Hashprice prints a cycle low. $42.60 per petahash per day — down 41% from the April pre-halving baseline of $72.30. The tape is not trading that number. It is trading ETF flows, a macro calendar, and a six-week range that has refused to break in either direction.
The second number matters more. Foundry USA, AntPool, and ViaBTC constructed 61.4% of all block templates across the trailing 1,000 blocks. Not 61.4% of labeled hashrate — 61.4% of actual block construction, template by template, coinbase by coinbase. The distance between those two measurements is where the decentralization argument quietly dies.
Signal confirms. Action required.
Context: The Subsidy Cut Nobody Priced
April 2024. The block subsidy drops from 6.25 BTC to 3.125 BTC. Consensus models — mine included — called for an 8% to 12% hashrate contraction within two difficulty epochs. The contraction arrived late, arrived shallow, and then reversed outright. Network hashrate now sits near 780 EH/s, a record, while revenue per unit of compute has fallen off a cliff.
That divergence is the story. Compute keeps entering. Revenue per unit keeps falling. Miner economics are not converging toward equilibrium. They are converging toward consolidation.
The mechanism is mechanical, not sentimental. A miner with a fixed power contract, an amortized ASIC fleet, and a debt covenant does not shut down when hashprice falls below operating cost. It shuts down when the lender forces it, when the power contract expires, or when the hardware can be redeployed to a cheaper jurisdiction. Everything else is theater.
Which is why the marginal miner is not a strategic actor. It is a price taker with a three-month runway.
The macro backdrop does not help. Spot BTC products have absorbed supply steadily, but not at a pace that offsets new issuance plus treasury liquidation. I spent the weeks before the January 2024 ETF approval parsing the SEC's comment letters on the Fidelity and BlackRock filings — the custody language, specifically — and the conclusion then was the same as it is now: institutional absorption is a slow, compliance-gated process, not a bid. It does not catch a falling hashprice.
The result is a market that chops. Chop is not indecision. Chop is positioning. In a range, the edge does not come from direction. It comes from knowing who is forced to sell.
Right now, the forced sellers are identifiable.
Hashprice is the cleanest single metric here because it collapses block subsidy, transaction fees, and network difficulty into one number per unit of compute. It is the miner's version of a funding rate. When it compresses for two consecutive difficulty epochs without a corresponding hashrate exit, you are looking at deferred capitulation, not resilience.
The current range is not an accident of liquidity. It is the price discovery phase of a supply overhang — new issuance, miner treasury, and unlocked creditor distributions all meeting a bid that arrives in scheduled tranches through ETF creation baskets. There is no single catalyst that resolves that. There is a duration.
Core: The Template Layer Is the Control Layer
Here is the part that gets skipped in every "number of nodes" chart.
Bitcoin miners do not construct blocks in the way the colloquial model implies. A miner running a Stratum client receives a block template from its pool. The pool decides which transactions enter that template, in what order, and which get excluded. The miner hashes the header. The miner does not curate the block.
Template construction is the actual control surface of the protocol, and it is concentrated in three entities.
This is not a theoretical concern about a 51% attack. A 51% attack requires an attacker to spend real capital destroying the value of the asset it is attacking. Nobody holding 780 EH/s of exposure does that. The realistic risk is quieter and already normalized: template-level filtering of specific transaction classes by pools operating under specific regulatory regimes.
Two of the top three pools by construction share are operated by entities with direct exposure to jurisdictions that have demonstrated a willingness to compel transaction screening. Neither has been observed filtering. That is not the same as being unable to.
The attempted fix exists. OCEAN's DATUM protocol and BIP 310 push template construction back to individual miners, allowing local transaction selection. Adoption is marginal — a rounding error against 61.4%. A governance fix that requires miners to voluntarily surrender pool-provided latency optimization will lose to latency every time.
The latency race explains the concentration better than any conspiracy theory. Pools compete on stale-rate reduction — the fraction of submitted shares that arrive too late to count. A pool with better block propagation and a denser relay network pays miners a slightly higher effective rate for identical hashrate. Over twelve months, that spread compounds. Miners migrate. The share chart slopes upward, and everyone involved is behaving rationally.
Foundry USA runs the largest US-facing operation and inherited a substantial share of the North American hashrate migration after 2021. AntPool and ViaBTC sit on the other side of the same industrial logic, with cheaper capital and lower coordination costs. None of them is doing anything wrong. That is precisely what makes the outcome durable — no villain is required for concentration to persist.
I watched this exact failure mode before. During my 2017 audit of early state-channel prototypes on the OmiseGO testnet, I found a liveness assumption buried three layers deep in the spec — sound on paper, unenforceable in production, because the party expected to enforce it had a direct economic incentive not to. The code shipped. The assumption did not hold. Same pattern, different layer.
Concentration analysis is the same discipline I applied to BAYC holder wallets in 2021, when a 15% single-syndicate position preceded a 40% floor move. The lesson transfers: measure the distribution, not the headline. A network with 20,000 reachable nodes and three template builders is a network with three template builders.
