Bitcoin's Decentralization Matrix

Duke H3 · Aug 28, 2026 · 23 min read

For over a decade, the health of Bitcoin's decentralization was reduced to a single number: how many full nodes are running. This was always an oversimplification, promoted by people who should have known better. And now, after a proposal failed to reach consensus, the same crowd is flipping the narrative from a fairy tale into a horror story.

The fairy tale went like this: noble, independent node runners employ the miners; therefore the system is decentralized and incorruptible. The horror story goes like this: a cabal of mining conglomerates conspired on a Saturday night to corrupt the miners and seize control of the network, and only a band of brave warriors can save Bitcoin from their clutches. So they are forking of with bitcoin knots.

But are these stories true? Or is reality more complicated than the story. Well the truth is the system's health has never depended on a single number. It has always depended on a matrix of parameters---users, miners, mining nodes, verification nodes, exchanges, software providers---all working together across distinct layers, competing within the layer, and cooperating between the layers. To understand if the system is healthy, you need to understand how those layers work, and the dynamics that drive the actors.

Bitcoin as a system

Bitcoin is best understood by starting with its fundamental relationship: the one between users and mining nodes. Satochi describes this relationship

A note on terminology. Throughout this text I use "mining node" to distinguish the block-producing nodes from other nodes in the system. Satoshi himself was not consistent---page one of the whitepaper calls them "honest nodes," while chapter 3 describes them as the "Timestamp Server." By "user" I mean anyone who interacts with the Bitcoin system in any form, whether self-custodying or through a trusted intermediary. There are currently around 200 million of them.

The relationship is straightforward. Users work to gather real world resources and seek an honest system that will store value over time. The mining nodes sequence the chain in a transparent, irreversible, and verifiably honest way, and are rewarded for this in the present through block rewards---a fixed subsidy and a dynamic transaction fee.

The mining process

At the operational core of the system lie the mining nodes. They organise miners into mining pools and sequence the chain using the hashpower they source from those miners, giving Bitcoin its true power: the ability to irreversibly order transactions in an honest, transparent, decentralized, and predictable way.

Proof of work provides intrinsic value

This proof of work, and nothing else, provides Bitcoin its intrinsic value---the one Peter Schiff keeps missing. When a user accepts a payment and it confirms, that user can be certain the coins are theirs and the network cannot take them back, neither by outright theft nor by involuntary expansion of the supply. The product the mining nodes sell to the users is transparent, decentralized, irreversible sequencing---not eternal storage.

To achieve this, the system currently expends roughly 900 million TH/s and rewards the mining nodes with approximately $200,000 every ten minutes. No other system on the planet does this. It is the harnessing of this power---and this power alone---that makes users willing to pay around $65,000 per coin.

Most of this value is passed down to the people operating mining rigs. These miners operate on a for-profit basis. The mining nodes are sovereign vis-à-vis their peers, by system design. They guard each other against misuse of the system and compete with each other for the hashpower that the miners provide. A mining node who produces an invalid block is punished by its peers: the block is not recognised, a competing block is favoured instead, and the offender loses its block reward and the long-term ability to compensate its miners for their work---risking that miners leave to sell their service to a competing mining node.

By design, the barrier to entry for a mining node is low. As the whitepaper states, the system is hyper-competitive. Even if every existing mining node were captured by a hostile cabal, new honest nodes would come online, and the miners---protecting their long-term economic interests---would migrate to them, leaving the cabal with nothing.

This is why proof of work matters. There is a persistent misunderstanding that you can change the mining algorithm at will and "fire the miners" when they displease you. You can---but the decentralized, sequenced irreversibility that makes Bitcoin valuable walks out the door with them. It does not come from the algorithm; it comes from the miners' accumulated work. Change the algorithm and you lose that work, and with it irreversibility, the core property that makes the system worth anything to users. Any system that loses its irreversibility can no longer reliably sequence blocks, will lose all value in its coin, and will irrecoverably become a dumping ground for all kinds of garbage. This has been tried and tested before---"Bitcoin Gold" and "Bitcoin SV" are two such chains that tried and failed.

Governance

Bitcoin is a proof-of-work system, and this extends to its governance. To signal for or against a change, you must mine a block---and to mine a block, you must expend real energy. There is no vote without work. The mining nodes channel this stake: they build the block template, set the version bits, and direct the miners' earned work into governance signals. The miners choose which mining node to follow; the mining node chooses what to signal. Neither can act without the other. This is how consensus is formed.

