Hash price — the daily revenue earned per petahash of compute — moved from $27.70 in June to $40 by late August. That is the only sentence in the CryptoQuant weekly that carries independent information. The 78% aggregate revenue rebound, the 962 EH/s hashrate print, the "reasonable returns" characterization — every one of those is either a downstream consequence of that single number or a restatement of the Bitcoin price itself. My task is to separate the signal from the echo.
I have built a career on this kind of separation, usually at smaller scale and always against the grain of the prevailing narrative. In 2020, during the first DeFi summer, I left a corporate risk seat to audit Curve Finance's liquidity pools independently. I hand-traced the invariant calculations for the 3Pool and found that the parameterized fee structure opened a subtle arbitrage window for high-frequency traders during volatility spikes. The pool's headline yield was a marketing figure. The invariant was the truth. I sold a forty-page report on that distinction for $15,000 and have applied the same method to every protocol I have touched since: locate the invariant, ignore the banner.
For Bitcoin mining the invariant is a three-variable relationship — the protocol subsidy, the transaction fee pool, and the network hashrate. Miner revenue is the product of the first two. Hash price is that product divided by the third. Any recovery claim that does not survive this decomposition is not a recovery. It is a repricing. Hype evaporates; solvency remains. What follows is a line-by-line teardown of the number, what it contains, what it conceals, and precisely where the break points sit.
To read the rebound you must first read the constraint that produced it.
In April 2024 Bitcoin executed its fourth halving. The block subsidy fell from 6.25 BTC to 3.125 BTC per block. At ten-minute intervals that is 144 blocks per day, so daily issuance dropped from roughly 900 BTC to roughly 450 BTC. This was not a market event. It was a deterministic, pre-announced 50% cut to the single largest revenue line of every mining operation on the planet. Ledger integrity precedes market sentiment — and here the ledger is unambiguous.
A mining operation is a conversion machine. It converts electricity and capital into hashrate, and hashrate into a probabilistic claim on block rewards. Revenue arrives through two channels: the block subsidy, fixed by protocol, and transaction fees, which float with on-chain demand. Costs are almost entirely fiat-denominated — electricity priced in kilowatt-hours, hardware amortized over its useful life, cooling, labor, pool fees, and increasingly the cost of capital itself. Profit is the spread between two currencies: revenue denominated in BTC, costs denominated in dollars. That asymmetry is the entire story of mining risk, and it is the thing the 78% headline obscures.
Between May and August 2024, per the weekly's own information points, miners operated in a "severe loss" regime. From August 21 the characterization shifted to "reasonable returns." That sequence — loss, then recovery, without any protocol change — is the standard script of a post-halving margin squeeze, not the output of a technical upgrade. Nothing in Bitcoin's consensus rules changed between May and August. No soft fork activated. No efficiency breakthrough arrived. The only variables that moved were price and hashrate. One of them is exogenous to the miner. The other is a consequence of the first. Stability is a calculated illusion when the stabilizing variable is owned by someone else.
This matters because the mining sector has spent fifteen years building a rhetorical apparatus around the idea that its economics are self-correcting — that hashrate, difficulty, and price form a negative-feedback loop that naturally finds equilibrium. There is truth in that. Difficulty does adjust. Marginal miners do shut down. But the loop is slow, capital-intensive, and path-dependent, and in the short run it produces exactly the kind of violent margin swings we observed between May and August. The word "self-correcting" is doing a great deal of quiet work in most mining commentary. What it actually means is: miners lose money for a while, then some of them quit, and the survivors get cheaper blocks. That is not stability. It is attrition.
To evaluate the 78% figure properly, I want to set out the four structural facts that frame it, because everything downstream depends on them.
First, the subsidy is now 3.125 BTC per block, or roughly 450 BTC per day. At a BTC price near $84,000, that is approximately $37.8 million of protocol-issued revenue per day, before a single transaction fee is counted. Second, daily aggregate miner revenue at its recent peak was reported near $48 million. The difference — roughly $10 million, or about 21% — is the transaction fee contribution. Third, network hashrate stood at 962 EH/s as of the late-July reading, up from 899 EH/s, a 7% increase. Fourth, miner outflows were running near 12,000 BTC per day, with the weekly noting that extreme outflow events had not recurred.
Hold those four numbers. The entire analysis lives inside them, and two of them are internally inconsistent in ways that deserve their own treatment.
Start with the 78%.
