Over the past seven days, Bitcoin's network hashrate has clawed back above 960 EH/s, recovering from a drawdown that briefly pushed it below 900. On the surface, this reads as a clean bullish print. CryptoQuant's latest weekly note frames it precisely that way: daily miner revenue up 78%, from $27M to $48M; a Puell-style sustainability metric flipping from "extremely underpaid" to "reasonably paid"; wallet balances among 100–1,000 BTC operators stabilizing after a 20% drawdown; and selling from Satoshi-era miners down roughly 70%. The narrative assembles itself. Capitulation is over, the supply overhang is lifting, and the coins are rotating into strong hands.
I have audited this exact kind of claim before, and the structure is always the same. When I pulled the fee data sitting underneath the headline revenue figure, I found a ratio of 0.57%. That single number does not dismantle the bull case. It reveals that the bull case was never about fundamentals in the first place.
Miner flow analytics is now a mature sub-discipline, dominated by CryptoQuant, Glassnode, and IntoTheBlock. The methodology is consistent across vendors: cluster wallet addresses believed to belong to mining entities, track net balance changes, and cross-reference against revenue proxies such as the Puell Multiple. The approach is legitimate. The data is on-chain, not conceptual. What varies is granularity and transparency, and both matter more than most readers assume.
The CryptoQuant report under discussion does one thing well. It segments by holder size — 100–1,000 BTC addresses, modern large operators like MARA and Riot, and the Satoshi-era cohort — rather than collapsing all miners into a single aggregate. This segmentation is the analytical highlight. It lets us see that the 100–1,000 BTC cohort shed 20% of its balance and then stabilized, while the old-guard cohort reduced selling by 70%, a cohort whose holdings total roughly 590,000 BTC.
The Puell Multiple itself deserves scrutiny. It divides daily miner dollar revenue by its own 365-day moving average. That construction means any sharp price rally mechanically lifts the metric, independent of cost structure. A 45% price spike does not require miners to become more efficient; it only requires them to be paid more dollars for the same hash. Reading the metric as evidence of improved miner health conflates revenue with profitability, and the two are not the same thing.
History offers a clean parallel. During the 2021 NFT boom, I watched valuations detach from any measurable utility, and I argued then that digital scarcity without enforceable rights was speculation, not an asset class. The miners' pause has the same texture. A data point is being elevated into a thesis because it fits the prevailing mood, not because it survives scrutiny. In both cases the analytical failure is identical: treating a symptom of rising prices as a cause of them.
But the report also carries a structural warning it does not resolve. Somewhere in the September data, a time inconsistency appears: a figure cited as "December 2025" is compared against "early September." If the report was published in September, December has not occurred. This is either a typo for December 2024, a mislabeled time field, or a publication date that does not match the data. Any of those possibilities contaminates the entire time series. A dataset that cannot agree with itself about the calendar cannot be trusted to describe a cycle.
Here is the deconstruction that matters. Miner revenue rose 78%, from $27M to $48M per day. Transaction fees over the same window moved from $195k to $275k — a 41% increase, and still far below their historical peak. Divide the fee figure by total revenue and you get 0.57%. Miners are earning less than one percent of their income from actual network demand. The remaining 99.4% is block subsidy plus price appreciation. In other words, the entire revenue expansion is a function of the BTC price rising 45%, not of the network being used more.
There is a longer structural story buried in the subsidy line. After April 2024, the block subsidy fell to 3.125 BTC. Every four years it halves again. The design intends for transaction fees to eventually replace the subsidy as the security budget. That transition is not happening. Fees at 0.57% of revenue mean the network is nowhere near a fee-funded security model, and each halving pushes miners further toward pure price dependence. The "miners are holding" narrative is not a story about confidence. It is a story about a security budget shrinking faster than the fee market can fill it.