This process is formalised in BIP-9. Each proposed change is assigned a bit in the block header's version field. Mining nodes set that bit in the blocks they produce to signal support. Every 2,016 blocks---one retarget period---the network tallies the signals. If a supermajority threshold is reached within a defined window, the change activates. If not, it expires. No amount of discussion, lobbying, or node policy can substitute for this process: only mined blocks count.

Trust through verification

All of this happens under the watchful eyes of users, who are at liberty to run a verification node to observe the process. If users feel the system is misbehaving, they have two ways to punish it. The first is straightforward: sell their accumulated coins, pushing the price down and directly reducing the value of future block rewards. The second is harder but more constructive: start their own mining node and compete for hashpower.

There is a third option that has been much debated: refusing to relay transactions or blocks that do not fit the user's preferences, through their verification node. But this option is a mirage---it feels powerful but is, by design, ineffective. Verification nodes have no power to bully mining nodes by withholding their blocks from other mining nodes---the mining nodes are fully capable of sharing blocks with each other directly. And users whose transactions are censored by their peers are at liberty to run their own verification node and reach the mining nodes themselves.

What matters, therefore, is not the relay policy that individual verification nodes choose to enforce, but the consensus policy that the mining nodes negotiate among themselves. A user is of course free to take a moral stand and refuse to relay transaction X for subjective reason Y. But the system is specifically designed to route around such behaviour. This is Bitcoin's censorship resistance, designed to withstand attacks from states.

Supporting services

Software providers, exchanges, and other services form an outer layer of the system. They are important but not primary actors---mining nodes choose what software to run, users choose what exchanges to trust. The system's power does not rest with them.

There is, however, a structural question that deserves careful examination: unlike miners and mining nodes, software developers are not funded through the system's own incentive mechanism. The consequences of this gap, and what it means for the health of the system, will be the subject of a follow-up article.

Bitcoin's decentralisation matrix

This all leads to Bitcoin's decentralisation matrix, where we can model the incentives and see if there are problems ahead for Bitcoin. All estimates assume $65,000 per coin and post-halving reward of 3.125 BTC per block. Derivations are in Appendix B.

The user layer

| Actor | Role | Peers | Trend | Revenue/year | Barrier to entry | Recourse | |------------------------|---------------------------------|-------|---------|------------------|--------------------|-----------------------------| | Users | Provide value, demand integrity | ~200M | Growing | Source of value | None | Sell coins, start mining node | | Verification nodes | Observe and verify | ~15k | Stable | $0 | ~$500 | Alert users | | Software wallets | Key management, signing | ~80 | Growing | ~$50M§ (fees) | ~$50k | Users switch | | Hardware wallets | Secure key storage | ~5 | Stable | ~$250M (sales) | ~$500k | Users switch | | Exchanges | On/off ramp to fiat | ~500 | Growing | ~$30B§ | ~$1M+ | Choose what to list |

The mining node layer

| Actor | Role | Peers | Trend | Revenue/year | Barrier to entry | Recourse | |---------------------------|-------------------------------|-------|---------|------------------|--------------------| -------------------------| | Mining nodes | Order transactions, signal | ~15 | Stable | ~$200M (fees) | <$1,000 | Lose miners, lose reward | | Miners | Supply proof-of-work | ~1M | Growing | ~$10.5B (rewards)| ~$4,000/rig | Choose mining node | | Hardware manufacturers| Build mining rigs | 3† | Stable | ~$5B (sales) | ~$100M+ | Miners switch | | Software providers | Write the implementation | 2‡ | Stable | ~$10M (grants) | ~$10M/year | Mining nodes fork/switch |

§ All crypto, not Bitcoin-only. Bitcoin-specific share is lower.

† Bitmain (~60% market share), MicroBT, and Canaan. Together >95% of all ASIC production.

‡ Bitcoin Core and Bitcoin Knots (a patchset of Core). Some pools run privately patched versions of Core for block template construction.

Software distribution across node types (2025)

| Software | Mining nodes | Verification nodes | |----------------|----------------------------------|--------------------------| | Bitcoin Core | ~14 of 15 pools (~99% hashrate) | ~75–80% (~14,000 nodes) | | Bitcoin Knots | 1 pool (Ocean, ~1–2% hashrate) | ~20–25% (~4,400 nodes) |

Knots reached 25% of verification nodes but commands only ~1–2% of hashrate through a single mining node. This illustrates the article's core point: verification node counts do not translate to governance power. Only mined blocks count.

What actually happened on Saturday

On Saturday 8 August 2026, block 961,632 was reached---the point at which BIP-110 mandatory signalling activated. After this block, Bitcoin Knots nodes were programmed to automatically reject non-signalling blocks.