Revenue is quantity times price. For a miner, the quantity is block rewards earned, and the price is the dollar value of BTC. Between the trough and the peak of the window under discussion, Bitcoin rose approximately 45% from its low. If miner revenue rose 78% while the coin rose 45%, the residual 33% must come from somewhere. There are only two candidates: a change in the quantity of rewards captured, or a change in the measurement window and its endpoints. The protocol did not change the subsidy between May and August — the halving was in April, and the next subsidy cut is roughly four years away. So the quantity of protocol rewards is fixed at 450 BTC per day regardless of what price does.
That leaves fees and the window. If fees expanded meaningfully, revenue could outrun price. But the weekly's own arithmetic puts fees at roughly 21% of total revenue — about $10 million against a $48 million peak. For the 78% revenue move to be more than a price move, the fee component would have had to expand faster than the subsidy component, and nothing in the data suggests it did. The more parsimonious explanation is that the 78% is measured from a trough in both price and fees to a peak in both, and that the double-counting of the two trough-to-peak moves produces a headline that exceeds the price move alone.
This is not a criticism of the number's accuracy. It is a criticism of its interpretation. A 78% revenue rebound that is 100% explained by a 45% price move plus a favorable fee window is not evidence of sector health. It is evidence of sector leverage to price. And leverage cuts both ways. Precision is the only risk mitigation, and the precise statement is this: miner revenue is a linear function of the Bitcoin price, scaled by a slowly-varying reward quantity, and perturbed by a volatile and currently small fee component. Nothing in that sentence describes a self-sustaining recovery.
Now isolate the cleaner metric. Hash price — revenue per petahash per day — rose from $27.70 in June to $40. This is the better instrument because it divides out hashrate growth. If hashrate is rising while hash price is also rising, the implication is that revenue grew faster than hashrate. If hashrate is rising while hash price is flat or falling, the marginal miner is being diluted even as the aggregate pie grows. The move from $27.70 to $40 is a 44% improvement in unit economics, and it is the number that actually reflects what a miner earns per unit of capital deployed.
The distinction between the 78% and the 44% is the distinction between aggregate and per-unit. Aggregate revenue can rise while per-unit revenue falls, provided hashrate grows faster than revenue. That has been the structural condition of Bitcoin mining for most of its history, and it is why the sector is a treadmill: you must keep buying machines to keep your share, and buying machines increases the hashrate that dilutes everyone. A 44% hash price recovery in a single quarter is genuinely favorable. But hash price of $40 is not a return to the conditions of prior cycles, when hash price routinely sat above $60 and sometimes above $100. Context matters. The recovery is real. The level is still depressed relative to history, and the direction of the difficulty adjustment argues that even $40 may not persist.
Here is the mechanism the weekly mentions but does not pursue: difficulty.
Bitcoin targets a ten-minute block time. When hashrate rises, blocks arrive faster than target, and the protocol raises difficulty every 2,016 blocks — roughly every two weeks — to restore the target. Difficulty is the network's throttle. It is also, for miners, a tax on their own expansion. When 962 EH/s of hashrate is pointed at the network, difficulty rises to match, and the same 450 BTC per day is spread across more compute. Hash price falls.
The consequence is a one-way ratchet in the short term. Hashrate rose 7% between the late-July and August readings. That growth is not free: it reflects capital already committed — machines ordered months ago, hosting contracts signed, power agreements executed. Those machines will come online regardless of what price does next, because the capital is sunk. And as they come online, they push difficulty up, and they push hash price down. The $40 hash price is therefore not a new equilibrium. It is, with high confidence, a local high — a reading taken before the difficulty adjustment fully absorbed the hashrate increase that produced it.
This is the automatic stabilizer working in reverse. The mechanism that supposedly rescues miners in bad times — difficulty falling as unprofitable miners shut off — is the same mechanism that punishes them in good times, by diluting the recovery. The weekly frames the hashrate increase as a sign of strength, and in one sense it is: capital is being deployed, miners are not capitulating, the network is more secure. But strength at the network level and profitability at the operator level are different things, and they move in opposite directions more often than the commentary admits.
I have seen this exact confusion before, in a different market. When I assessed the Bored Ape floor collapse in 2022, the prevailing narrative treated the floor price as a measure of collection health. It was not. The floor was a measure of the last marginal seller's willingness to transact, and it could be manipulated, and roughly 12% of it was — wash trades engineered to support NFT-backed lending. The lesson generalizes. A headline metric that aggregates a heterogeneous population will almost always misrepresent the median member of that population. Aggregate hashrate is a headline. Hash price is the median member. When they diverge, trust the median.
