Bitcoin Mining Economics — Miner Treasury Guide

The unit economics of Bitcoin mining and what they mean for miner treasuries: halvings, hashprice, fleet efficiency, HODL-versus-sell strategies and the AI data-center pivot, with sources cited.

How Bitcoin mining revenue works

Bitcoin miners earn revenue from exactly two streams, and both are paid in bitcoin. The first is the block subsidy: newly issued BTC that the protocol grants to whichever miner finds a valid block. Since the April 2024 halving, the subsidy is 3.125 BTC per block, and with a block arriving roughly every ten minutes, the network issues about 450 new BTC per day. The second stream is transaction fees: the fees attached by users to the transactions included in each block.

In mid-2026, fees remain a small fraction of the total. The weekly average fee per block was about 0.0249 BTC in early August 2026, per the Hashrate Index roundup of August 3, 2026 — under 1 percent of the 3.125 BTC subsidy. The Block reported that fee revenue fell to a 12-month low earlier in 2026, underscoring how dependent miner economics still are on the shrinking subsidy. The Block's fee-share chart tracks this ratio monthly; outside of episodic spikes (Ordinals inscriptions in 2023, the Runes launch at the 2024 halving), fees have typically contributed low single-digit percentages of miner revenue since the last halving.

This matters because the subsidy is not permanent. Every 210,000 blocks — roughly every four years — it halves. Each halving mechanically cuts the industry's dominant revenue stream in half overnight, and the industry's long-term viability depends on some combination of a higher bitcoin price, lower costs, and eventually a deeper fee market picking up the slack.

Hashprice: the industry's single most useful number

Hashprice is the expected revenue a miner earns per unit of hashrate per day, usually quoted in dollars per petahash per second per day (USD/PH/s/day). The metric was popularized by Luxor's Hashrate Index, and it compresses the four forces that drive mining revenue — bitcoin's price, network difficulty, the block subsidy, and transaction fees — into one number. When bitcoin's price rises or difficulty falls, hashprice goes up; when difficulty rises or the subsidy halves, it goes down.

As of August 3, 2026, USD hashprice stood at about $32.10 per PH/s per day, per the Hashrate Index — a level the same report describes as "at or below breakeven for many miners depending on operating cost and machine model type." For context, hashprice spent much of the 2021 bull market in the hundreds of dollars per PH/s/day; the 2026 level reflects a bitcoin price around $63,000–65,000 (roughly 48 percent below its October 6, 2025 all-time high of $126,080, per Fortune's August 7, 2026 price update) combined with a network that is still enormous by historical standards. The Block also maintains a public hashprice index chart.

A useful rule of thumb follows directly from the definition: a miner is profitable when its all-in cost to produce a petahash-day of work is below hashprice. That cost is dominated by electricity, which is why the same hashprice can mean healthy margins for one operator and losses for another.

Hashrate and difficulty in mid-2026: a rare contraction

Network hashrate — the total computational power securing Bitcoin — peaked above 1 zettahash per second (1,000 EH/s) in late 2025 and has since pulled back. As of early August 2026, the 7-day average sat around 932 EH/s per the Hashrate Index, with CoinWarz showing short-window estimates fluctuating between roughly 920 and 1,000 EH/s depending on the sampling period. (Hashrate is inferred from block times, so short-window readings are noisy; multi-day averages are more meaningful.)

Difficulty — the protocol's self-adjusting throttle that retargets every 2,016 blocks (about two weeks) to keep block times near ten minutes — tells the cleaner story. Difficulty stood at 126.23T after a -0.74 percent adjustment on July 25, 2026 (Hashrate Index), roughly 20 percent below its November 2025 peak near 156T, per reporting collected by ChinaTechNews and CryptoPotato. Notably, Blockspace, citing Hashrate Index data, reported that 2026 marks only the second time in Bitcoin's history that difficulty has fallen year over year — the other being the 2021 China mining ban.

The drivers are compound: a lower bitcoin price, post-halving revenue compression, rising power costs in key hubs, curtailment events in Texas, regional disruptions, and — uniquely to this cycle — miners voluntarily redirecting electricity and data-center capacity from bitcoin mining to AI computing. Falling difficulty is the network's shock absorber: as unprofitable machines switch off, the remaining miners earn more BTC per unit of hashrate, which is why BTC-denominated hashprice has actually ticked up even as USD hashprice sits near cycle lows.