The people believing in the fairy tale fully expected the mining nodes to start signalling for BIP-110. Their stated reasoning was twofold: by withholding relay, verification nodes could incentivise the mining nodes to "flip" the bit; and BIP-110-compliant miners would reject non-signalling blocks, in effect forcing either the submission of 97% of the hashpower to their will or a trial by hashpower. The theory, in their own words, was that "game theory will make the miners comply." They cited the SegWit activation years ago as precedent, carefully curating the story to fit their narrative.

For anyone observing from the sidelines who had read the statistics above, what they were attempting was clear: running a lorry off the road in a game of chicken while sitting in a Trabant. There was only one possible outcome.

The contest

At 19:35 UTC, AntPool mined block 961,632---the first block after activation---without signalling BIP-110. The main chain had spoken.

BIP-110 chain: Roughnecks, channelling hashpower through Ocean's BIP-110 DATUM endpoint, rejected AntPool's block and mined their own competing block 961,632, formally splitting the chain. Roughnecks then mined a second block (961,633). Both blocks were mined at Bitcoin's full 127.48 trillion difficulty level.

Main chain (non-signalling): The rest of the network continued building on AntPool's block. F2Pool mined block 961,633 at 20:18 UTC. Ocean's own non-BIP-110 pool mined block 961,634 at 20:29 UTC. Foundry followed with block 961,635 at 21:02 UTC.

There are rumours that Foundry initially took a neutral position and paused mining operations. This is unconfirmed. What is beyond doubt is that AntPool did no such thing. Perhaps Jihan Wu wanted to take vengeance on Luke for the SegWit2x defeat back in 2017.

The outcome

No third BIP-110 block materialised. The minority chain froze---stuck at block 961,633 with the full network difficulty but a fraction of the hashpower. At approximately 03:40 UTC on 9 August, Roughnecks held a team meeting. Shortly before 04:00 UTC, they announced they were standing down, characterising the retreat as "an escalation to the next step" and advising BIP-110 miners to "stop until further notice."

Ocean's hashrate collapsed from approximately 36 EH/s to 1.25 EH/s---a 96.5% drop. By the time Roughnecks lowered their colours, the main chain was 111 blocks ahead. Not a single block on the main chain had signalled BIP-110.

Today, all of the mining nodes are sequencing the non-BIP-110 chain. For Bitcoin's purposes, this chapter is closed and consensus is restored.

What remains is drawing the conclusions and writing history.

This base on a theory that the syncronising nodes could be blackmailed into activating the signaling bit by withholding transit. As described above, this is a scenario that the system is specificly designed to withstand. The minors, fully aware that the only hash power that signaled support for bip-110 was half of Ocens miners or about 1% hash rate chose to do nothing. And continue as usual. The only desnting mining group roughnecks chose to activate bip-110 and hence ignore any non signaling blocks. This pitched them as the desending part with about 1% of what we assume is bought (rented) hash power. This brougt them into a short hash war against the rest of bitcoins combined hashpower. While there are roumors here and there about what actualy went on, the result was never in doubt.

There are rumours that the mining nodes conspired against the system. This does not have to be the case.

Mining nodes select whatever version bits they choose when building a block, and miners can follow whichever nodes they want. If a node signalling the desired bits does not exist, new nodes can come online---Ocean, for example, allowed miners to select their own block template. This is a setup very far from a world of dark figures on Zoom calls dictating terms.

The proposed change---BIP-110---simply had not reached consensus, and was far from it. The matrix tells us why. BIP-110 achieved support from ~25% of verification nodes (running Knots), but only ~1–2% of hashrate (Ocean). Under BIP-9, only mined blocks count. No amount of verification node adoption can substitute for mining node signalling, and no amount of mining node signalling can substitute for mined blocks.

If the mining nodes had acted differently---forcing miners onto a minority chain---they would have lost the miners in a heartbeat, and the block reward with them. The "cabal" could have had as many Zoom calls as they liked. Any action other than what we observed would have produced the same outcome, with additional punishment for the dissenting mining nodes.

The system worked exactly as designed.

Conclusion

Bitcoin's decentralization is not a single number you can point to. It is a matrix: users providing value, mining nodes ordering transactions, miners supplying proof-of-work, and verification nodes independently auditing the result. Each layer constrains the others. Each layer has its own form of sovereignty and its own limits.

The system is as healthy as it has ever been---not because any one group is virtuous, but because the incentive structure across all layers makes honest behaviour the only profitable strategy.

