Tracing the immutable breath of the Bitcoin timechain, I found something odd on the morning of 31 July. SBI Crypto's 24-hour average hashrate had fallen to 0.452 EH/s. A month earlier, on 30 June, the same metric read 16.222 EH/s. In between, the seven-day average had slipped to 5.817 EH/s, a single-month decline of roughly 64 percent. Then, after 29 July, the pool simply stopped producing blocks. The numbers did not look like an attack. They looked like a protocol for shutdown.
The obvious interpretation is self-defeating: another mining pool dies, therefore centralization wins. But the raw data refuses to cooperate. SBI's exit did not create the concentration that everyone is now pointing at. On 20 July, when SBI was still connected to its miners, the three largest pools - Foundry USA, AntPool, and F2Pool - had already claimed 64.8 percent of attributed block production. On 27 July, the same group claimed 60.8 percent. The 60 percent threshold was cracked before SBI closed its doors. SBI's disappearance is not the cause of the concentration problem. It is the dust settling on it.
This is a story about how aggregate metrics mislead, how 'pool share' is a statistical artifact, and how hashrate centralization is best understood not by watching the pool that dies, but by reading the block templates of the pools that remain. Sitting in my Hong Kong office, far from any Japanese data center, I can only examine the public telemetry. But the forensic signature is clear.
The Actor and the Protocol
SBI Crypto is not a garage operation. It is the mining arm of SBI Group, one of Japan's largest financial conglomerates. For years it operated a Stratum-based mining pool, aggregating the work of remote miners, constructing candidate block templates, and distributing the bitcoin block reward when one of those miners found a valid share. This is the unglamorous layer of Bitcoin that most investors never touch. Miners sign into a pool with a simple URL and a worker credential. The pool's server becomes the coordinator. The pool chooses which transactions are eligible for inclusion, which transaction fees matter, and how the reward is split.
The technical infrastructure is mature. Stratum is not a new protocol; it has not changed in any fundamental way that affects this story. Bitcoin's core consensus layer - Proof of Work, difficulty adjustment, UTXO modeling - remained completely unchanged during the SBI wind-down. No protocol rule was violated. No consensus bug was triggered. The only thing that changed was a service provider deciding to stop offering a service.
Yet, in the digital economics of Bitcoin, that mundane service has outsized significance. A mining pool is a coordination layer, not a consensus layer. It aggregates fragmented physical hardware into a statistical machine. It smooths the variance of mining for small operators. It turns a lottery into a salary. When a pool closes, the physical miners do not disappear. They reconnect. They point their hashes elsewhere. The network's security boundary is not the pool's server; it is the will of those miners to keep their machines running.
SBI's closure was announced as an orderly wind-down. The pool operators did not switch off the Stratum endpoint in a panic. They staged the process, gradually decreasing the pool's effective hashrate, giving miners time to redirect their rigs. The telemetry reflects that. A pool that was producing a real share of blocks on 30 June was, by 31 July, nearly invisible. That is not a technical failure. It is an operational decision.
The market narrative, however, wants a villain. It wants to say that a Japanese giant gave up, and that Bitcoin becomes more concentrated with each defection. The forensic reality is more nuanced. SBI was already small. On 5 July, its seven-day average hashrate had fallen to about 5.817 EH/s. Before the closure had fully executed, its attributed block share had dropped to 0.72 percent, equivalent to roughly 6.8 EH/s. These are not numbers that threaten the network. These are numbers that suggest an operator running a business that no longer pencils out.
The Math of a Quiet Exit
Let me walk through the numbers, because the order of operations matters.
On 30 June, SBI reported a seven-day average hashrate of 16.222 EH/s. By 30 July, that same seven-day average had fallen to 5.817 EH/s. That is a loss of 64 percent of its reported capacity in a month. Then, on 31 July, the 24-hour average collapsed to 0.452 EH/s. That is not a pool that is struggling. That is a pool that has functionally turned off its lights. After 29 July, SBI did not produce another block. The Stratum endpoint was, in effect, dead.