This is the blind spot in every "miners stopped selling" headline. A miner stops selling when the price rises above their cost basis. It is a reflex, not a conviction. When BTC traded near $58k, the same Puell-style metric read "extremely underpaid." At $83k, it reads "reasonably paid." Interpolating between those two points places the aggregate breakeven somewhere in the $65k–$75k band. If price retraces 15–20%, the selling pressure does not merely return — it returns mechanically, because the incentive that suppressed it was never structural.
I built a comparable stress model in 2020, simulating Aave's liquidity pools against a 50% ETH drawdown. The lesson then applies now: a system that looks stable under current conditions can be entirely conditional on those conditions. Miner balance sheets are the same. "No obligation to sell" is not the same statement as "choosing to hold."
Then there is the methodology black box. The report never discloses its wallet clustering algorithm — how "miner-associated wallets" are defined, whether exchange mining-pool addresses are included, whether internal transfers are netted out. This matters because in August, miners moved BTC to NYDIG, a regulated US custodian. When coins move from self-custody to a custodian, on-chain analysis records an outflow. The coins were not sold. The "selling stopped" signal and the "custody migration" signal can be the same event wearing different clothes.
What the report cannot show is leverage. It contains no funding rates, no open interest, no options skew, and no data on whether miners are borrowing against their holdings rather than selling them outright. This is the largest information gap in the entire dataset. A miner who pledges BTC as loan collateral and spends the proceeds is a seller in economic terms, yet appears as a holder on-chain. Without derivative and lending data, the on-chain balance is a partial ledger at best.
Now weigh the signals by their hardness. Hashrate recovering from an 18% drawdown to 13% is a genuine capital-expenditure signal — no operator spends on electricity and rigs without a medium-term view. Satoshi-era selling down 70% is a conviction signal, though the sample is thin: roughly 600 BTC in September, statistically noisy. Against these, the wallet-balance stabilization is soft, and the "miners will become long-term holders" thesis — the report's largest promise — is explicitly conditional and unverified. The report itself admits miners are not yet accumulating. That admission is the most honest line in the document, and it quietly concedes that the bullish confirmation does not exist yet.
The consensus reading is that miner selling has decoupled from price and that supply pressure has been removed. I think the opposite is closer to true: miners have not decoupled from anything. They are more coupled to price than at any point since the 2024 halving, because the halving cut their subsidy in half while fee demand failed to replace it. When 99% of your revenue is price plus subsidy, you are a leveraged long on BTC with fixed operating costs. That is not a holder. That is a forced seller waiting for a trigger.
This is a reflexive structure, and reflexivity runs both ways. Higher prices suppress selling, which tightens float, which supports higher prices, which further suppresses selling. The loop feels like strength while it lasts. But it inverts with equal symmetry. Once price falls below the breakeven band, suppressed supply returns alongside any coins previously pledged as collateral, and the same reflex that lifted the market amplifies the decline. Positioning on a reflexive uptrend is not investing in a fundamental; it is renting one.
The "supply pressure removed" framing also overstates the marginal impact. Post-halving issuance runs roughly 450 BTC per day. Against daily spot volume, that is a rounding error. Miner selling is a second-order variable except during genuine capitulation events, and nothing in this data describes capitulation. What the data describes is a quiet period — the absence of stress, not the presence of strength.
The deeper contrarian point concerns data integrity, and it connects to a principle I have argued since 2017: code is law, but man is the loophole. The chain records what wallets do. It does not record intent. As mining financializes — NYDIG custody, borrowing, options hedging — the on-chain footprint of a miner becomes an increasingly poor proxy for their actual economic behavior. Hidden selling via collateralized loans and hedges is invisible to flow analytics. The signal degrades precisely as the sector matures. Correlation is a map, not the territory, and here the map is drawn by an algorithm nobody has published.
So position accordingly. The signal to watch is not "selling stopped" — that is a lagging artifact of a price move already priced in. The signal is net accumulation: miner wallet balances turning sustainably positive for consecutive weeks. If that appears, it is a far stronger confirmation than the current pause. If price slides toward the $65k–$75k breakeven band first, watch whether the pause holds. My suspicion is that it will not, and that the miners who look like holders today will look like sellers again the moment the reflex inverts.