Appendix A: Active mining nodes (August 2026)

Total network hashrate: ~1,000 EH/s (one zettahash). Data from mempool.space and pool self-reporting.

| Pool | Hashrate share | Software | Notes | |----------------|---------------|---------------|----------------------------------------| | Foundry USA | ~25% | Core (patched)| US-based, backed by DCG | | AntPool | ~20% | Core (patched)| Bitmain-owned | | F2Pool | ~13% | Core | One of the oldest pools | | ViaBTC | ~11% | Core | | | SpiderPool | ~9% | Core | | | MARA Pool | ~5% | Core | Marathon Digital | | Braiins Pool | ~3% | Core | Formerly Slush Pool (first ever, 2010) | | Luxor | ~2% | Core | US-based, professional focus | | Binance Pool | ~2% | Core | Exchange-integrated | | Ocean | ~1–2% | Knots | Luke Dashjr, only Knots mining node | | EMCD | ~1% | Core | | | Clover Pool | ~1% | Core | Formerly BTC.com | | Poolin | <1% | Core | | | SBI Crypto | <1% | Core | Japanese | | Others | <1% | Various | Small and solo pools |

The top 4 pools control ~70% of hashrate. By design, this concentration is not a systemic risk: the barrier to entry for a new mining node is under $1,000, and miners will migrate to any honest node that offers competitive terms.

Appendix B: How the numbers were calculated

Miner revenue (~$10.5B/year)

Post-halving block reward is 3.125 BTC. At 144 blocks/day: 3.125 × 144 × 365 × $65,000 ≈ $10.7B. This excludes transaction fees, which add roughly 5–10% depending on network congestion.

Mining node revenue (~$200M/year)

Pool fees are typically 1–2% of block rewards passed through to miners. At ~$10.5B in total rewards: 1.5% × $10.5B ≈ $160M. Some pools earn additional revenue from MEV-style transaction ordering. $200M is a rough upper estimate.

Hardware manufacturer revenue (~$5B/year)

Only one of the three major manufacturers is publicly traded:

| Manufacturer | 2025 Revenue | Source | |-------------|-------------|-------------------| | Canaan | $530M | SEC filing (audited) | | Bitmain | ~$3–4B (est.)| Private. ~60% market share, estimated from Canaan's share | | MicroBT | ~$1B (est.) | Private. No public financials |

Bitmain and MicroBT are private companies and do not publish financial results. The estimates are derived from Canaan's known revenue and relative market share positions. The ~$5B total should be treated as an order-of-magnitude estimate, not a precise figure.

Software provider funding (~$10M/year)

Bitcoin Core development is funded through grants and employment by several organisations:

| Organisation | Core-specific funding | Developers | |-------------------|-----------------------|------------| | Brink | ~$1.5–2M | ~8 | | Chaincode Labs | ~$1.5–2.5M | ~6–8 | | Spiral (Block) | ~$2.5–3M | Core portion of ~$10M budget | | Localhost Research | ~$2–2.5M | ~5–6 | | OpenSats + others | ~$2–3M | Various |

Total: ~$10–12M/year funding ~30–40 active developers producing ~2,500 commits/year. Bitcoin Knots is maintained by a single developer (Luke Dashjr).

Source: Bitget analysis of Core funding, NobsBitcoin 2023 report.

Appendix C: Who owns the hashpower?

Most mining nodes (mining pools) are service providers, not miners. They provide the mining node infrastructure and charge 1–2% fees. The hashpower itself is owned by thousands of independent mining operations that can switch pools at will. This distinction reinforces the article's argument: mining nodes do not own the work, they organise it.

| Pool | Hashpower ownership | Notes | |----------------|-------------------------|--------------------------------------------| | Foundry USA | Mostly external | Aggregates large US miners (Marathon, Riot, etc.). Backed by DCG | | AntPool | Mixed (vertically integrated) | Operated by Bitmain, who manufactures ~60% of ASICs. New Antminers ship pre-configured for AntPool. Research suggests several smaller pools route hashrate through AntPool infrastructure | | F2Pool | Mostly external | Aggregates independent miners | | ViaBTC | Mostly external | Aggregates independent miners | | SpiderPool | Mostly external | Aggregates independent miners | | MARA Pool | Mostly self-mining | Marathon Digital's own machines | | Braiins Pool | Mostly external | Formerly Slush Pool. Pool service provider | | Luxor | Mostly external | Pool service provider | | Binance Pool | Mostly external | Exchange-integrated, aggregates users | | Ocean | Mostly external | Allows miners to select own block template |

Exact ratios of self-owned versus external hashpower are not published by any pool. The above classifications are based on public reporting and known corporate structures.

The Bitmain/AntPool vertical integration---the company that builds the hardware also runs one of the largest pools---is the most significant concentration of power in the mining node layer. However, the miners pointing their machines at AntPool are still free to switch to any competing pool at any time.