The immediate impact on Bitcoin's aggregate hashrate is easy to overstate. Even at its peak, SBI's 16 EH/s represented a small fraction of a global network measured in hundreds of exahashes. The source data used by Hashrate Index and mempool.space points to an aggregate effect on the network total of less than 2.5 percent. A 2.5 percent drop in network hashrate is not trivial, but it is also not an existential event. The difficulty adjustment algorithm is designed to absorb exactly this kind of movement. If miners leave, difficulty falls. If miners return, difficulty rises. The system breathes.
What is more interesting is the direction of the math. If SBI's miners were real, they did not evaporate. Some set of ASICs that were pointing at SBI's Stratum server had to go somewhere. The switch cost for a miner is nearly zero. A miner changes a few lines in the configuration file: the stratum URL, the worker name, maybe a port number. The hardware does not care which pool it talks to. The consequence is that aggregate pool-level data cannot tell us where the hashrate went. We see the source disappear. We cannot see the destination. That statistical blind spot is one of the most underappreciated limitations of public mining data.
If a meaningful portion of SBI's 16 EH/s migrated to one of the three large pools, then the concentration problem worsened. If it migrated to smaller pools such as Luxor, then the closure might actually have improved the distribution. The telemetry alone cannot distinguish between these two futures. Anyone who claims that the 60 percent threshold is now worse because of SBI is making an assumption that the data does not support.
That is the first lesson of a forensic approach to mining metrics: never confuse an entity's reported hashrate with the physical hardware that once supported it. Reported hashrate is a measure of service level, not of machine ownership. When the service ends, the machines go elsewhere, and the story changes.
Attributions Are Not Hardware
The most dangerous number in the entire SBI story is also the most quoted. The three largest pools - Foundry USA, AntPool, and F2Pool - are said to control more than 60 percent of Bitcoin's hashrate. The arithmetic is straightforward: Foundry with 26.67 percent, AntPool with 17.13 percent, F2Pool with 16.21 percent. Summed, they total approximately 60.01 percent. It is a clean, terrible number. It fits on a dashboard. It fuels alarmist headlines.
But this number is not a measurement of physical hashrate. It is a measurement of attributed blocks. A block is attributed to a pool based on the metadata embedded in the coinbase transaction, the pool's payout structure, or the block template it created. If a pool solves a block using a template that carries the pool's fingerprint, the block counts toward that pool's share. That is a reliable enough way to track relative market share over time. It is not, however, a real-time map of who controls mining hardware.
Hashrate is probabilistic. Pool share numbers should be read as rolling averages, not as point-in-time verdicts. The 60.01 percent reading captures one instant. It could easily be 62 percent the next day or 58 percent by the end of the week. In fact, the source data shows that the top-three share was 64.8 percent on 20 July and 60.8 percent on 27 July. That is a four-point swing in one week, using the same methodology. This is not a stable state. It is a noisy signal.
The forensic problem is that these measurements do not capture the flow of miners. A pool may lose a 10 EH/s customer who moves to a rival. The losing pool's attributed block share will fall, but only after enough unlucky blocks have accumulated to reflect the change. The winning pool will see its share rise only as it wins blocks with its expanded workforce. Between those two events, the official share data can look misleadingly stable. This is the same problem that I saw when analyzing DeFi protocols after the 2022 collapse: aggregate TVL numbers were often smoothed over several days, hiding the sharp exits that had already happened.
So when the mainstream readout says that three pools control 60 percent of hashrate, I translate it more carefully. Three pools have recently been credited with a majority of solved blocks. That is a material fact. But it says nothing about whether those pools could compel the underlying miners to do anything beyond what the pool's payouts and policies reward. Hashrate concentration is a structural concern, not a command-and-control mechanism. The miners still own the hardware. They can leave as quickly as they arrived.
