Two Pools Now Mint Half of Bitcoin: The Post-Halving Concentration Signal

Two pools. Fifty-one percent of blocks. Hashprice at record lows. Execute. Over the trailing 30 days, Foundry USA and AntPool have together produced more than half of every block added to Bitcoin's chain. That is not half of advertised hashrate on a dashboard. That is half of the actual block templates that became canonical history. The subsidy is 3.125 BTC per block. The fee share of miner revenue has slid back under 5% on most days since inscription demand cooled. And difficulty keeps adjusting upward into an operator base that cannot absorb it. That combination is the signal. Not a price signal — a structural one. The market is filing it under "mining, boring" while it quietly rewires who decides which transactions make it into the next block. The concentration is not ideological. Post-halving economics force the smallest, most leveraged operators to sell rigs or rent hash to whoever settles fastest. That counterparty is almost always the largest pool. Concentration here is a mechanical output of the halving. Mechanical outputs are tradeable. Bitcoin's fourth halving cut the per-block subsidy from 6.25 BTC to 3.125 BTC. Every halving does the same thing: it halves the revenue attached to a fixed unit of work and forces the cost curve to reprice. The difference this cycle is where the cost curve landed. Hashprice — the dollar value of one unit of hash per day — printed repeatedly into territory where the least efficient fleets operate at or below cash cost. For operators running older air-cooled silicon on retail power contracts, the marginal block is a loss-making block. They do not power down because they want to. They power down because the electricity bill clears before the block reward does. When they power down, three things happen in sequence. Hashrate leaves the network. Difficulty eventually adjusts down, restoring margin to whoever remains. And the hash that did not physically leave — the hash that got rented, repointed, or sold — flows toward pools offering the most predictable payout. That last step is the structural one. A miner choosing a pool is not choosing a philosophy. A miner is choosing a payout scheme, a latency profile, and a fee. Full Pay-Per-Share pays on submitted shares whether or not the pool finds a block. That is a guarantee, and guarantees cost money. When margin compresses to zero, operators buy the guarantee. The pools that can underwrite it are the pools with the balance sheet. Scale begets scale. There used to be a buffer here. Transaction fees. During peak inscription activity, fees covered a meaningful slice of block revenue and kept marginal operators solvent through the subsidy cut. That buffer has thinned. Fee share has fallen back to single digits on ordinary days, which means the network's cost curve now rests almost entirely on the 3.125 BTC subsidy and on hashprice. Neither is forgiving. So the headline number — two pools above 50% — is downstream of an accounting choice made by thousands of distressed operators. That is why it is sticky. It will not mean-revert because someone writes a governance post about decentralization. The ratchet is economic, not cultural. Arb window closing. Execute. Now the technical layer most coverage skips. A pool does not mine. A pool aggregates. The operator runs a node, builds the block template — the ordered list of transactions plus the coinbase — and hands that template's header to miners via Stratum. Miners grind nonces against the header the pool chose. Whoever finds a valid hash broadcasts the pool's template. The pool takes the credit. The miner takes the payout. Read that again. The pool chooses the transactions. The miner chooses only the nonce space. Hashrate can be spread across a million machines and still produce a single entity's block template. That is the gap between hashrate decentralization and block production decentralization, and it is not a rounding error. It is the entire governance surface. If three operators build the templates, three operators decide whether a given transaction is included, delayed, or filtered. That is not a theoretical attack on consensus. It is a live property of the current production layer. I have watched this layer up close. When I audited early rollup prototypes out of Seoul, the recurring failure mode was never the cryptography. It was the operator sitting between users and settlement — the sequencer, the relayer, the coordinator. Centralized chokepoints wrapped in decentralized branding. Bitcoin's mining layer has the same shape, just with better thermodynamics. The pool is the sequencer. The template is the ordering. And "decentralized block production" has been a slide deck for as long as "decentralized sequencing" has. The counterargument is Stratum V2. Stratum V2 lets the individual miner construct the template while the pool merely coordinates. On paper, that moves template authority from pool back to hasher. In