Energy costs: what separates profitable miners from the rest

Electricity is the overwhelming majority of a miner's marginal cost, so profitability reduces to two variables: the fleet's efficiency (how many joules it burns per terahash of work) and the power price (dollars per megawatt-hour). The Hashrate Index breakeven table for early August 2026 illustrates the spread starkly: at prevailing hashprice, fleets more efficient than 14 J/TH could tolerate power costs up to about $109/MWh, while older fleets in the 25–38 J/TH range broke even only below roughly $41/MWh.

In practice, the miners still expanding in 2026 share a few traits:

Cheap, firm power — long-term contracts, owned generation, or stranded/flared energy, typically well below $50/MWh all-in.
New-generation fleets — sub-16 J/TH average efficiency, which halves the energy bill per terahash versus 2020-era hardware.
Demand-response revenue — in markets like ERCOT (Texas), miners earn credits for curtailing during grid stress, effectively lowering their net power cost.
Low leverage — debt service at $12+ billion of aggregate sector debt (per crypto.news reporting) is what turns thin margins into forced selling.

Miners without these traits face the classic squeeze: revenue per machine falls with hashprice while their costs are fixed, and the rational response — switching machines off — is what shows up in aggregate as the 2026 difficulty decline.

ASIC efficiency: the J/TH arms race

Mining hardware is measured in joules per terahash (J/TH) — energy consumed per unit of work — and each generation of application-specific integrated circuits (ASICs) has pushed the number down. Machines of the 2016-era Antminer S9 class ran near 100 J/TH; 2020-era S19-class hardware brought that to roughly 30 J/TH; the 2023–2024 S21 generation reached the high teens and low teens. The frontier in 2026 is Bitmain's S23 family: the air-cooled S23 runs at roughly 11 J/TH, and the liquid-cooled S23 Hydro is the first production machine below the 10 J/TH threshold at about 9.5 J/TH (approximately 580 TH/s at 5,510 W), per reviews by Simple Mining and the 2026 J/TH rankings compiled by BT-Miners.

Two consequences follow. First, efficiency gains are slowing — each generation now shaves a few J/TH rather than halving the figure, so hardware refresh cycles no longer rescue laggards the way they did in 2018 or 2021. Second, because every halving doubles the revenue bar a machine must clear, older hardware ages out faster than its physical lifespan: a 25 J/TH machine that was competitive in 2023 is, at August 2026 hashprice, only viable on nearly free power.

The next halving: April 2028

The next halving arrives at block 1,050,000, when the subsidy drops from 3.125 BTC to 1.5625 BTC. Countdown trackers such as CoinWarz and CoinGecko project it for roughly mid-April 2028, with the exact date drifting with realized block times. Daily issuance will fall from about 450 BTC to about 225 BTC.

For miners, the 2028 halving is the deadline embedded in every capital decision being made today. Absent a doubling of bitcoin's price or of fee revenue, hashprice mechanically drops by nearly half at the halving. That arithmetic — applied to an industry already operating near breakeven in 2026 — is a large part of why so many public miners are racing to convert their power and land assets into AI infrastructure revenue that does not halve every four years.

Miner treasuries: who HODLs and who sells

Public miners differ sharply in what they do with the bitcoin they produce, and 2026 has redrawn the map. In the 2024–2025 bull market, several large miners adopted explicit "full HODL" treasury strategies, funding operations with equity and debt while stacking mined BTC. The 2026 downturn reversed much of that: crypto.news, citing sector trackers, reported that publicly traded miners sold more than 32,000 BTC in Q1 2026 alone — a single-quarter record exceeding their combined sales for all of 2025 — with MARA, Riot, CleanSpark, Core Scientific, Cango, and Bitdeer among the sellers.