The 60% That Was Already There
Let me be explicit about the timeline, because the chronological detail is the crux of the analysis.
On 20 July, with SBI still nominally active, the combined share of Foundry, AntPool, and F2Pool was 64.8 percent. On 27 July, one day before the final phase of the shutdown, it was 60.8 percent. SBI's closure was not formally complete until later. The 60 percent threshold was not an aftereffect. It was the pre-existing condition.
This matters because almost every commentary piece treats SBI's exit as a new wound. In reality, the wound was already there. SBI was no longer a meaningful counterweight. Its share had fallen to 0.72 percent of attributed blocks, equivalent to about 6.8 EH/s. A pool with less than one percent share cannot be the defender of decentralization. Its departure does not break a previously balanced system. It merely removes a player who had already lost the game.
The deeper point is that the 60 percent concentration is not driven by the closure of a small pool. It is driven by the structural advantages of large pools. Foundry benefits from its American institutional brand and close ties to large asset managers. AntPool benefits from its Asian supply-chain relationships and diversified financial services. F2Pool benefits from more than a decade of operational history and a broad global miner base. These advantages compound. They are not fragile; they are the direct result of scale economies in an industry where capital costs, hardware procurement, and counterparty trust determine who survives.
A mining pool is not a protocol. It is a business. It has customers, margins, risk exposure, and a balance sheet. SBI's exit belongs to the same pattern that shaped the rest of the mining industry after the 2024 halving: block subsidies were cut in half, revenue pressure increased, and any pool with high operating costs or a weak value proposition had to reconsider its existence. The closure of SBI is an industrial rationalization story. The centralization story was already in the data long before.
Why SBI Left
The economic incentives behind SBI's decision are not mysterious. A mining pool earns its revenue through fees charged to miners, typically in the range of one to four percent of the block reward plus transaction fees. That revenue is derived from the miner's production. When the block subsidy was cut from 6.25 BTC to 3.125 BTC per block, the absolute value of each unit of hashrate dropped by half. A pool's fee income drops accordingly unless the miner's hashrate rises or the fee rate rises. In a competitive market, raising the fee rate is difficult. Miners will simply switch to a cheaper pool.
Japan added another layer of pressure. Electricity costs in Japan are notoriously high by global mining standards. Any mining operation that owns its own machines in Japan is paying a premium for power. SBI's exit is consistent with a capital allocation decision that the expected return on mining infrastructure no longer justified the operational cost. It is not necessarily a prediction that Bitcoin is doomed. It is a prediction that SBI's specific mix of hardware, power costs, and balance-sheet expectations could not compete with Foundry or AntPool.
This is why I pay more attention to the pool fee market than to the closure headlines. When a pool exits, its miners migrate. The migration gives the remaining pools an opportunity to compete. Some will lower fees to attract volume. Others will raise fees because they no longer need to compete. The likely outcome is a two-tier market: large pools with institutional miners and financialized services, and small pools that compete on transparency, data analytics, or ideological purity. That is not decentralization in the ideal sense, but it is a market finding its equilibrium.
What This Does to Security
There is a recurring confusion in the crypto media between Bitcoin's consensus security and the centralization of its mining service layer. Consensus security refers to the difficulty for an attacker to produce a chain of blocks that rewrites history or censor transactions. That security is a function of total honest hashrate, the cost of re-mining, and the economic weight of the chain. A pool closing does not directly reduce consensus security if its former miners remain active elsewhere. The hashrate does not vanish; it is redistributed.
The SBI event underscores that nuance. The Bitcoin core protocol - Proof of Work, difficulty adjustment, UTXO model - was not modified. The Stratum service was switched off, but the network's ability to settle transactions did not weaken. The difficulty adjustment algorithm will respond to whatever the true network hashrate happens to be. If the total dropped by 2.5 percent, the next difficulty adjustment will make mining slightly easier. If the total recovered because SBI's miners migrated, the adjustment will be minimal. From the perspective of transaction finality, the system remains stable.
