Bitcoin Halving Cycle Math: The Miner Shakeout Before Halving Five

Why Halving Math Is Really a Miner Solvency Problem

Most coverage of Bitcoin’s halving cycle asks a price question: does the four-year rhythm of ever-smaller peaks still hold, or has it finally broken. A companion piece on this site tackles that debate directly, weighing stock-to-flow against the power law model, Morgan Stanley’s seasonal framing and the case that Bitcoin is simply repricing as a macro asset (see Bitcoin Halving Cycle Math: The Road to Halving Five). This piece asks a narrower, more mechanical question instead. Set the price debate aside for a moment: what does the halving schedule actually do to the people who run the machines securing the network, and what does their behavior across four halvings say about what happens at the fifth.

The answer turns out to be more measurable than most price forecasts. Bitcoin’s block reward fell from 6.25 BTC to 3.125 BTC in April 2024, and it is scheduled to fall again, to 1.5625 BTC, at block 1,050,000, an event most trackers now place around mid-April 2028, per CoinWarz’s live halving countdown. Between now and then, somewhere around 900 exahash per second of installed hashing power has to keep covering its power bills on that shrinking subsidy, at a hashprice sitting near its lowest level in roughly five years, according to Luxor’s Hashrate Index. Who is already being pushed out, how the survivors are adapting, and what the arithmetic looks like heading into 2028 is the subject of everything that follows.

Bitcoin itself has already made one brutal round trip this cycle, from an October 2025 all-time high near $126,000 to a June 2026 low below $60,000, a drawdown examined in detail in Bitcoin Halving Cycle Math: What the 2026 Drawdown Actually Proved. Miners lived through the same round trip with leverage attached, because mining revenue depends on price, block reward and total network hashrate all at once, not on price alone.

What follows moves from the mechanics, the subsidy schedule and the hashprice math built on top of it, through the balance sheets of the three largest public miners, into the pool-level plumbing that actually assembles blocks, and finally into the historical precedent for what a sustained margin squeeze does to an industry, before closing with explicit, checkable predictions for the two years leading into halving five.

The Subsidy Schedule Miners Actually Live By

Bitcoin’s issuance schedule is not a policy that gets revised by a vote. It is hardcoded into the GetBlockSubsidy function in Bitcoin Core, which cuts the block reward in half every 210,000 blocks, roughly four years, until the subsidy eventually rounds down to zero and the 21 million cap is fully issued, an event not expected until around the year 2140 across 33 total halvings. Four halvings in, the pattern is simple to state and expensive to live through: half the new bitcoin, twice the pressure on anyone whose business model depends on selling it to cover costs.

The table below lays out the full schedule to date, plus the market’s current best estimate for the fifth halving.

As of mid-July 2026, most countdown trackers put halving five somewhere around 600 to 650 days out, with block 1,050,000 expected in the first half of April 2028. More than 95% of the eventual 21 million BTC supply is already in circulation, per CoinGecko’s supply tracker, which means each successive halving does less to the total stock of bitcoin in existence and more to the day-to-day cash flow of the industry that secures it. That distinction matters more for this piece than for almost any other halving angle: holders experience a halving as a supply-shock narrative, miners experience it as an overnight 50% pay cut with no ability to renegotiate. Only four more halvings after this one will ever occur, in roughly 2032, 2036, 2040 and every four years after that until the subsidy rounds to zero near 2140, at which point miners will depend entirely on transaction fees, a topic revisited later in this piece.

Hashprice: The Number That Actually Decides Who Survives

Miners do not budget around the spot BTC price on its own. They budget around hashprice, the expected daily dollar revenue per unit of hashing power, usually quoted in dollars per petahash per second per day. The metric was popularized by Luxor’s Hashrate Index and combines BTC price, block reward and transaction fees in the numerator, dividing by total network hashrate in the denominator. It is the single figure that tells an operator, on any given day, whether a specific machine is making money or burning it.