The fee ratio deserves its own treatment, because it is the sector's long-term structural problem and the weekly only gestures at it.
Bitcoin's security budget has two components. The subsidy is finite — it halves every 210,000 blocks and reaches zero around 2140. The fee market is supposed to replace it. For that to happen, transaction fees must grow to fill the gap left by a shrinking subsidy. At present, fees are roughly 21% of miner revenue, and the subsidy is roughly 79%. That ratio is inverted from where a mature fee market needs to be. The long-run target, if Bitcoin is to remain secure without a subsidy, is a fee share north of 50%, ideally higher.
Twenty-one percent is not close. Worse, the fee component is unstable in a way the subsidy is not. The subsidy is deterministic. Fees are a function of block space demand, which spiked in 2023 and 2024 on the back of Ordinals and inscription activity — a speculative, cyclical demand source that has already cooled materially. If inscription activity continues to fade, the fee share will fall, and miner revenue will become even more tightly bound to the Bitcoin price. That is the definition of a concentration risk: one variable — price — carrying the weight of the entire revenue model.
There is a subtler problem here that almost no one quantifies. When fees are a small share of revenue, miners have weak incentive to include high-fee transactions, because the marginal fee is trivial relative to the block subsidy. The economics of block template construction become dominated by subsidy capture, not fee maximization. This is not a failure of miner rationality; it is a correct response to the incentive structure. But it means the fee market's development is partly throttled by the very subsidy that is supposed to be replaced. The two components of the security budget are not independent — they are in tension, and the tension resolves, perversely, in favor of the subsidy.
Audits reveal what code conceals, and the code here is the issuance schedule. It conceals nothing about the subsidy — that number is public and predictable to the satoshi. What it conceals is the fragility of the fee channel, because the fee channel is not in the code at all. It is in the behavior of users, and user behavior is the least auditable input in the system.
The 365-day moving average sits at approximately $80,000, and the weekly correctly identifies it as the critical level. What the weekly does not say — because it is a data digest, not a risk document — is that this level is overdetermined. It is simultaneously a technical support level and, at roughly $80,000, the approximate break-even for the marginal miner.
That coincidence is not an accident. The 365-day average and the marginal cost of production converge because the marginal cost of production is, over long horizons, a price floor. Miners who cannot cover costs with current prices shut down, reducing supply, supporting price. Miners who can cover costs keep running. The equilibrium sits at the cost of the least efficient producer who remains online. In a market where the marginal producer's all-in cost — power, hardware amortization, overhead — lands near $80,000, the 365-day average will tend to gravitate toward that level, and price will tend to bounce off it, for mechanical reasons rather than mystical ones.
But this floor is conditional, and the condition is the composition of the mining fleet. If the marginal cost is $80,000 because the marginal miner runs older, less efficient hardware on expensive power, then the floor is soft: a price break forces those miners offline, the floor drops, and the surviving fleet has a lower marginal cost. If the marginal cost is $80,000 because the fleet has already upgraded and the survivors are efficient, then the floor is harder, and a break is more likely to be brief. The weekly's data — hashrate rising rather than falling after the halving, implying that inefficient machines have already been retired — leans toward the harder-floor interpretation. The fleet that survived the halving is not the fleet that entered it.
That is a genuinely favorable structural read, and I want to give it its due. But it comes with a corollary the bulls miss. If the surviving fleet is efficient and the floor is hard, then the downside is limited — but so is the upside, because the same efficiency that makes the floor hard makes the industry quick to add hashrate when margins improve, which caps the recovery through dilution. A hard floor is a good thing for solvency. It is a bad thing for the kind of explosive margin expansion that produces mining stock manias.
Now the outflow number, which is the most misread figure in the weekly.
The report notes roughly 12,000 BTC per day of miner outflow and observes that extreme outflow events have not recurred. The natural reading — miners are not panic-selling, sentiment is stable — is the one most commentary adopted. I want to flag that this reading is almost certainly wrong, or at least unverified, and that the weekly itself does not support it.
Twelve thousand BTC per day against a daily production of roughly 450 to 500 BTC is a ratio of approximately twenty-four to one. Miners do not sell twenty-four times their daily production every day. They cannot; they would exhaust their treasuries within weeks. A figure that large relative to production almost certainly measures wallet movements rather than sales. Miner outflows, as typically constructed, capture any movement out of wallets tagged as belonging to miners, including internal transfers between custody arrangements, collateral postings for loans, rebalancing between hot and cold storage, and consolidations ahead of custody migrations. None of those are sales. Treating the outflow figure as a proxy for sell pressure is a category error.