That does not mean the concentration data is irrelevant. It means we should be precise about what kind of risk the concentration creates. A pool that controls a large portion of attributed blocks has the power to decide which transactions enter its block templates. It can choose to ignore certain addresses, certain transaction types, or certain fee levels. It can theoretically delay inclusion of a contested transaction. It can also decide whether to include Ordinals and BRC-20-style transactions or to promote a particular Bitcoin Core policy. These are real forms of power. They are economic and political, not cryptographic.
The mitigations are structural. Miners can switch pools in seconds. If a pool abuses its template authority, the miners who disagree will migrate. This is the tension at the heart of pool centralization: pools have influence, but miners have the ultimate veto. The historical track record of miners organizing against pool policy is mixed, but the technical capacity exists. The switching cost is a single config edit. The risk is not that a pool becomes a dictator; the risk is that miners become too comfortable, too slow, or too financially entangled to notice.
I have spent much of my career auditing smart contracts where the code is the final authority. Mining pools are different. They are not deterministic. They are organizations. They have employees, legal obligations, and revenue targets. The 'code' that governs their behavior is partially open-source software and partially a private server running behind an API. Silence in the code speaks louder than audits - and in this case, the silence is the absence of SBI's server. That silence tells us nothing about what the remaining pools are doing inside their private infrastructure.
The Churn Below the Top
The ranking table tells another story that the mainstream narrative tends to ignore. Foundry is solidly in first place with 26.67 percent. AntPool holds second with 17.13 percent. F2Pool is third with 16.21 percent. The three together form the 60 percent club. But below them, the field is moving. Luxor is rising. Braiins is falling. NeoPool has disappeared from the public charts. These movements are the real front lines of the mining industry.
Luxor's rise makes sense. It has positioned itself as a data-driven pool, offering hashrate derivatives and analytics that appeal to sophisticated miners. In a bear market, where margin matters more than raw optimism, services that help miners optimize their treasury are attractive. Braiins, by contrast, has historically appealed to ideological miners who value open-source software. That appeal may be eroding as the financial sophistication of miners increases. And NeoPool, which once appeared as a niche competitor, no longer holds a visible position. This is creative destruction at the small-pool level.
The presence of churn is actually a healthy sign. A market with no movement is a market with no pressure. The three large pools hold a majority share, but the minority tail is not static. New pools can gain ground by offering better data, better fee structures, or better geographic reach. The barrier to entry is not technological; Stratum software is mature and openly available. The barrier is trust and distribution. Miners have to believe that a pool will pay them honestly and will not disappear with their accrued balance. That belief is built over years of transparent behavior.
If SBI's exit teaches anything, it is that trust disappears quickly. SBI's miners did not wait for a formal announcement to migrate; their hashrate had been falling for weeks before the final shutdown. They saw the signals and acted. This is the silent discipline of profit-maximizing agents. They do not care about the narrative. They care about their next payout.
The broader market implication is that the top three pools' share is not immutable. It is a snapshot of a battle that is happening every day. Foundry's lead is not guaranteed. AntPool's second place is not permanent. F2Pool's position is not a birthright. Any of them could falter on operational security, fee structure, or regulatory pressure. The 60 percent number is a condition, not a verdict.
The Contrarian Read: The Metric Is Not the Threat
Now I want to push against the most comfortable conclusion. The comfortable conclusion is that we must fear the 60 percent concentration because it threatens Bitcoin's decentralization. My read is more contrarian: the 60 percent number, as commonly presented, is the wrong threat model.