The formula also explains why a halving is mechanically brutal for hashprice even before anything else changes. Cut the block reward in half and, holding BTC price and network hashrate constant, hashprice halves overnight by definition. In practice price and hashrate never stay constant for long, which is exactly why the four-year cycle argument covered in our companion pieces on this topic and the mining-margin argument in this piece are really two views of the same underlying mechanism, just measured differently.

The trajectory over the past year has been rough by that mechanical logic. Hashprice traded in a roughly $55 to $63 per PH/s per day range around the middle of 2025, as Bitcoin ran toward its October all-time high, then collapsed to a five-year low near $28 to $30 per PH/s per day by the first quarter of 2026, according to CoinShares’ Q1 2026 Bitcoin Mining Report. As of mid-July 2026, trackers put it back in a $29 to $33 per PH/s per day range depending on the index used, still down by roughly half from a year earlier.

For context, the last time hashprice sat this low was 2020, in the weeks after the COVID-19 crash, when the entire market, not just mining, was pricing in the worst case. That similar levels have returned now, even with Bitcoin trading well above its 2020 price in absolute terms, is a clear sign that the halving’s arithmetic, not sentiment alone, is doing much of the work.

The Efficiency Arms Race: From 98 Joules to Sub-14

If hashprice is the revenue side of a miner’s ledger, joules per terahash (J/TH) is the cost side, and it has fallen harder than almost any other input cost in the industry over the past decade. The metric measures how much electrical energy a machine needs to produce one terahash of SHA-256 hashing per second; lower is better, and the generational improvement has been dramatic enough to mostly offset four halvings on its own.

That last figure is not a rough estimate: Bitmain’s own specifications put the Antminer S21 XP at 13.5 J/TH, running 270 TH/s at roughly 3,645 watts. Compared with S9-era hardware at around 98 J/TH, that is roughly a sevenfold efficiency gain in eight years, which is the main reason total network hashrate has kept climbing even as individual-unit economics have gotten tighter. Historically, a new flagship ASIC generation has landed within a year or two of each halving, almost like clockwork, which is no accident: efficiency is the one lever a miner fully controls, unlike price or network-wide hashrate.

The practical consequence shows up in breakeven power pricing. Using the standard formula, breakeven cost in dollars per kWh equals hashprice (in dollars per PH/s per day) divided by 24 times efficiency in J/TH. At July 2026’s roughly $29 to $30 per PH/s per day hashprice, a 13.5 J/TH S21 XP breaks even around $0.09 per kWh; a 21.5 J/TH S19 XP-class machine breaks even nearer $0.056 per kWh; and 98 J/TH S9-class hardware needs power under roughly $0.012 per kWh, a price that barely exists outside a handful of stranded-gas or heavily subsidized sites. That single piece of arithmetic is why S9s are now effectively museum pieces, while sub-15 J/TH fleets sitting on sub-$0.06 power are still cash-flow positive.

Cash Cost Versus All-In Cost: Reading a Miner’s Real Margin

Public miners’ earnings often look confusing because two very different cost lines get reported side by side: cash cost per BTC, which covers power and direct operating expenses, and all-in cost per BTC, which layers on depreciation of the mining fleet itself. A miner can be profitable on one basis and underwater on the other in the very same quarter.

Riot Platforms’ first-quarter 2026 10-Q filing with the SEC is a clean worked example. Cost to mine one bitcoin, excluding depreciation, was $44,629, against a production value of $75,964, comfortably cash-profitable. Including depreciation, the all-in cost jumped to $96,283, or 126.7% of production value, meaning the same quarter was a loss on a fully-loaded GAAP basis. Both figures are true at once; they just answer different questions about the same three months.

Across the public-miner sector more broadly, CoinShares put the weighted-average cash cost per BTC at just under $80,000 in the fourth quarter of 2025, while JPMorgan analyst Nikolaos Panigirtzoglou estimated an all-in production cost of roughly $78,000 and noted Bitcoin traded around 19% below that level for about five consecutive months into mid-2026. As Panigirtzoglou put it: “When bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower.” CoinShares estimated 15 to 20% of the global fleet was running at a loss under this framing. For a fuller, company-by-company breakdown of margins, see Bitcoin Mining Margins Explained: Inside the 2026 Profit Squeeze.

