non-GAAP measure the SEC does not define. That gap, real filings wrapped around an unstandardized headline number, is exactly why margin literacy falls to the reader. For the broader enforcement backdrop, see our field guide to how SEC crypto enforcement works in 2026.

Macro policy matters too, indirectly but powerfully, because it moves the Bitcoin price that sits atop every margin. A miner’s fortunes in the back half of 2026 will turn as much on the September FOMC as on any chip upgrade. You cannot model that away. You can only know where a company sits on the cost curve when it arrives.

Frequently Asked Questions

What is a good cost to mine one Bitcoin in 2026?

It depends on whether depreciation is included. On a cash basis, low-cost operators mine for well under $50,000 per coin, while the industry’s weighted-average cash cost was about $79,995 in late 2025, per CoinShares. Including depreciation, all-in costs run far higher and sometimes exceed the coin’s market price, which is why cash-profitable miners can still post GAAP losses.

Why do Bitcoin miners lose money even when they say they are profitable?

Because ‘profitable’ usually refers to cash or gross margin, which excludes the non-cash depreciation of machines and facilities. A miner can generate positive operating cash while its all-in cost per coin exceeds the price. Add the mark-to-market swings on Bitcoin held on the balance sheet, and reported net income can be deeply negative in a quarter the mines ran normally.

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

Hashprice is the daily revenue one unit of hashrate earns, quoted in dollars per petahash per day. It bundles the block subsidy, transaction fees, difficulty, and the Bitcoin price into the top line of every margin. Near $31.73 in August 2026, it sat close to multi-year lows, meaning only efficient fleets with cheap power kept a margin.

How do I calculate a mining machine’s breakeven electricity price?

Divide the hashprice by 24 times the machine’s efficiency in joules per terahash. At a hashprice near $31.73, a 13.5 J/TH machine breaks even around ten cents per kilowatt-hour on electricity alone, while older machines need far cheaper power. Remember this is electricity-only, so other costs push the true breakeven lower.

Is the AI pivot actually improving Bitcoin miner margins?

For some operators, yes: leasing power and space to AI tenants earns more per megawatt than mining, and analysts expect AI to supply a majority of listed miners’ revenue by late 2026. But announced contract values are spread over many years and are often unfunded, so the benefit is real mainly where the deal is signed, financed, and anchored by a creditworthy tenant.

Yuki Tanaka covers Bitcoin mining, energy markets, and the economics of proof-of-work for HOGE Wire.