Three Pools, One Template: The Post-Halving Hashrate Squeeze Nobody Is Pricing

Over the past 90 days, three mining pools have signed roughly 61% of all Bitcoin blocks. Not hashrate — blocks. The distinction matters, because block signing is the exact point where transaction selection happens, and almost nobody is watching that layer. I have run surveillance on this surface for the better part of a decade. My job is not to forecast price. It is to find where the order book is lying, where the flow is synthetic, and where a structural assumption has quietly stopped being true. This is one of those moments. Let me show you the plumbing. The April 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC — a 50% haircut to the single largest revenue line for every miner on the network. What did not halve: electricity contracts, ASIC amortization schedules, debt covenants, and the physical cost of relocating machines to wherever the marginal kilowatt-hour is cheapest. The result is a hashprice — revenue per unit of hashrate per day — that has spent most of the past two years compressing toward the operating cost of the least efficient rigs still plugged in. When hashprice sits near the shutdown threshold for the marginal miner, you do not get a smooth equilibrium. You get a purge. Network hashrate has kept climbing into the squeeze, which is the counterintuitive part. Hashrate rising while hashprice falls means capital is being deployed at negative marginal returns — typically the signature of listed miners defending production numbers for quarterly reporting, or of machines that are already sunk cost and will run until they physically fail. Difficulty adjustment lags this by design. Two-week epochs mean the network does not reprice to a hashrate shock until well after the shock has already done its damage to margins. That lag is where miners die. Miner treasury behavior confirms it. Public miners have been net sellers into strength for consecutive quarters, and the flow routes through OTC desks rather than spot — which suppresses the visible price impact while the supply quietly changes hands. That is not capitulation. It is structured distribution. That is the setup. Now the consequence. Here is the mechanical chain most coverage skips. A miner does not build its own block template unless it runs a full node and its own template engine. Most hashrate — the overwhelming majority — points at a pool via Stratum, and the pool constructs the candidate block: which transactions go in, in what order, and which are excluded. The entity that constructs the template controls transaction inclusion at the margin. Not consensus. Consensus is untouched. But the practical censorship surface on Bitcoin is not the node layer. It is the template layer, and the template layer is consolidating. I reconstructed pool attribution across a rolling 90-day window using coinbase output tagging and stale-block analysis. Three pools dominate. The tail — solo miners, regional pools, template-marketplace experiments — has thinned measurably. Three structural drivers explain it, and none of them are ideological. Payout mechanics. Large pools offer predictable payout schedules and credit lines. Small pools cannot smooth variance. In a post-halving margin regime, variance is what kills you, not the mean. Fee market access. Post-halving, fees are a larger share of revenue than at any point in Bitcoin's history. Pools running sophisticated template engines capture more fee revenue per block than pools that don't. Miners follow the fee capture. Hashrate derivatives. Hashrate financialization routes through a handful of counterparties, and those counterparties have balance-sheet relationships with a handful of pools. Liquidity begets liquidity. The second point is where my microstructure instinct fires. The Bitcoin fee market is quietly becoming an order-flow auction, and the auction is run by three desks. When blocks are full, template construction is a knapsack problem: maximize fee per weight unit. Whoever solves that knapsack best earns more. That is not decentralization. That is a performance race, and performance races have winners. Template fingerprints are the tell. Two pools can share a name and build different templates; one pool can run two names and build identical ones. I do not track brands. I track structure. Stratum V2 was supposed to fix the template layer by pushing template construction back to individual miners. Adoption is real but shallow — the pools that benefit most from central template construction have the least incentive to deploy it, and the mining firmware ecosystem has been slow to expose it. Meanwhile the security budget math is uncomfortable. Subsidy declines geometrically; fees must eventually carry the network. In a bear regime, fee revenue is thin, and thin fee revenue means the marginal block is worth less than the marginal cost of producing it for a growing share of the network. Here is the angle you will not read in mainstream coverage. Everyone worries about hashrate concentration because they have been trained to worry about a 51% attack. That framing is stale. A 51% attack on Bitcoin is economically irrational — you would destroy the asset you are attacking, and the hardware is not fungible enough to hide the act. The real exposure is subtler and more durable: template-level exclusion is cheap, deniable, and legally defensible. A pool does not need 51% to delay a transaction. It needs to be the pool that happens to build the next few blocks, and to have a compliance policy nobody audits. This is not theoretical. Post-2024, several large pools have publicly described transaction filtering tied to sanctions screening. Filtering is defensible when it is narrow. Filtering becomes a structural problem when three entities control the majority of templates and the filtering logic is opaque. Based on my audit experience across exchange and custodian infrastructure, I can tell you how this plays out: policy creeps. It starts with OFAC-listed addresses. It ends with "we exclude certain transaction patterns." Nobody votes on it. There is no governance proposal. It becomes a line in a template engine. I have seen this exact creep pattern in exchange listing policies and custodian withdrawal rules. The first version of the policy is always narrow and defensible. The fifth version is not, and nobody can point to the moment it changed. Liquidity doesn't vanish when a market turns structurally fragile — it relocates to whoever is willing to price the risk. On Bitcoin, the risk being repriced right now is not hashpower. It is template diversity, and no instrument prices it. The uncomfortable synthesis: Bitcoin's base layer is consolidating into fewer template producers at the same moment its scaling layers are fragmenting into more execution venues. Those two trends are not opposites. They are the same trend — capital and control concentrating at the layers that are hardest to observe. Now layer on the other half. While Bitcoin's template layer consolidates, the rollup ecosystem keeps fragmenting. Dozens of live Layer2s compete for the same addressable user base, and that user base is not growing at the rate the architecture implies. The result is not scaling. It is slicing. Here is the forensic tell. Bridge net flows on the long tail of rollups are flat to negative, while TVL denominated in native tokens holds roughly steady. That combination is a Red Flag: dollar value is being propped up by token price, not new capital. Strip the price effect and sequencer revenue does not cover data availability cost. Sequencer economics make it worse. Most rollups still run a single sequencer with a documented escape hatch that has never been stress-tested in production. The decentralization roadmap is a document, not a deployment. EIP-4844 was supposed to change that equation by making blobspace cheap. It worked — DA costs collapsed — and it changed almost nothing about user growth. Cheaper scaling does not create demand. It just makes the fragmentation cheaper to sustain. Arbitrage is the market's way of telling you where real liquidity sits. When the same asset trades at a persistent basis across fifteen rollups, that basis is not fifteen opportunities. It is one fragmented pool being arbitraged by bots with faster bridges than you. The scarce resource was never blockspace. It is users. And the rollup boom has been an exercise in dividing a fixed number of users across a growing number of execution environments. Watch what happens when incentives taper. Most of these ecosystems are paying for activity with their own emissions. When emissions fall, the TVL that was rented walks — and it walks faster than it arrived. Three signals, in order of importance. First, the fee-to-subsidy ratio. If fees climb as a share of miner revenue, template construction becomes a competitive business and concentration accelerates. If fees collapse, the marginal miner shuts down and hashrate consolidates through attrition instead. Either path ends in fewer pools. Second, template diversity. Track distinct template fingerprints, not pool names. Pool names lie; template structure does not. Third, net bridge flows on Layer2s outside the top three. If the long tail keeps bleeding deposits while valuations hold, the repricing is mechanical, not narrative — and it will not be gradual. Surveillance is not about predicting the break. It is about knowing which assumption is load-bearing before it fails. Right now, two of them are.