The most dramatic shift is MARA Holdings, formerly the flagship HODLer among miners. MARA's holdings fell to 35,577 BTC as of June 30, 2026 — down roughly 29 percent year over year and about 34 percent from the 53,822 BTC it held at the end of 2025 — as it sold coins to fund operations, reduce debt, and finance its infrastructure buildout, per The Crypto Basic's summary of MARA's Q2 2026 report and MARA's own Form 10-Q for the quarter ended June 30, 2026. Of that total, only about 26,307 BTC was unrestricted; roughly 9,270 BTC was loaned or pledged as collateral. Riot Platforms likewise sold 3,778 BTC in Q1 2026 at an average near $76,626 — more than twice its quarterly production of 1,473 BTC, per Riot's Q1 2026 production update and crypto.news.

CleanSpark now runs an actively managed treasury: in July 2026 it produced 586 BTC, sold 579 BTC (229 on the spot market plus 350 via exercised call options at an average of $66,133), and ended the month holding 13,931 BTC, of which 4,070 BTC was posted as collateral or receivable on derivative positions, per its July 2026 operational update. Hut 8 restructured entirely: it moved its mining business into majority-owned American Bitcoin Corp. in 2025, and the two entities now hold bitcoin separately — Hut 8's own reserve was roughly 10,278 BTC as of May 2026 per Bitbo's treasury tracker, while American Bitcoin grew its stack from about 7,021 BTC on March 31, 2026 to about 8,002 BTC on June 30, 2026, per its SEC Form 8-K.

Company	BTC held	As of	2026 treasury posture
MARA Holdings	35,577	June 30, 2026 (SEC 10-Q)	Selling to fund debt reduction and the Long Ridge acquisition; ~9,270 BTC loaned or pledged
Riot Platforms	15,680	August 2026 (The Block)	Sold more than it produced in Q1 2026 to fund data-center expansion
CleanSpark	13,931	July 31, 2026 (company update)	Actively managed treasury; monthly spot and options-based sales
Hut 8	~10,278	May 2026 (Bitbo tracker)	Strategic reserve retained; mining spun into American Bitcoin Corp.
American Bitcoin Corp.	~8,002	June 30, 2026 (SEC 8-K)	Still accumulating mined BTC through 2026

Holdings figures move monthly; the live, sourced numbers for every miner we track are on the mining company holdings page, and recent disposals appear in latest Bitcoin sells. Smaller listed miners such as Cipher Mining, Bitfarms, and Bitdeer are tracked there as well.

The AI/HPC pivot: mining's second business model

The defining strategic story of 2025–2026 is bitcoin miners converting their scarcest asset — grid-connected, energized data-center capacity — into long-term AI and high-performance-computing (HPC) contracts. The deals are large enough to dwarf mining revenue:

IREN secured a $9.7 billion AI cloud contract with Microsoft, deploying Nvidia GB300 infrastructure from its Childress, Texas campus, and by mid-2026 had closed $3.65 billion in investment-grade financing to deliver it.
TeraWulf signed 10-year AI hosting agreements with Fluidstack — initially 200+ MW announced via its investor site, later expanded into a deal CoinDesk valued at $9.5 billion, with Google backstopping $1.3 billion of Fluidstack's lease obligations.
Cipher Mining stacked a 168 MW, ~$3 billion Fluidstack agreement (Google backstopped $1.4 billion and took an equity stake, later expanded) with a $5.5 billion, 15-year AWS lease for 300 MW of AI capacity.
CleanSpark signed its first HPC lease in July 2026 — a 20-year, 175 MW triple-net lease at its Sandersville, Georgia campus with an investment-grade technology tenant, worth about $6.6 billion in contracted revenue — alongside a Texas land acquisition for a further 300+ MW campus.
MARA agreed in April 2026 to acquire Long Ridge Energy in a $1.5 billion deal — a 505 MW gas plant and 1,600-acre campus in Ohio — explicitly to power AI data centers.
Core Scientific is the cautionary counterexample: CoreWeave's $9 billion all-stock acquisition was rejected by Core Scientific shareholders on October 30, 2025, who argued it undervalued the company; Core Scientific continues as an independent AI/HPC-focused operator.

The market has noticed: analysts increasingly value these companies on power contracts, land, and contracted AI revenue rather than on BTC production. For adoption tracking, the pivot cuts both ways — it stabilizes miner balance sheets (reducing forced BTC selling in future downturns) but also means less hashrate growth and, at several firms, a treasury that is now a financing tool rather than an accumulation strategy.