Record Hashrate, Record-Low Hashprice: The Signal Nobody Predicted

One of the stranger features of 2026 is that network hashrate and hashprice have been moving in opposite directions at the same time. Network hashrate has held near all-time highs, with trackers showing a range of roughly 860 to 915 exahash per second in July 2026 depending on methodology, even as hashprice sat at a five-year low.

Difficulty tells the same story from another angle. It peaked just above 155 trillion in late October 2025 after a single 6.31% upward retarget, then posted three consecutive negative adjustments, the first such streak since July 2022, before settling back into the low-to-mid 130-trillion range by mid-2026, per CoinShares’ mining report. Separate industry analysis of the same period pointed out that the first quarter of 2026 produced Bitcoin’s first year-over-year hashrate decline in a first quarter in six years, an early capitulation signal even though the absolute hashrate level stayed historically high, because efficient new-generation rigs kept coming online at roughly the same pace that old S9 and S17-class units were switched off.

This is the mechanism Panigirtzoglou described playing out in real time: price falls, the highest-cost machines go dark first, hashrate and difficulty adjust down with a lag, and hashprice for the remaining, more efficient fleet stabilizes a little. It is a slower, quieter process than a price crash, which is exactly why it tends to get less attention than it deserves.

The pattern is not unique to this cycle. After each of the prior three halvings, difficulty historically took a brief step down or plateaued for a few retarget periods as the least efficient miners of that era switched off, before resuming its longer-run climb once newer hardware caught up to the lower reward. What is different this time is the scale and the overlap with the AI buildout: previous post-halving dips were purely a mining-industry story, while this one is unfolding alongside a parallel, much larger capital cycle in AI data-center construction that is competing for some of the same power and, increasingly, some of the same balance sheets.

The Public Miner Scorecard: MARA, Riot, CleanSpark and the AI Pivot

The largest Nasdaq-listed pure-play Bitcoin miners have spent the past year answering the margin squeeze in different ways, but converging on the same broad strategy: use power infrastructure and balance-sheet BTC to fund a move into AI and high-performance-computing hosting, rather than betting solely on a Bitcoin price recovery.

MARA’s first-quarter 2026 results come with a footnote worth reading closely: “energized” hashrate includes rigs that are plugged in but not necessarily running at full tilt, so it can overstate delivered output relative to figures reported as “operating” hashrate elsewhere. The company’s net loss widened to roughly $1.26 billion for the quarter, driven mostly by a paper markdown on its BTC holdings rather than by operating losses alone.

CleanSpark’s own June 2026 operational update put average operating hashrate at 42.6 EH/s against 50 EH/s of nameplate capacity, with the gap attributed largely to uptime and curtailment. Unlike MARA and Riot, CleanSpark has stayed closer to a pure-play mining model, still building out sites in Sandersville, Georgia and Texas rather than announcing large AI hosting contracts of its own so far.

Riot’s pivot has an activist push behind it. In a letter dated February 18, 2026, Starboard Value argued that Riot’s remaining uncommitted power capacity, roughly 1.4 gigawatts across its Corsicana and Rockdale sites in Texas, could generate more than $1.6 billion in annual EBITDA if leased at rates comparable to recent AI data-center deals, an amount that would dwarf the company’s roughly $6.3 billion market capitalization at the time, per CoinDesk’s coverage of the letter. Riot’s stock jumped nearly 7% on the news.