practice, adoption is thin — because the pool's business model assumes the template is the pool's, and the miner's business model assumes not caring. FPPS pays the same whether the miner picks the transactions or not. There is no line item on a miner's P&L for "template sovereignty." Until there is, Stratum V2 stays a conference topic. The concentration feeds itself through a second channel: the ASIC supply chain. When hashprice craters, the secondary market floods with used machines. The buyers with the cheapest capital — again, the large operators with power contracts and balance sheets — absorb that supply. New-generation silicon goes to whoever can pay for it and site it. Siting requires power. Power requires relationships. The barriers compound. The difficulty adjustment is the only automatic stabilizer in the system, and it lags. It retunes every 2,016 blocks — roughly two weeks — against a hashrate that moves in hours. During a capitulation cascade, that lag is the window where the marginal operator bleeds. The pool with a balance sheet survives the lag. The solo operator does not. Every cascade, the survivor list gets shorter. I ran the numbers across the last difficulty epoch to test whether the concentration is cyclical or structural. The test is simple. If it were cyclical, pool shares would oscillate with hashprice — distressed miners drifting out, then returning when margin recovers. What the data shows instead is a ratchet. Shares move up and stay up. Miners do not migrate back to small pools when margin returns, because the payout guarantee is worth more than the ideological premium. The ratchet clicks one direction. Signal confirms. Action required. Here is the part that should concern anyone holding BTC for its decentralization premium. That premium is priced into the asset's valuation thesis. If block production is functionally controlled by a handful of operators, the premium is mispriced — not because those operators are malicious, but because the cost of coordinated action has fallen below the level the market assumes. Coordination is a phone call, not a conspiracy. And the pool that declines the call loses hashrate to the pool that takes it. That asymmetry is the whole game. Let me be precise about the risk vector, because the lazy version is wrong. The lazy version says a pool could 51% attack Bitcoin. That is not the live risk. A sustained majority attack would destroy the attacker's own revenue stream. The live risk is softer and cheaper: transaction filtering at the template level, below the threshold of consensus failure. A pool does not need 51% to delay your transaction. It needs only to build the template without you. With two pools above 50%, the coordination cost of a filtering regime is a fraction of what a decentralization premium implies. This is where my audit history sharpens the read. In the early rollup testnet work, the exploitable surface was never the state-transition math. It was the operator's discretion over which transitions got posted on-chain. The math was sound. The discretion was not. Bitcoin's template layer is that same story at planetary scale. The proof-of-work is sound. The discretion is concentrated. Here is the unreported angle. Everyone is watching the hashrate pie chart. Almost no one is watching the difficulty adjustment schedule against the pool-share ratchet, and no one is pricing the interaction. The consensus narrative treats pool concentration as a governance problem to be solved by education and Stratum V2 adoption. That framing is wrong in a way that matters. The concentration is not a bug awaiting a fix. It is the terminal state of a mining market that has been financialized. Hash is rentable. Hashprice has derivatives. When you can rent a majority of hash for a known dollar cost, the number of miners was never the security metric. The cost of attack was. And the cost of attack is now a function of hashprice, which is a function of the subsidy, which halved. So the real blind spot is this: the market treats mining concentration as a decentralization story when it is actually a derivatives story. Hashprice forward curves now price miner capitulation the way they price everything else. If you want to know whether the concentration ratchets further, do not read governance forums. Read the hashprice forward. It is telling you the marginal operator's breakeven sits above spot, which means more hash flows to the pools that can underwrite the wait. Gas spike imminent. Wait. Signal confirms. Watch the next difficulty adjustment. If Foundry and AntPool hold above 55% of blocks through it, the decentralization premium baked into BTC's valuation thesis needs a reprice — and the trade is not short BTC. It is long the operators who own the templates, and short the narrative that the templates are neutral. Floor holding. Momentum shifting. The question is not whether two pools control half the blocks. It is whether the market notices before a third one joins.