Right now, on the Bitcoin network, three mining pools are ordering the majority of every block that comes out of the difficulty algorithm.
Foundry USA. Antpool. ViaBTC. On the rolling seven-day window, those three routinely sign more than half of all blocks found. Not "more than half of one pool's share." More than half of the entire chain's block production, aggregated.
That is not a headline the desks are running this quarter. The bull market is loud. The ETF flow tape is loud. Nobody wants to stare at who actually sequences transactions on the base layer — which is, functionally, the same question regulators spent two years asking about Ethereum rollups and never got a clean answer to.
I pulled the pool attribution data off my own node and cross-checked it against two explorers that still publish parent-pool tags. The concentration is not new. What is new is how little it costs to sustain.
Pulse on the chain, breath in the market. The chain is telling a story the tape refuses to price.
The fourth halving cut the block subsidy from 6.25 BTC to 3.125 BTC. That is a 50% instantaneous haircut to the largest single line on every miner's income statement — not a gradual slope, a cliff edge, and it landed on a network that had spent the prior eighteen months pricing in exactly that cliff.
Hashprice is the number that matters here: revenue per petahash per day, blended across subsidy and fees. Before the halving it sat near triple digits. After, it compressed into the mid-forties to mid-fifties, with violent excursions in both directions depending on whether fee revenue showed up.
And fee revenue showed up, briefly.
The Runes launch and the Ordinals wave after the halving pushed average fees into triple-digit territory for a stretch. There were blocks where fees accounted for more than 70% of the reward. Miners who had survived on subsidies suddenly had a fee market. Then the wave receded, and fee share of the block reward collapsed back under 5% for most of the following quarters — punctuated by short-lived spikes that made the fee curve look like a heartbeat monitor with a bad connection.
The design intent is clean. When subsidy halves and fees are lumpy, the marginal miner — older ASICs, retail electricity rates, no hedging desk — switches off. Weak hands capitulate. Hardware migrates to lower-cost jurisdictions. The network self-corrects.
Except hardware does not evaporate. It changes custody.
Every rig that a small operator sells during a margin squeeze ends up in a hosting contract somewhere. And hosting contracts, in practice, are pool contracts. Foundry's US footprint, Antpool's Asia-Pacific depth, and ViaBTC's blended book have all absorbed machines that used to point at smaller independent pools. Network hashrate climbed through the whole cycle anyway — because the compute didn't leave, it re-registered.
That is the mechanical part. Here is the part that should keep you up.
Pool concentration is the oldest critique in Bitcoin and the least examined. The standard rebuttal is that pools are not miners. Individual hashers can point their rigs elsewhere with one config change. Attribution is a vote, not a marriage.
That rebuttal was strong when switching costs were near zero. It has gotten softer.
Three things changed, and none of them are technical. Hosting arrangements now bundle power contracts with pool defaults — you get industrial rates because the facility already has a pool relationship, and repointing means renegotiating the power, not just the config file. Firmware and fleet-management dashboards ship with a default pool baked in, and defaults are destiny. Hashrate derivatives and forward contracts get written against pool participation, which turns "which pool do you point at" into a financial term rather than a technical preference.
None of that is a conspiracy. It is just the boring gravity of industrial consolidation. But it means the phrase "hashrate is decentralized because anyone can point a rig anywhere" describes a world that existed eight years ago.
Running where the liquidity flows fastest — and right now the liquidity in mining flows toward three balance sheets.
Consider what a pool actually controls. It builds the block template. It decides which transactions are in and in what order. It decides whether to include the OFAC-screened set. It sets the payout scheme — FPPS, PPLNS, or a proprietary blend — which determines how much variance each individual hasher absorbs. That is four levers of real power, and the miner on the other end sees a dashboard, not a control panel.
Stratum V2 was supposed to return template construction to individual miners. The protocol exists. The adoption does not. Two years after the spec matured, the share of hashrate running job negotiation locally remains in the low single digits, because implementing it requires operational effort that a pool-hosted operation has no incentive to spend.
Now stack that against the second layer.