Core: Fee Revenue Stopped Covering the Gap
For most of 2023 and early 2024, fee revenue was a meaningful subsidy. Inscription and Runes activity pushed the fee share of miner revenue above 30% on peak days. Block space was scarce. Mempool cleared in hours, not blocks.
That window closed.
Fee share of total miner revenue has settled back into the 4% to 8% band. Mempool depth is thin enough that a standard-priority transaction clears within two blocks at 2 sat/vB. There is no congestion premium left to harvest.
Gas spike imminent — on the wrong chain. On Bitcoin, the congestion premium is gone, and with it the last variable that could have made marginal miners whole without subsidy.
Strip the fee revenue out and the arithmetic is unforgiving. At $42.60 hashprice and a fleet efficiency of 25 J/TH, breakeven power cost sits near $0.071 per kWh. Industrial contracts in Texas and West Virginia clear below that line. Contracts in Norway, Germany, and parts of Kazakhstan do not. It is not difficult to guess which fleet retires at the next difficulty retarget.
The jurisdictional variable is underrated in this calculation. A miner paying $0.055/kWh and a miner paying $0.09/kWh can run identical hardware and reach opposite conclusions about whether to keep hashing at the current print. Concentration is therefore not a hardware story. It is an electricity and regulatory story, and both of those inputs are political.
Core: Capitulation, Measured Properly
The hash ribbon has not inverted. That is a lagging confirmation signal, not a leading one. The leading signals are balance-sheet signals.
Publicly listed miners have three levers: sell treasury BTC, issue equity through at-the-market facilities, or draw on credit. Two of the three are dilutive, and the third is expensive at current rates. Watch ATM utilization disclosures, not the hashrate chart. When a miner issues equity into a sideways tape to fund operations, it is telling you its treasury runway is shorter than its debt maturity.
Floor holding. Momentum shifting.
Which brings in the second-order trade nobody is pricing. Miner treasury liquidation does not hit the spot tape evenly. It hits over-the-counter desks first, gets absorbed into structured products, and reaches the lit market as a slow drip. That drip is running right now, quietly, into a range.
Difficulty retargets every 2,016 blocks and adjusts to the trailing hashrate average. The lag cuts both ways. When hashrate exits, difficulty follows about two weeks behind, which briefly restores margin for survivors. That margin window is where the next round of consolidation is decided, because the operators who can hold through the compression are the ones who capture the recovery.
The DeFi side has a mirror image worth watching. I flagged the mechanism in 2020 during the Uniswap V2 liquidity mining cycle, and it has not changed: incentive programs buy TVL, and when emissions stop, the TVL leaves within a single epoch. The number on the dashboard was never users. It was rent. Bitcoin miners are running a structurally identical playbook with a different subsidy source — the block reward. When that rent falls below operating cost, the fleet does not migrate toward sustainability. It stops.
The reflexivity is the same pattern I shorted in 2022 when the Terra peg mechanism came apart. In both cases, the exit condition is endogenous: the thing that keeps the system running is the thing that runs out.
Contrarian: The 51% Conversation Is the Wrong Conversation
Every institutional research note on Bitcoin security opens with the same chart — hashrate up and to the right — and closes with the same conclusion: the network has never been more secure.
That conclusion is a non sequitur. Hashrate measures the cost of rewriting history. It does not measure the cost of censoring the present.
The realistic attack surface is a compliance directive, not a mining farm. A pool that filters a specific address class at template construction produces no orphaned blocks, triggers no consensus alarm, and leaves no forensic trace, because an excluded transaction is indistinguishable from a transaction that was never broadcast. There is no observable difference between "no one sent it" and "three pools agreed not to include it."
Decentralization that has never been stress-tested is not decentralization. It is an untested assumption with a 61.4% concentration ratio.
The same oversight applies one layer up. Layer 2 sequencers across the major rollups remain single-operator nodes under the hood, with decentralized sequencing on the roadmap for two years running. Everyone knows it. Nobody prices it, because the sequencer has never been down long enough to matter. That is not a security argument. That is a historical accident.
And the frequently cited security budget panic — the projection that fees must eventually replace the subsidy — is directionally correct but analytically lazy. It treats miners as price setters. They are not. They are price takers on a commodity with a fixed supply schedule. The subsidy will fall. The fleet will compress. Construction share will concentrate further, because the surviving operators are the ones with the cheapest power and the deepest balance sheets — which is exactly the outcome that makes template-level control durable.
The market is debating whether Bitcoin is safe. The market is not debating who builds the blocks.
Takeaway
Three things to watch, in order of signal density.
Fee share of miner revenue at the next difficulty retarget. Below 10% and the subsidy-dependent fleet compresses again.
Pool construction share across the next 2,016 blocks. A move above 65% for the top three is a structural threshold, not a headline.
Public miner treasury disclosures. Equity issuance into a range tells you more than any hashrate chart.
Arb window closing. Execute.
The question is not whether Bitcoin survives its fourth halving. It is whether anyone notices that three entities decide what survives on it.