Miner capitulation dynamics

"Miner capitulation" describes the reflexive spiral that ends every mining down-cycle: falling hashprice pushes marginal operators below breakeven; they first sell treasury bitcoin to cover fixed costs, then switch off machines, then in the worst cases restructure or sell assets. The on-chain signature is a sustained difficulty decline paired with elevated miner outflows to exchanges. The 2026 episode ticks every box: a roughly 20 percent difficulty drawdown from the November 2025 peak (ChinaTechNews), record quarterly treasury sales (crypto.news), and aggregate sector debt near $12.7 billion.

Capitulation is self-limiting by design. Every difficulty retarget transfers revenue from the miners who left to the miners who stayed, until the marginal operator is again at breakeven. Historically these episodes — 2018, the 2021 China ban, late 2022 — have coincided with cyclical price bottoms more often than they have caused further declines, though past pattern is not a predictive claim. What is different in 2026 is the exit ramp: capacity leaving bitcoin mining is not being scrapped, it is being re-leased to AI tenants, which means some of this hashrate is unlikely to return even if hashprice recovers.

How miner holdings fit into institutional adoption tracking

Miners occupy a distinct niche in the institutional Bitcoin landscape. Unlike treasury companies or ETFs, they are structural producers of bitcoin: their holdings can grow without a single market purchase, and their disposals are a persistent, measurable source of sell pressure — roughly 450 BTC of new supply per day flows first through miner treasuries. Aggregate miner balances declining from about 1.86 million BTC toward roughly 1.8 million by mid-2026 (crypto.news) is therefore a supply-side datapoint, not just a corporate-treasury one.

On BTC Crawl, public miners appear alongside other corporate holders on the all holders list and in the dedicated mining companies view. Because miner treasuries now change month to month — sometimes by thousands of coins — each figure carries its source URL, verification status, and last-verified date, and figures sourced from company disclosures (monthly production updates, SEC filings) are treated as primary. Aggregator estimates are labeled as reported rather than verified. The full rules are on the methodology page. When a miner's disclosed balance drops, the change shows up in latest Bitcoin sells rather than being silently overwritten — the point is to make the flow visible, not just the stock.

FAQ

Is Bitcoin mining still profitable in 2026?

Only for efficient operators. At the August 2026 hashprice of about $32 per PH/s per day (Hashrate Index), fleets under roughly 14 J/TH can profit at power prices up to about $109/MWh, while older 25–38 J/TH fleets need power below about $41/MWh to break even. Many marginal operators are unprofitable, which is why difficulty has fallen roughly 20 percent from its late-2025 peak.

What is hashprice?

Hashprice is expected mining revenue per unit of hashrate per day, usually in USD per PH/s per day. Defined by Luxor's Hashrate Index, it combines bitcoin's price, difficulty, the block subsidy, and fees into one profitability benchmark. It was about $32.10/PH/s/day on August 3, 2026.

Do miners sell their Bitcoin?

Increasingly, yes. Public miners sold a record 32,000+ BTC in Q1 2026 alone (crypto.news). MARA — long a strict HODLer — reduced its stack about 34 percent from end-2025 to 35,577 BTC by June 30, 2026, and CleanSpark and Riot sell portions of monthly production. Hut 8 and American Bitcoin, by contrast, were still holding or growing reserves as of mid-2026. Track disposals on latest Bitcoin sells.

When is the next halving and what does it do?

At block 1,050,000, projected for roughly mid-April 2028 (CoinWarz countdown). The block subsidy falls from 3.125 to 1.5625 BTC, cutting daily issuance from about 450 to about 225 BTC and — all else equal — nearly halving hashprice overnight.

Why are Bitcoin miners building AI data centers?

Because their core asset — large, energized, grid-connected sites — is exactly what AI compute needs, and AI hosting revenue is contracted for 10–20 years and does not halve. IREN's $9.7B Microsoft contract, Cipher's $5.5B AWS lease, TeraWulf's Google-backstopped Fluidstack deals, and CleanSpark's $6.6B Georgia lease all converted mining infrastructure into long-duration cash flows during 2025–2026.

Bitcoin Mining Economics and Miner Treasuries