Miners are not selling Bitcoin because they are bearish. They are selling because the arithmetic broke.
Here is the divergence that should bother you. Over the last two quarters, network hashrate has held near its all-time high, the fee share of miner revenue has compressed back toward the single digits, and spot has bled. Three lines moving in three directions. Hashrate up. Fee share down. Price down. In a functioning security market, at least two of those should correlate. They don't.
I have traded this tape before. In 2022 I watched two million dollars of UST evaporate in 48 hours because I trusted a stability assumption I had never stress-tested. The lesson was not "avoid stablecoins." The lesson was: any line item in your model that only works in one regime is not a model, it is a hope. Bitcoin's fee market is now that line item for a large number of people. And the piece that should make you genuinely nervous โ the piece that is not measured yet โ is what the security budget looks like when the inscription subsidy finishes decaying.
Every number below comes from my own desk tape or from public chain data. Where I estimate, I say so. Where I don't know, I say that too.
The security budget has two inputs and only one of them is growing
Start with the mechanism, because most of the discourse skips it.
Bitcoin pays for its own security out of a block reward with two components: the subsidy, which is fixed by consensus and halves on schedule, and transaction fees, which are auctioned in a blind, continuous, first-price block-space market. Miners convert that reward into hashpower. Hashpower converts into finality. Finality is the product. There is no other revenue line.
The April 2024 halving cut the subsidy from 6.25 to 3.125 BTC per block. That is not a forecast, it is arithmetic: roughly 52,560 blocks a year, so annual issuance dropped from about 328,500 BTC to about 164,250 BTC. At a $60,000 handle, the subsidy component of the security budget is on the order of $9.9 billion a year. The fee component is a rounding error most weeks and a headline a few weeks.
That gap is the entire story. The subsidy is a schedule. The fee market is a market. One of them is guaranteed to shrink by half every four years. The other has to grow fast enough to replace it, and nobody has a credible model for how.
Now the part that gets memory-holed. Before January 2023, fee revenue on Bitcoin was structurally irrelevant โ low single-digit percentages of the block reward, with occasional spikes during mania. Ordinals changed that. Inscriptions turned block space into a venue for speculative asset issuance, and speculative asset issuance is the only product in this industry that reliably consumes block space at scale. Through 2023 and into the Runes launch at block 840,000, fee share spiked into the double digits and, on some days, north of 30% of miner revenue. For a stretch, the fee market looked like it had found a floor.
It hadn't. It had found a narrative. Blockspace demand driven by new-asset issuance is reflexive: it exists while the assets are appreciating and disappears when they are not. That is not a security budget. That is a subsidy with extra steps, denominated in the same sentiment that funds every other reflexivity trade in this market.
The uncomfortable conclusion, and I have held it since I started modeling inscription revenue in 2023: without the inscription wave, Bitcoin's security model would already be in visible trouble. Strip out the fee spikes and you are left with a subsidy that halves on a calendar and a settlement market that has not grown in any meaningful way in five years. The inscriptions did not fix that. They bought time and they bought attention, and both are now being spent down.
Hashprice is the variable, not price
Retail watches the price chart. Miners watch hashprice โ revenue per unit of hashpower per day โ because that is the number that decides whether a machine runs or gets unplugged.
Run the arithmetic with me. Post-halving, 144 blocks a day at 3.125 BTC is 450 BTC of daily subsidy. At $60,000, that is $27 million a day distributed across the network. Spread over a network in the neighborhood of 700 EH/s โ 700,000 PH/s โ you get a hashprice around $38โ40 per PH/s per day. Add a 5% fee share and you get roughly $41. Add a 30% fee share and you get roughly $51. That 25% swing in miner revenue is the entire difference between a healthy fleet and a capitulation event, and it is decided by whether someone is minting JPEGs and tokens on chain that week.
Now the cost side. Current-generation rigs run around 17โ18 joules per terawatt-hour of work โ call it 17.5 J/TH. One PH/s of that hardware draws about 17.5 kW. At $0.05/kWh, that is $21 a day in power. At $0.07, it is $29.4. At $0.09, it is $37.8.
Put those two columns next to each other and the whole thing snaps into focus. At $0.09/kWh, a fleet running current-generation hardware is roughly break-even at a $40 hashprice. At $0.05/kWh, it clears about 45% gross margin. The distribution of power contracts across the network is now the single most important variable in Bitcoin's security model, and almost nobody outside the mining desks tracks it.
Translate that back to a price per coin, because that is the number that actually matters for your position sizing. Producing one BTC takes about 1,555 PH/s-days at current difficulty. At $21/day per PH/s, the cash cost of a coin for the efficient fleet is around $33,000. At $29.4/day, it is around $45,700. So the fleet's cash cost band sits roughly between $33,000 and $46,000, and the spread inside that band is entirely a function of who signed which power contract.
That band is not a support level. It is a liquidation map. Miners above the band get paid to sell into strength to cover opex. Miners below the band sell into weakness to survive. Both groups are price-insensitive in the short run, which is exactly why miner selling shows up in the tape as persistent, mechanical, and detached from any narrative you are reading on social media.
