On September 9, a CryptoQuant weekly note reached the trade press with one headline: Bitcoin miners had stopped selling. The supporting numbers read clean. Miner wallet balances had fallen from roughly 64,000 BTC to 51,000 BTC, then stabilized. Daily miner revenue climbed from $27 million to $48 million. A Puell-style sustainability gauge flipped from "extremely underpaid" to "reasonable compensation." By the afternoon, the summary circulating was "supply pressure removed."
I ran the arithmetic before I ran the narrative. Miners earned $48 million a day. Transaction fees contributed roughly $275,000. That is 0.57 percent. The remaining 99.43 percent was block subsidy, valued at a spot price that had risen 45 percent. Remove the price move and revenue growth is zero. What is left is a revaluation of a fixed issuance schedule, not a change in demand.
That is the whole finding. The rest is decoration.
I do not trust the pitch; I audit the structure. The structure here says miners did not decide anything. They responded to a variable.
Context
Miner flow analytics is a mature discipline, not a frontier. Glassnode, CryptoQuant, and IntoTheBlock have tracked miner-associated wallets for years, clustering addresses by coinbase transaction patterns, pool payout graphs, and known pool tags. The method is standard. What varies is the packaging.
CryptoQuant's framing is the industry's default framing: miners as a supply cohort whose behavior predicts price. The logic runs clean in a bull market. Post-halving, miners produce roughly 450 BTC a day. If they hold, that supply does not reach the order book. If they sell, it does. Track the wallets, infer the pressure, price the pressure. It is appealing because it converts an opaque macro question โ where is Bitcoin going โ into a countable one: how many coins moved.
The September note told readers the coins stopped moving. Miners shifted from net distribution to net neutrality. Balances that had been declining since early summer flattened. The Puell Multiple, which divides daily miner dollar revenue by its own 365-day average, moved out of the "underpaid" band. Hashrate recovered from a drawdown of roughly 18 percent to roughly 13 percent. And in a detail that got less attention, Satoshi-era miners cut their selling by about 70 percent.
The tone was measured. The conclusion was not.
To understand why, you have to hold two clocks in your head at once. The first is the halving clock. Every 210,000 blocks โ roughly four years โ the block subsidy is cut in half. The April 2024 halving took it to 3.125 BTC per block. Each halving mechanically halves the issuance revenue of every miner on the network while their cost base โ electricity, hardware amortization, labor, debt service โ stays fixed. The only variable that can absorb the shock is price. This is the core design tension of Bitcoin mining: the protocol cuts your income on a schedule and bets you can find a higher price before the cut becomes fatal.
The second clock is the market clock. Price moves faster than cost. When price doubles in a year, every miner's margin expands, and the marginal seller โ the operator running the oldest hardware at the highest power cost โ stops needing to sell. That is not capitulation ending. That is the pressure valve opening because the input changed.
The historical "miner capitulation" signal is real, and it is a bottom marker, not a top marker. When price falls below the cost of production for a large fraction of the network, miners sell inventory to service debt, hashrate drops as machines go dark, and the forced selling exhausts itself. That exhaustion has historically coincided with bear-market floors. The signal inverts when price rises. The miners who were forced sellers become voluntary holders. The September note is describing the second state, and describing it as if it were the first.
There is also a source problem. Every number in the report traces to a single provider. CryptoQuant shares this weekly note exclusively with one outlet, which means there is no cross-verification against Glassnode, IntoTheBlock, or an academic dataset before publication. I am not accusing anyone of fabrication. I am noting that a data vendor has a commercial incentive to publish reports that generate attention for its paid subscription product, and that an exclusive arrangement removes the one check that would catch a framing error before it propagates. Here the structure is a single narrator with a product to sell.
And there is a data-quality flag the trade press ignored entirely. The report places a balance figure of roughly 64,000 BTC at a date that appears to postdate the September publication window โ a timeline inversion that cannot be resolved from the text. That may be a typo for the prior year, a dating error in the source, or a transcription fault in the summary. I cannot tell from here. But I can tell you what it means procedurally: when a dataset contains an internal contradiction that no one downstream caught, the error rate of the rest of the dataset is unknowable. Data integrity is not a per-figure property. It is a property of the pipeline. A pipeline that ships a visible contradiction has an unmeasured error budget everywhere else.
Core
Let me build the case against the conclusion in the order the data allows.