The three-layer model

Bitcoin's core architecture is a relationship between three sovereign entities arranged in layers:

| Layer | Entities | Role | |--------------------------|---------------------|---------------------------------------------| | User layer | ~200--300M users | Provide value and demand integrity | | Mining node layer | Mining pool nodes | Order transactions into irreversible blocks | | Mining layer | Hashpower operators | Supply the proof-of-work the mining nodes need |

Each entity is sovereign relative to its peers within the same layer but dependent on the layers above and below it. Users depend on mining nodes to order their transactions. Mining nodes depend on miners for hashpower. Miners depend on users for the value of the block reward.

Notice what is absent: software developers, verification nodes, and other supporting services. These are important, but they are not primary actors. The system's fundamental health does not depend on them, and their influence is, at most, secondary. They appear elsewhere in the matrix---as an outer defensive layer---but not at the core.

The user layer

Users sit at the top of the system. They work in the real world, acquire bitcoin, and reward the system with their trust. Their belief in the future integrity---and hence value---of the system is what funds the mining nodes in the present.

This matters because there is a persistent misunderstanding that you can change the mining algorithm at will and "fire the miners" when they displease you. The decentralized, sequenced irreversibility that makes Bitcoin valuable does not come from the algorithm; it comes from the miners' accumulated work. Change the algorithm and you lose that work, and with it the core property that makes the system worth anything to users.

The mining node layer

At the operational core of the system lie the mining nodes, run by competing mining pools. These nodes sequence the chain using the hashpower they source from miners, giving Bitcoin its true power: the ability to irreversibly order transactions in an honest, transparent, decentralized, and predictable way.

This, and nothing else, is Bitcoin's intrinsic value---the one Peter Schiff keeps missing. When a user accepts a payment and it confirms, that user can be certain the coins are theirs and the network cannot take them back, neither by outright theft nor by involuntary expansion of the supply. This is fundamentally the product that the mining nodes are selling to the users. Transparent, decentralised irreversible sequencing, not eternal storage.

The sovereignty of mining nodes is bounded. They are not sovereign against the miners below, whose work they depend on to find the next block, nor against the users above, who provide the value behind the rewards.

To achieve sequencing, the system currently expends roughly 900 million TH/s and rewards the mining nodes with approximately $200,000 every ten minutes. No other system on the planet does this. It is the harnessing of this power---and this power alone---that makes users willing to pay around $65,000 per coin.

The mining layer

Most of the value that mining nodes receive from users is passed down to people operating mining rigs. These miners are loyal only to their pocketbooks, and they have demonstrated this loyalty on many occasions. Act dishonestly, and the wrath of the market delivers instant punishment.

Miners crave a return on investment that typically takes years to recover. No mining node could hope to retain mining power if it acted against the interests of users---say, by inflating the supply or forcing through a soft fork that lacks consensus. The punishment for doing so is immediate and terminal.

Supporting services

The verification layer

Bitcoin is not a "trust me, bro" system. The mining nodes work under the watchful eyes of thousands of verification nodes operated by the users that relay transactions and provide an important outer layer of defense: keeping invalid transactions out and valid transactions in, regardless of who wants to stop them.

Verification nodes do not mine, so they cannot punish mining nodes by forking them off. But the users who operate them do have their hands on the sell button, and they serve as canaries in the coal mine---an early warning system that the core layers are behaving correctly.

The distinction matters: the verification nodes' job is not to steer the system. Their job is to verify it, primarily for the benefit of the individual node runner. You run a node so that you can independently confirm the system is working as specified. If you don't like what you see, the most potent signal you can send is to sell your coins.

Trying to block a transaction that complies with consensus---even if your friends agree with you---is, by design, a futile effort. This is Bitcoin's censorship resistance.

An important clarification: verification nodes do not relay blocks between mining nodes. Mining nodes are operated by professional organizations with direct, high-speed links to each other. Verification nodes ferry transactions from users to mempools and new blocks from mining nodes to users---from the periphery to the core and back, not within the core itself.

A group of verification nodes, no matter how coordinated, cannot prevent a new block from propagating between mining nodes. To believe that verification nodes can blackmail mining nodes by threatening to withhold relay unless certain version bits are set is to fundamentally misunderstand how the system works. This is not a power the system bestows upon them.

What actually happened on Saturday

There are rumours that the mining nodes conspired against the system. This does not have to be the case.

The proposed change simply had not reached consensus, and was far from it.

Conclusion

The system is as healthy as it has ever been---not because any one group is virtuous, but because the incentive structure across all layers makes honest behaviour the only profitable strategy.