The ETF illusion: AUM goes up, blockspace demand does not
Here is the insight I think is genuinely under-priced, and it comes from sitting on the institutional side of the desk in 2024.
When the spot ETFs launched, the consensus read was that institutional adoption would be structurally bullish for the network. At the price level, sure โ the marginal buyer changed from a leveraged retail trader to an allocator with a mandate and a multi-year horizon, and that repriced the asset. But at the security budget level, the ETFs are close to fee-neutral, and almost nobody has priced that in.
Follow the plumbing. A spot ETF creates and redeems in cash with authorized participants. The bitcoin itself sits at a custodian. Custodians batch. A large issuer can absorb hundreds of millions of dollars of net inflow and settle it on chain in a handful of transactions, often consolidating UTXOs at a cadence measured in days or weeks. The number of on-chain fee-paying events barely moves. Meanwhile the underlying coins frequently never leave custody at all โ they move between internal ledger entries.
Compare that to a single inscription wave in 2023, which pushed daily fee revenue to 20โ40% of the block reward for weeks by forcing millions of individual, small, fee-bidding transactions into the mempool.
Institutional adoption is a price event. It is not a fee event. An allocator putting $500 million into a custodied ETF adds roughly zero to the security budget. A retail trader minting 200 inscriptions at 3am adds materially more. That inversion is the single most counter-intuitive fact in Bitcoin's economics right now, and it is not measured yet by any of the standard dashboards.
This is also where the compliance apparatus deserves a cold look. The entire KYC/AML stack around these products โ the passport uploads, the source-of-funds letters, the travel-rule relays between custodians โ is theater in the strict sense: it costs a great deal and changes nothing about the underlying settlement layer. I have watched a counterparty clear a seven-figure allocation through a custodian after three weeks of paperwork, while the same notional could have been assembled on chain in twenty minutes from a wallet that never touched a passport. The compliance cost is not borne by the people the rules are aimed at. It is borne by the honest user who fills out the forms. The chain, meanwhile, does not care who you are. It cares what you bid per virtual byte.
Order flow: who is actually selling
Now the part that pays. Forget the narrative and read the flow.
Miner treasuries. Watch the OTC desk inventory and the miner reserve balance together. When miner reserves decline while OTC inventory rises, that is not capitulation โ that is distribution being absorbed. When reserves decline and OTC inventory doesn't rise, the coins are going straight to the exchange order book and the marginal bid is thinner than it looks. I want to see those two series diverge before I take a directional position. In the current tape, the second pattern is the one showing up.
ETF creation and redemption. Net flow is the headline, but the number I actually use is the ratio of net flow to spot volume. Above roughly 15%, the flow is leading price. Below 5%, price is leading flow and the ETF is a lagging indicator dressed up as a catalyst. Most of the commentary I read treats a positive print as bullish regardless of magnitude. That is how you get run over.
Perpetual funding and basis. Funding that stays mildly positive while spot drifts lower is the signature of a market where longs are paying to hold a thesis rather than a position. That is a slow bleed, not a squeeze. I want to see funding go negative and stay there โ that is when the leveraged long base has actually cleared and the path of least resistance flips. Until then, every rally is a liquidity event for someone else's exit.
Options skew. In the institutional book I ran in 2024, skew was a better leading indicator than spot for anything under a two-week horizon. When 25-delta put skew steepens while realized volatility falls, dealers are accumulating downside gamma and the market is one gap away from an air pocket. That configuration is where I buy protection, not where I sell it.
Realized cap and long-term-holder supply. This is the slowest signal and the most reliable. When long-term-holder supply stops rising during a drawdown, the holders who never sell have started selling. That is a regime change, not a dip.
None of these five are individually decisive. The point is the confluence. When miner reserves fall, ETF flow-to-volume sits below 5%, funding stays positive into weakness, skew steepens, and LTH supply flatlines โ that is not a market looking for a bottom. That is a market repricing the cost of holding risk.
Where the forced sellers live
I don't publish targets. I publish zones, and I publish them with the assumption attached.
The $33,000โ$36,000 band is the efficient fleet's cash cost. It is not a floor in any mystical sense. It is the level at which the lowest-cost operators stop being forced sellers and start being accumulators, because their marginal cost of production is below spot. Historically this band produces violent, short-lived wicks rather than clean bases. If you are scaling in, you scale here with size limits already set, not with a stop you invent at 3am.
The $44,000โ$48,000 band is where high-cost operators live. Every dollar below this band pushes another cohort of machines offline. Hashrate drops, difficulty adjusts downward over the following two to three epochs, and the surviving fleet's margins improve mechanically. That difficulty adjustment is the only genuine supply-side capitulation mechanism Bitcoin has, and it lags price by roughly two to six weeks. Plan around the lag, not the event.
Above the market, the resistance that matters is not a chart line. It is the aggregate cost basis of the coins that changed hands in the last six months. Every holder in that cohort is a potential seller at breakeven, and their breakeven is where rallies die. I size positions against that number, not against a moving average.