MARA CEO Fred Thiel has been the most blunt public voice on where all this is heading. “By 2028, you’ll either be a power generator, be owned by one, or be partnered with one,” he said. “The days of being a miner plugged into the grid are numbered,” according to comments reported by CoinGeek. Thiel frames plain mining in similarly stark terms: “it’s a zero-sum game. As more people add capacity, it gets harder for everybody else. Margins compress, and the floor is your energy cost.” CoinShares head of research James Butterfill has quantified how fast the pivot is happening industry-wide, saying miners could derive “as much as 70% of their revenues from AI by the end of this year, up from roughly 30% today,” according to The Block, though he added a Bitcoin recovery to $100,000 was “not an unrealistic assumption.” Needham & Co analyst John Todaro summed up why the pivot pencils out on paper: “The revenue per megawatt and EBITDA margins are far higher for HPC and AI colocation than for mining.”

The pivot is not confined to the three names above. CoinShares has tracked more than $70 billion in cumulative announced AI and HPC contracts across the public mining sector, including Core Scientific’s roughly $10.2 billion, twelve-year colocation deal, Hut 8’s roughly $7 billion, fifteen-year, 245-megawatt lease, and TeraWulf’s roughly $12.8 billion in total contracted HPC revenue. Much of that has been funded with fresh debt rather than existing cash flow, which raises its own question: whether these companies are financing genuine diversification or simply borrowing against a bet that AI hosting demand holds up as well as its contracts promise.

All three headline miners carry their BTC holdings at fair value on the balance sheet, so a meaningful share of every quarterly result is really a leveraged bet on Bitcoin’s price layered on top of the mining and hosting business itself. That is worth remembering before reading too much into any single quarter’s headline loss.

Who Actually Mines the Blocks: Pool Concentration and the Stratum V2 Race

Almost no individual miner finds blocks on its own; the variance is too high for that to be practical below an industrial scale. Instead, hashpower is pooled, and pooling concentrates a surprising amount of influence over what actually goes into a block in the hands of a small number of pool operators.

The top five pools above control close to four-fifths of Bitcoin’s hashrate between them, per Hashrate Index’s pool tracker, a concentration level that periodically revives 51%-attack anxiety online. Researchers generally point to a narrower, more realistic risk instead: under the original Stratum V1 protocol, it is the pool operator, not the individual miner, who assembles a block’s contents, which raises transaction-censorship concerns rather than double-spend risk. F2Pool, for instance, has again been flagged for filtering transactions linked to OFAC-sanctioned addresses.

The industry’s answer is Stratum V2, whose Job Declaration feature lets an individual miner build its own block template while the pool simply smooths out payouts and validates completed work. In May 2026, seven pools together representing about 75% of network hashrate, including Foundry, AntPool, F2Pool, SpiderPool, MARA Pool, Block Inc. and DMND, joined a working group to standardize the rollout. The first production block built this way was mined the following month.

There is also a regulatory backdrop worth noting for readers weighing whether any of this touches securities law, a question examined in depth in The Howey Test in 2026: How the SEC Decides What’s a Security. The SEC’s Division of Corporation Finance addressed proof-of-work mining specifically, a position later folded into a broader interpretive release in March 2026 concluding that mining, whether solo or pooled, does not itself involve a securities transaction under the Howey framework, because the reward flows from the miner’s own computational contribution rather than the managerial efforts of a third party. That legal clarity is a large part of why MARA, Riot and CleanSpark can operate as ordinary Nasdaq-listed industrial companies rather than fighting a registration battle the way some staking-as-a-service providers have.

The Shakeout Precedent: What Happened Last Time Margins Broke This Badly

Sustained sub-cost hashprice is not a new phenomenon, and history offers a fairly clear preview of how it tends to resolve: not through the network shrinking evenly, but through the highest-cost operators failing outright while efficient operators absorb their share of the network.

An earlier, milder version played out in 2018 and 2019, when Bitcoin’s price fell from around $20,000 to under $3,500 and a wave of smaller miners and hosting operations shut down or sold hardware for scrap, though nothing on the scale of what followed in 2022.