I have made this exact mistake in my own work and learned from it. When I built the forensic report on BAYC collateral, the naive approach was to count transfers. Counting transfers produced a distorted picture because a large fraction of transfers were the same asset moving between affiliated wallets. The correction was to build a graph of addresses and to weight transfers by the economic distance between the endpoints — a move between two addresses controlled by the same entity counts as zero. The same discipline applies to miner outflows. The number is only informative after you separate genuine disposals from internal bookkeeping, and the weekly does not perform that separation.
The correct inference is therefore narrower than the headline suggests. We can say that extreme outflow events have not recurred — meaning no single-day episode of the kind associated with capitulation. We cannot say that miners are accumulating, or that sell pressure is low, or that sentiment is stable. Those conclusions require an address-level analysis that the aggregate does not support.
Floor prices are illusions of liquidity, and the same is true of outflow aggregates. Both measure the last observable transaction, not the underlying willingness to sell. Both can be inflated by internal activity. Both are quoted as facts when they are, at best, estimates with wide error bars.
There is a second internal inconsistency in the data that the weekly does not resolve, and it bears on the reliability of the whole dataset. The weekly reports daily aggregate revenue at a peak near $48 million and separately reports the subsidy at roughly 450 BTC per day, which at $84,000 equals about $37.8 million. The residual, $10.2 million, is attributed to fees. That implies fees of about 21% of revenue. But 21% fee share is high relative to the fee share observed in most of 2024 outside the peak inscription windows. Either the peak revenue reading coincided with an unusually active fee environment, or the revenue figure includes components — such as transaction fee subsidies or pool-level adjustments — that are not strictly protocol revenue. Without the underlying block-level data, the fee share cannot be verified, and a 21% assumption is doing a lot of load-bearing work in the analysis.
This is not nitpicking. If the true fee share is 12% rather than 21%, then miner revenue is even more price-dependent than the headline suggests, and the 78% rebound is even more completely explained by price. The direction of the error matters, and the weekly does not give us enough to determine it. Precision is the only risk mitigation — and here the available data does not permit precision. That is itself a finding.
Let me turn to the parts of the mining stack that the weekly, as a data digest, does not address, but which determine whether the recovery is durable.
The first is hardware turnover. A halving does not merely halve revenue; it raises the efficiency bar that a machine must clear to be profitable. When the subsidy fell from 6.25 to 3.125 BTC, every machine's revenue per unit of power fell by half, and machines whose efficiency fell below the new bar became uneconomic. The older generation of ASICs — the workhorses of the previous cycle — were retired or repurposed. The fleet that remained is newer and more efficient. This is why hashrate did not collapse after the halving, and why it could rise even as margins compressed: the surviving machines are simply better.
The implication is that the mining industry has been quietly recapitalized by the halving itself. That is bullish for the survivors and brutal for the laggards, and it means the marginal cost of production has fallen, which lowers the price floor. A lower floor sounds good for miners, but it also means the market can sustain a lower Bitcoin price without triggering capitulation, which is a subtle bearish signal for price itself. The efficiency gain that protects miners' margins also removes a support from the market. These two effects are the same event viewed from opposite sides.
The second unaddressed factor is pool concentration. Network hashrate is not distributed evenly; it aggregates into a small number of mining pools that construct block templates on behalf of their constituent miners. This concentration carries a governance dimension the weekly ignores. The pool operator decides which transactions to include and in what order, which is a form of soft power over the network — the ability to censor or to extract ordering value. The weekly's risk framework does not flag this, but it belongs in any honest assessment. A network whose security is distributed across hundreds of thousands of machines is not the same as a network whose block production is coordinated by a handful of pool operators. The distinction is structural, and it does not show up in the hashrate number.
The third unaddressed factor is the growing coupling between mining and external finance. Miners borrow against machines, against future production, against hashrate itself. Hashrate derivatives and machine-backed lending have turned mining into a leveraged, pro-cyclical business. In a rising market, leverage amplifies expansion and margin. In a falling market, it forces liquidation, and liquidation is not gradual — it is a cliff. The 2022 credit crisis in mining was a leverage event, not a price event; the price fell, and the leverage turned a margin compression into a solvency crisis. The current recovery, if it is driven by price and financed by leverage, carries the same tail risk. The weekly's risk matrix mentions operational and regulatory risk but does not weight the leverage channel, which is the one that has actually destroyed miners.