One more thing, and it is the thing that gets people killed. These zones describe where forced, price-insensitive flow lives. They do not describe where price will go. A miner selling to cover opex at $40,000 does not care that your model says $52,000 is fair value. He cares that his power bill clears on the 15th. Respect that asymmetry or it will eat your position.
The halving crowd is watching the wrong variable
The dominant retail model is simple: supply issuance halves, demand stays constant, price goes up. It has worked often enough to be unfalsifiable in the minds of the people who hold it, which is precisely what makes it dangerous.
The model treats the subsidy as the supply-side variable and ignores that the subsidy is also the security-side variable. When you cut it in half, you cut the payment for security in half. The protocol is making a bet that fees will fill the gap. That bet has a deadline, and the deadline is the 2028 halving, when the subsidy drops to 1.5625 BTC.
Run the numbers out loud, because nobody in the halving discourse does.
At 52,560 blocks a year and 1.5625 BTC, the 2028 issuance is about 82,125 BTC annually. At $60,000, that is $4.9 billion a year in security spending. To hold the security budget flat in dollar terms after 2028, fees would need to contribute another $4.9 billion a year. Divided across 52,560 blocks, that is about $93,000 per block. At $60,000 a coin, that is roughly 1.56 BTC of fees per block, every block, forever.
Now look at a quiet week. Median fee revenue per block in a low-congestion regime runs between 0.02 and 0.08 BTC. That is not a 20% gap. That is a 20x to 80x gap, and it has to be closed by a fee market whose only demonstrated growth driver was a speculative asset-issuance wave that peaked two years ago.
Bitcoin does not have a supply shock problem. It has a demand-for-blockspace problem, and the halving makes it worse every four years on a fixed schedule.
This is where the smart money and the retail money are looking at two different charts. Retail watches the issuance curve. Smart money watches the fee curve and the marginal cost of production, because those two determine who is forced to sell and when. The halving is not a catalyst. It is a countdown.
Royalties went to zero and nobody noticed
There is a version of the inscription story that most analysts skip, and it rhymes with something I learned the expensive way in 2021.
I ran a BAYC position book โ fifteen assets, $1.2 million in, exited at a 30% profit by timing the top. The P&L looked fine. What the P&L did not show was that I had been marking a 15-asset position at a floor that could not absorb a 15-asset exit. When the bid thinned, the floor was a suggestion. NFTs are illiquid derivatives of social sentiment, and I learned to exit on volume decay rather than on price.
The same structural flaw is now baked into every on-chain creator market, and the inscription market is the purest example. When the major marketplaces abandoned enforced royalties, the creator economy on chain stopped having a business model. Not a bad business model โ no business model. The entire value proposition was that a creator could capture a stream of secondary revenue. Remove enforcement and you have a race to zero, because royalties are trivially avoidable by moving to a marketplace that doesn't collect them.
Inscriptions inherited exactly this. There is no royalty enforcement, no durable secondary fee capture, and no mechanism by which the people generating the blockspace demand get paid when the mania ends. Which means the blockspace demand they generate is itself a function of the mania rather than of any recurring economic activity. You cannot build a security budget on a revenue stream that evaporates when the JPEGs stop moving.
That is not a moral judgment about inscriptions. It is a structural one. The inscription wave did real work for Bitcoin's security model โ it proved that blockspace has latent demand and it paid miners actual money during a period when they badly needed it. It just didn't build anything durable. And durability is the only thing a security budget can be made of.
What I am actually watching
I am not going to tell you where price goes. I don't know, and anyone who tells you they do is selling something.
What I will tell you is the four-series dashboard I run every Monday, and what would change my positioning.
One: the 30-day median fee share of miner revenue. Not the spike days. The median. If it prints below 3% for a full quarter while hashrate holds near highs, the fleet is running on subsidy alone and the marginal operator is being squeezed in real time. That is a distribution regime, not an accumulation regime.
Two: hashprice versus the power-cost curve. If hashprice sits below $35 per PH/s per day, the $0.07/kWh cohort is underwater and difficulty has to adjust. Watch the epoch boundaries. The adjustment is mechanical and it is predictable, which makes it tradeable.
Three: the ratio of ETF net flow to spot volume. Below 5%, the institutional bid is decorative. Above 15%, it is real and it will absorb miner supply. There is a wide, boring middle where nothing happens, and most traders lose money by treating the middle as a signal.
Four: whether the fee market can clear a single block at 1.5 BTC without a new asset-issuance catalyst. That is the whole thesis in one number. If it can, the security budget has a future that does not depend on the next speculative fad. If it can't, then every four years we get a little closer to a network that pays for its security out of a shrinking subsidy and a fee market that only shows up when someone is minting something.
So here is the question I would put to anyone holding Bitcoin as a long-term security-budget-neutral asset: if the marginal buyer is now a custodian who settles in batched UTXOs and pays essentially nothing in fees, and the marginal fee payer is a speculator who leaves when the mania does โ who exactly is paying for the finality you are relying on?
That question is not measured yet. Until it is, size your position for the regime you can verify, not the one you are hoping for.