The clearest precedent is 2022. Core Scientific, then one of the largest US-listed miners, filed for Chapter 11 bankruptcy protection in Texas on December 21, 2022, citing the prolonged decline in Bitcoin’s price, rising electricity costs and a rising network hashrate that combined to crush its margins, with estimated liabilities between $1 billion and $10 billion. Compute North, a mining-hosting provider that counted several public miners as customers, filed its own Chapter 11 in September 2022, owing as much as $500 million to at least 200 creditors. Celsius Mining’s operations went through the same bankruptcy process that year. Core Scientific eventually reorganized and relisted on Nasdaq in January 2024, but only after wiping out a large share of prior equity holders along the way.

Trade coverage of the mining sector is already framing 2026 in similar terms. Reporting on post-halving mining economics points to a widening efficiency gap between best-in-class and legacy fleets as reason enough for acquirers to buy existing efficient hashrate rather than upgrade older sites themselves, with further consolidation expected through 2026 regardless of where Bitcoin’s price goes next. The mechanism is the same one described earlier in this piece: hashprice near cost, difficulty adjusting down with a lag, and the weakest balance sheets running out of room first.

Modeling Halving Five: Three Scenarios for Spring 2028

One thing about halving five is not in question: the subsidy drops from 3.125 BTC to 1.5625 BTC per block, cutting new daily issuance from about 450 BTC to about 225 BTC. Everything else, Bitcoin’s price, total network hashrate and therefore hashprice, is an assumption. The table below is simple illustrative arithmetic, not a forecast, built by holding the halved issuance fixed and varying the price input. Treat it as a way to reason about the range of outcomes, not a prediction of which one arrives.

The base case roughly tracks where several of the forecasting frameworks in our companion piece on halving-five modeling already expect the cycle to sit by 2028, assuming the four-year rhythm holds at all. Under the bear case, expect the shakeout described above to simply continue and probably intensify. Under the bull case, expect renewed hashrate growth and likely new entrants, undoing some of today’s consolidation.

Each scenario also assumes fee revenue stays roughly where it is today, under 1% of block rewards. If fee demand were to spike well beyond that, during a period of sustained on-chain congestion, for example, it would cushion any of the three cases somewhat, though current data shows no sign of that trend yet, and miners have generally treated fee income as a rounding error in their planning rather than a line item to build a strategy around.

Fees, the Security Budget and Bitcoin’s Long Game

Transaction fees remain a small share of miner revenue. CoinShares’ data shows fee income consistently under 1% of total block rewards in recent quarters, nowhere near the level some researchers argue will eventually be needed to fund network security once the subsidy approaches zero over the coming century.

Thiel has been candid that the fee-driven transition away from subsidy dependence “hasn’t happened,” and that the math “gets very tough after 2028” without sustained Bitcoin price growth well above 50% a year, a pace nobody should treat as a base case. This is not a new anxiety; block-reward-versus-fee debates go back to at least the block-size disputes of the mid-2010s. Some researchers have floated a rough rule of thumb that fees eventually need to cover something like 20% or more of total miner revenue for the security budget to stay comfortable deep into the subsidy’s decline, though estimates like this vary widely and depend on assumptions about future price and hashrate that nobody can pin down decades in advance. What is new heading into halving five is that, for the first time, the largest miners have a live alternative revenue stream in AI and HPC hosting, rather than pure hope for fee growth to fall back on.

What Proof-of-Work Miners and Proof-of-Stake Validators Have in Common

Bitcoin is not the only network wrestling with a shrinking guaranteed subsidy. Ethereum validators earn a mix of protocol issuance, priority fees and MEV, a different mechanism but a structurally similar question: can transaction-fee demand eventually fund network security on its own, without relying on fixed issuance. That question, and how validator payouts actually break down in practice, is covered in detail in Validator Economics Explained: How Stakers Actually Get Paid.

The two systems answer the question differently. Bitcoin’s answer is a hard, scheduled halving every 210,000 blocks, fully predictable and impossible to negotiate. Ethereum’s issuance instead floats with total stake participation and is partly offset by fee-burning, a softer, continuously adjusting version of the same tension. Miners also carry a cost structure validators mostly avoid: sunk hardware that depreciates on a schedule set by the efficiency curve described earlier, plus long power contracts, whereas staked capital is comparatively liquid and can exit far faster than a warehouse full of ASICs. Liquid staking derivatives add a further wrinkle on the Ethereum side, letting capital keep earning validator-style yield while remaining tradable or usable as collateral elsewhere, something with no real equivalent for capital already sunk into a mining rig.