First: the revenue is a price artifact. Miner income is a function of two variables. Issuance is fixed by protocol at 3.125 BTC per block. Price is not fixed by anything. When a report says miner revenue rose 78 percent, it is describing a price move with a formula attached. The fee line is the only portion that reflects network demand, and the fee line barely moved โ from $195,000 to $275,000, a 41 percent increase on a base so small it is a rounding error against the total. Fees remain far below their historical peaks. The network's organic demand for block space did not improve. It stayed where it has been since the inscription-driven fee spikes faded. By the common working threshold analysts use, a healthy mining economy needs fees to carry 30 percent or more of miner revenue to survive the next halving without total dependence on price. The current figure is 0.57 percent. That is not a warning sign. It is a structural fact the bull case declines to price, because pricing it means admitting that Bitcoin mining's revenue model is a leveraged bet on price and nothing else.
Second: the Puell metric is a reflex, not a diagnosis. The Puell Multiple is mechanical. Revenue up, multiple up. A 45 percent price rally lifts the numerator of every miner's revenue calculation, and the multiple follows arithmetically. Reading "reasonable compensation" off a metric that is defined by price, immediately after a price rally, and calling it evidence of miner health is circular reasoning dressed as analysis. It tells you the price went up. It does not tell you the cost structure improved. The break-even band โ the level below which a large share of miners become forced sellers โ sits somewhere in the $65,000 to $75,000 range, implied by the same metric flipping from "underpaid" to "reasonable" across a $58,000-to-$83,000 span. That is a 15 to 20 percent cushion between spot and forced selling. Thin does not begin to describe it.
Third: the wallet clustering is a black box. This is where I stop trusting the number and start asking how it was made. "Miner-associated wallets" is a definition, not an observation. Does it include exchange-hosted mining-pool addresses? Does it strip internal transfers between a miner's own wallets? Does it count coins moved to a custodian? CryptoQuant does not disclose its clustering algorithm. Neither do its competitors. But a methodology that cannot be independently reproduced is a methodology that cannot be independently trusted, and here the methodology matters more than usual. The same report notes that in August, miners moved BTC to NYDIG, a regulated custody provider. A coin moved from a self-custodied miner wallet to a custody address registers on-chain as an outflow. It is not a sale. It is a relocation. If the September "stabilization" reflects miners parking coins with a custodian rather than selling them, the chain shows flat miner balances while the coins have entered a financial system that can lend, hedge, and eventually liquidate them off-chain. The sell pressure did not vanish. It changed venue.
Liquidity is a mirage; solvency is the only truth. A miner wallet balance is a liquidity reading. The solvency question โ can this operator service its debt at a given price โ is not visible on-chain at all. I learned this the hard way. In 2020, during DeFi Summer, I spent three months simulating impermanent-loss scenarios for a liquidity-mining protocol advertising 5,000 percent APY while my colleagues chased the yield. The math said the return was unsustainable and structurally equivalent to a slow rug. I wrote a forty-page memo. The firm ignored it. The position lost 60 percent when the mechanism unwound. The lesson was not that I was right. It was that the visible number โ the APY, the wallet balance, the daily revenue โ is a surface reading, and the surface reading is where narratives are built and where they fail.
Fourth: the custody migration is the real signal, and it cuts both ways. Miners moving to NYDIG is not a neutral event. It marks the financialization of mining. A custodied coin can be pledged as loan collateral. It can back a hedging program. It can sit inside a yield product. None of that appears as a spot-market sale, and all of it can become one the moment a lender calls. This is why "miners stopped selling" is a category error: the metric captures direct sales and is blind to structured ones. As custody penetration rises, the on-chain miner balance becomes a progressively weaker proxy for actual sell pressure. The signal degrades precisely as it becomes more popular. That is a general law of on-chain analytics, not a quirk of this report. Every metric that attracts capital gets gamed, and the gamed version replaces the honest one.
Fifth: the honest line in the report is the one the bulls skipped. Buried in the note is an admission that miners have not yet begun accumulating. They stopped selling. They did not start buying. That distinction is the entire trade. "Stopped selling" is a passive state you fall into when price exceeds cost. "Started accumulating" is an active decision to hold capital in reserve against a thesis. The first is a consequence. The second is a conviction. The report confirms the consequence and defers the conviction to a future it cannot verify. A trade built on "stopped selling" is a trade built on the absence of an action, and absence of action is the weakest form of evidence there is.