I want to give the bulls their due, because the contrarian reading of this dataset is stronger than the bearish one, and intellectual honesty requires stating it.
The bulls' central claim is that the death spiral did not happen. The halving cut revenue in half, and the network did not collapse — hashrate rose, difficulty rose, and miners stayed online. This is a real and important observation. The doomsayers predicted a cascade of miner shutdowns, a hashrate collapse, and a security crisis. None of that occurred. Instead, the network absorbed a 50% revenue cut and continued to grow. That is a demonstration of resilience that no amount of bearish framing can erase.
The bulls' second claim is that the price floor is real and rising. The 365-day moving average at $80,000 has held, and the marginal cost of production anchors it. This, too, is supported. The fact that price bounced from a level consistent with marginal production cost, and that hashrate rose rather than fell at that level, is evidence that the floor is not imaginary.
The bulls' third claim, and the most interesting, is that the halving's supply shock has not yet fully propagated. Daily issuance is now roughly 450 BTC, or about $37.8 million at current prices. Against a market where spot ETFs absorb capital daily and long-term holders accumulate, that issuance is small. If demand is even modestly inelastic, the halved supply is a persistent tailwind, and the miner revenue recovery is an early signal of it. Arbitrage exists only in structural inefficiency — and a halved, inelastic supply against steady demand is exactly the kind of structural imbalance that produces durable repricing.
I think the bulls are right about the mechanism and wrong about the timing and the magnitude. The mechanism — halved supply, resilient hashrate, a real cost floor — is sound. But the recovery it produces is capped by dilution, because the same hashrate resilience that prevents a spiral also ensures that any margin expansion is competed away. The bulls are describing a floor, not a launchpad. And the distinction between a floor and a launchpad is the difference between a value investment and a speculation.
The weekly's own hidden-information observations support this tempered reading. Difficulty will rise as the new hashrate is absorbed, pressuring hash price. The $40 reading is a local high. The fee share is unstable and dependent on a speculative activity that has cooled. Each of these is a cap on the upside, not a floor under it. The recovery is real. The extrapolation is not.
The most important thing the weekly does not say — and the thing a risk framework should say — is that the miner revenue recovery is a lagging indicator of the Bitcoin price, and lagging indicators are useful for confirming what has already happened, not for predicting what comes next. The 78% rebound confirms that price rose. It does not tell us that price will keep rising. It tells us that miners, who are structurally the most price-sensitive cohort in the market, had a profitable window. That window is valuable precisely because it is finite.
What should the reader do with this? Three things.
Watch the 365-day moving average at $80,000. It is the joint life line — technical support and marginal cost of production in the same number. A sustained break below it does not merely signal weakness; it triggers the negative feedback loop the bulls think cannot happen: miners go underwater, some shut off, hashrate falls, and the cost floor itself drops, removing the support that made the break look bounded. The loop is slow and the participants are levered, which is the worst combination.
Watch hash price, not revenue. Revenue is a price restatement. Hash price is the unit economics, and unit economics are what determine whether a marginal miner survives. A rising revenue headline accompanied by a flat or falling hash price is a dilution event dressed as a recovery. The $40 reading should be tracked against the next two difficulty adjustments, because those adjustments will tell you whether the recovery is being competed away.
Watch the fee share. A rising fee share is the only signal that Bitcoin's security budget is becoming self-sustaining. A falling fee share, against a shrinking subsidy, is the long-term structural risk that no short-term price rally can resolve. Twenty-one percent is not a fee market. It is a rounding error with ambitions.
The mining sector will present this quarter's numbers as vindication — proof that the industry can absorb a halving and keep growing. It can. But absorbing a shock and thriving are different states, and the data supports the first, not the second. The question worth carrying forward is not whether miners survived the halving. They did. The question is whether an industry whose revenue is a linear function of an exogenous price, whose cost floor is falling as its fleet improves, and whose fee channel is shrinking can ever be described as structurally healthy — or whether it is simply a leveraged, pro-cyclical bet on Bitcoin that happens to have survived one more cycle.
The ledger will answer that question the same way it always does: slowly, then suddenly, and without regard to what anyone believed. Ledger integrity precedes market sentiment. It also, eventually, overrides it.