What This Means for the 2026-2028 Predictions Cluster

Pulling the data above together into a set of concrete, checkable expectations for the two years between now and halving five:

• Expect further consolidation among smaller public miners and mining-hosting providers if hashprice does not durably recover above the mid-$40s per PH/s per day, following the Core Scientific and Compute North playbook from 2022.
• Expect the largest public miners to keep expanding AI and HPC hosting revenue rather than betting purely on a Bitcoin price recovery to survive halving five, extending the pattern already visible at MARA, Riot, CleanSpark, Core Scientific, Hut 8 and TeraWulf.
• Expect pool concentration to stay a live governance topic, with Stratum V2 and locally-built block templates as the practical mitigation rather than any near-term protocol-level cap on pool size.
• Expect hashprice and difficulty to keep adjusting downward first whenever Bitcoin’s price corrects, with the spot-price narrative catching up second, the same sequence described earlier in this piece.
• Expect the loudest halving-five headlines in 2028 to again be about the block reward and price, even though the more decisive story will be which miners’ balance sheets could absorb another 50% cut to daily issuance.

None of this resolves the bigger argument in our companion pieces about whether Bitcoin’s four-year cycle is still intact on the price side. What the mining data does settle is the supply side: the halving schedule is not up for debate, only the demand that meets it.

Frequently Asked Questions

When is the next Bitcoin halving, and how much will the block reward drop?

Most trackers place Bitcoin’s fifth halving in spring 2028, clustering around mid-April 2028 at block height 1,050,000, per CoinWarz’s halving countdown. The block reward falls from 3.125 BTC to 1.5625 BTC, cutting new daily issuance from roughly 450 BTC to roughly 225 BTC. The exact date shifts slightly with network hashrate between now and then, since blocks are not mined on a fixed clock.

What is hashprice, and why do miners watch it more closely than the Bitcoin price?

Hashprice is the expected daily revenue per unit of hashrate, usually quoted in dollars per petahash per second per day, a metric popularized by Luxor’s Hashrate Index. It combines BTC price, block reward, transaction fees and total network hashrate into the single figure that actually determines whether a specific machine is profitable that day, which is why it matters more to a miner’s survival than the spot BTC price on its own.

Why is Bitcoin’s hashrate near record highs while mining profitability sits at multi-year lows?

Two forces are pulling in opposite directions. Newer-generation machines like the Antminer S21 XP run at a fraction of the power per terahash of 2016-era hardware, so total network hashrate keeps climbing even as individual-unit revenue shrinks. At the same time, the 2024 halving cut the block reward and Bitcoin corrected sharply from its October 2025 high, so hashprice, revenue per unit of hashrate, fell to roughly a five-year low even while aggregate hashrate stayed close to its all-time high.

Will Bitcoin mining still be profitable after the 2028 halving?

It depends almost entirely on where Bitcoin’s price and network hashrate sit at the time, since the subsidy cut itself, from about 450 to about 225 new BTC per day, is already certain while demand is not. Industrial operations with sub-15 joule-per-terahash hardware and power costs under roughly $0.06 to $0.08 per kWh have stayed profitable through past halvings; higher-cost operators historically have not, which is why margin squeezes have repeatedly triggered consolidation and bankruptcies, as with Core Scientific and Compute North in 2022.

Does the SEC treat Bitcoin mining rewards as securities?

No. The SEC’s Division of Corporation Finance concluded that proof-of-work mining, whether solo or through a pool, does not involve a securities transaction under the Howey framework, because the reward comes from the miner’s own computational work rather than a third party’s managerial effort, a position folded into a broader interpretive release in March 2026.

HOGE Wire Markets Desk.