The bull market has been generous to Bitcoin-adjacent scaling. Stacks, Babylon's restaking experiments, BitVM-adjacent bridging research, and a dozen BTCfi projects all raised on the thesis that the base layer is the settlement truth and scaling belongs above it. Fine thesis. I wrote about parts of it approvingly.
But go and find me a production L2 on Bitcoin, or on Ethereum for that matter, where a single sequencer is not the de facto ordering authority. Go on. I'll wait.
Based on my time on a surveillance desk, I have watched sequencer status pages go down and watched the correlation move. When one rollup's sequencer stalls, its bridge TVL bleeds within minutes — not because the money is at risk, but because the exit queue is. That is the tell. A system with credible decentralized ordering does not develop a fear reflex around a single process restart.
A sequencer's revenue model is trivially simple: it orders transactions, it captures the spread, and it sells that ordering to whoever pays. There is no fraud proof fast enough to make a difference in the fifteen seconds that matter. There is no escape hatch that a retail user will actually find before the panic window closes. The escape hatch is a slideware promise dressed as a technical guarantee.
"Decentralized sequencing" has been a slide in a deck for two years. The upgrade path keeps getting rescheduled behind whatever ships revenue this quarter. Meanwhile the sequencer operator collects the ordering rent, decides the inclusion order, and is one legal entity in one jurisdiction.
You can call that a scaling solution. You can also call it a single centralized node with a very good PR department.
There is a third layer to this that nobody links to the first two, and they should.
Governance delegation.
The same consolidation logic runs through DAOs. Token holders do not research proposals. They delegate to whichever delegate has the loudest feed presence, and the top delegates accumulate voting power without accumulating any corresponding accountability. I have watched this happen in real time — proposals pass with 60%+ participation from fewer than a dozen wallets, most of them delegated proxies.
Pull the delegate list of any large DAO and sort by voting power. You will find a power law: the top decile controls the majority of live votes, and a meaningful slice of that weight is itself delegated from another ten thousand wallets that have never signed a transaction beyond the initial claim. Delegation was pitched as making governance accessible. It made governance rentable.
That is not governance. That is a representative system with no elections and no recalls.
And here is the coupling that matters: the entities best positioned to run industrial-scale mining hosting, operate rollup sequencers, and hold delegated governance weight are, overwhelmingly, the same short list of well-capitalized players. Concentration in one layer reinforces concentration in the others, because capital that dominates hardware also dominates ordering and also dominates votes.
Now overlay the ETF era. The same institutional capital buying spot Bitcoin exposure through custodial wrappers is, in several cases, the same capital with mining financing desks and rollup treasury positions. The instruments are different. The phone book is not.
Three pools. A dozen sequencers. A handful of delegates. Different ledgers, same shape.
Here is where I will argue against my own prior.
The cleanest assumption in crypto is that concentration is fragile — that a dominant pool is one regulatory subpoena or one mispriced block template away from losing share. History partially supports that. Pools have died. Pools have been censored and lost share.
But "temporarily" is doing a lot of work in that sentence.
Censorship at the pool level has never triggered a sustained hashrate exodus. It triggered a public argument, a few blocks of unusual ordering, and a quiet return to baseline. The enforcement mechanism — miners voting with their feet — turns out to be a mechanism almost nobody uses, because switching pools costs money and complaining costs nothing.
So the real risk is not a sudden 51% attack. The real risk is slow, boring, permanent ossification: a base layer whose ordering is dominated by three accountable-in-name-only entities, scaling layers whose state transitions depend on a process one company runs, and governance systems where the votes were decided before the proposal was posted.
None of that is a headline. All of it is a balance sheet.
Sensing the tremor before the earthquake hits is the entire job. The tremor is currently in the attribution data, not the price chart.
Watch three things over the next two quarters. Pool share drift on the 30-day window — not the seven-day noise, the trend. Sequencer uptime disclosures, specifically whether any production rollup publishes an actual multi-sequencer failover that has been stress-tested in the wild, not in a testnet demo. And delegate concentration in the three largest DAOs by treasury size — if top-10 delegate share keeps climbing while participation stays flat, the governance decentralization narrative is finished.
Seventy-two hours without sleep, zero doubts: the market is pricing Bitcoin as a settlement network and paying a premium for it. The question nobody is asking is who gets to decide what settles.