Then the sample-size problem on the Satoshi-era cohort. The report celebrates a 70 percent reduction in selling by early miners. That cohort holds roughly 590,000 BTC at an effectively zero cost basis. But the September reduction amounts to about 600 BTC. Six hundred coins out of a 590,000-coin position is statistical noise wearing a narrative costume. I am not disputing that early miners sell less than they used to. I am disputing that a single 600-coin month tells you anything about intent. The signal-to-noise ratio does not support the weight placed on it. And I would want to know how the clustering separates genuine Satoshi-era coins from early-adopter wallets that merely look old. The report references excluding Patoshi-pattern coins โ the non-standard nonce signature attributed to Satoshi's own mining โ which is reasonable hygiene but also an admission that the cohort boundary is drawn, not discovered. A boundary you draw is a boundary you can move.
Sixth, and most important for anyone trading this: the report is a description of the past sold as a forecast of the future. The event it documents โ miners stopping their sales โ is downstream of the price move that already happened. The market has already paid for it. By the time the note circulated, the 45 percent rally was in the tape and the "supply pressure removed" conclusion was the explanation, not the cause. This is what I call a results narrative. It explains why the price did what it did. It does not tell you what happens next. In my experience, results narratives cluster in the middle and late stages of a move, not at the start, because that is when journalists go looking for a fundamental to attach to a price that has already run. The report is not a leading indicator. It is a lagging one wearing a leading indicator's clothes.
What the report cannot see is the one thing that would matter most: leverage. There is no funding-rate data, no open-interest data, no options-skew data in the note. Without those, you cannot tell whether the miner-holding state is a genuine reduction in supply or a levered position that will unwind violently. A miner holding coins while carrying debt is not a holder. It is a forced seller waiting for a margin call. The absence of derivative data is the single largest blind spot in the entire analysis, and it is the blind spot the bull framing depends on.
Contrarian
Here is what the bulls got right, and I will give it full weight, because the reflex to dismiss everything is its own failure mode.
Hashrate recovery is a harder signal than wallet balances. Hashrate is capital expenditure made visible. A miner does not plug in machines, sign a power contract, and run ASICs in an environment of no confidence. The drawdown narrowing from 18 percent to 13 percent, read as capacity returning rather than capitulation, is a genuine medium-term vote of confidence by the operators who carry the cost. It is a costly signal, and costly signals are the only ones I weight. Talk is cheap. A power purchase agreement is not.
The financialization argument also has a legitimate bull reading I under-weighted on my first pass. If miners migrate to regulated custody, they gain access to institutional credit lines, hedging instruments, and structured products that let them smooth revenue across the cycle. A miner who can hedge downside with options does not have to dump coins into weakness. In that world, custody is not a hiding place for sell pressure. It is a mechanism that removes it. Financialization cuts both ways, and I was too quick to assign it a single sign. The honest position is that I do not yet know which way it resolves, and neither does the report.
And the base case deserves its due. On-chain data is still the least bad instrument we have. It is noisy, it is gameable, and its definitions are opaque, but it is measured continuously and it cannot be edited after the fact. The alternative โ trading on sentiment, equity-market proxies, and influencer framing โ is worse. The correct posture is not to discard miner flow data. It is to price it at the confidence level it actually earns, which is lower than the headline implies and lower than the trade press assigned it.
There is also a regulatory angle the report skips entirely, and it is worth naming because it is the part of this system that actually touches real people. The migration to regulated custody is sold as maturation. What it also does is build a compliance perimeter around the professional miners while leaving the retail participants outside it. Compliance costs are real, and they are paid by whoever wants to stay inside the regulated lane. The operators who cannot afford custody, legal review, and reporting requirements do not disappear. They move to the edges, where the rules are thinner and the risks are higher. The same dynamic governs KYC in every corner of this industry: the requirements are nominally universal, the enforcement is not, and the cost of appearing compliant falls hardest on the users least able to absorb it. A few well-funded wallets can route around any verification regime. The miner with two racks in a garage cannot. The report does not mention this, because custody migration looks like progress until you ask who is paying for it.
Takeaway
The trade this report implies is not "buy because miners stopped selling." It is "watch the break-even band." If spot drifts toward $65,000 to $75,000, the passive state reverses, the sell pressure returns, and the custodied coins are the ones most likely to move first, because they sit inside a leverage system that reacts faster than a self-custodied holder ever would. The confirmation that would actually matter โ two consecutive weeks of net miner-wallet inflow โ has not happened. Until it does, this is a rearview mirror being sold as a windshield.
Emotion is a variable I exclude from the equation. The equation here is simple. Revenue up because price up. Fees flat because demand flat. Selling paused because selling became unnecessary, not because holding became a choice. The narrative is one step ahead of the data, and it always is. Ask not what the miners stopped doing. Ask what they have not started.