The first flaw is methodological. Attributed block share is a lagging indicator. It is calculated from blocks that were found and broadcast with a pool tag. It does not measure instantaneous hashrate. A pool that is about to die can show a temporarily high share because its miners are running on old templates. A pool that is about to surge can show a low share because it has not yet been lucky enough to find its next blocks. Using a one-day or one-week share as a proxy for ongoing control over the network is like analyzing a decentralized exchange's liquidity based on a single block of swap events. It captures a moment, not an equilibrium.
The second flaw is the conflation of pool share with pool power. Even if a pool controls 40 percent of attributed blocks, that does not mean it controls 40 percent of miners' decisions. A pool operator cannot force a miner to continue pointing hashrate at it. The only real power a pool has is the ability to filter transactions and choose payout policies. That power is bounded by the threat of miner exit. Miners are not passive. They are watching the mempool, the fee market, and the pool's solvency. When a pool behaves badly, they leave. This is why no major pool has attempted a controversial transaction-selection policy for long without facing a backlash.
But there is a darker version of this power that deserves attention. A pool could theoretically use its template policy to censor a specific transaction or address, not because the network demands it, but because the pool's jurisdiction demands it. That would be a political act. The miners could vote with their hashrate, but the vote takes time. During that time, targeted transactions could be delayed. This is not a consensus attack. It is a legal attack. The threat is not 51 percent control. The threat is 20 percent control by a pool that is compliant with a particular government's demands, combined with a general atmosphere of not rocking the boat.
The third flaw is the tendency to treat all hashrate as equal. It is not. A pool with 20 percent of the hashrate but with direct relationships to a few large institutional miners has more coordination power than a pool with 25 percent spread across thousands of anonymous hobbyists. SBI's 16 EH/s was likely corporate-owned, not spread across retail miners. That kind of centralized enterprise hashrate is more volatile: a single board decision can remove it. Retail miners scattered across the globe are harder to switch off. The SBI event is proof of this. Its hashrate may have been controlled by one corporation, and when that corporation decided to exit, the hashrate went away overnight. A decentralized network of small miners would not have produced such a clean, sudden cliff.
This gives the story a surprising twist. SBI's closure might be seen not as a blow to decentralization, but as a subtle reminder that the most dangerous concentration is not in the public pool rankings. It is in the concentrated ownership of hardware by entities whose business models are not aligned with Bitcoin's longevity. The problem is not the pool; it is the fact that a large portion of the physical hash power is held by balance sheets that can be repriced overnight.
I saw this pattern in the 2022 LUNA/UST collapse. Everyone pointed to the Anchor Protocol's high yield as the cause. The real cause was a circular economic design that lacked a stable equilibrium. Here, everyone points to the 60 percent pool share as the cause of centralization. The real condition is a mining industry that rewards scale, compliance, and reliable cash flow. As long as that is true, the large pools will remain large, and the small pools will keep churning. SBI's exit was not a violation of that equilibrium. It was an expression of it.
What, then, should security-minded analysts actually monitor? I would monitor three things. First, the raw number of distinct block-producing entities over a rolling 30-day window. That number is less volatile and more meaningful than a single-day share. Second, the geographic distribution of miners connected to the top pools. If a single pool becomes the clearinghouse for a single jurisdiction, the legal risk profile changes. Third, the payout policies of the top pools. A pool that changes its fee structure or introduces KYC requirements for miners is a pool that is becoming more centralized in its decision-making, even if its hashrate share is flat.
None of these signals will make a dramatic headline. They are quiet, structural, and boring. But they are the kinds of signals that matter when the next SBI emerges.
A Note on the Method
I want to be transparent about the data sources. The numbers in this analysis come from public telemetry reported by SBI Crypto itself and from independent aggregators such as Hashrate Index and mempool.space. These are professional platforms that are widely used by miners and analysts. They are not flawless. Their attribution methodology relies on visible block tags, which can be spoofed or omitted. A pool can choose to run a template without a distinctive coinbase tag, in which case it would be classified as unknown. That means the official share figures may understate the true concentration. They are unlikely to overstate it.
In my own audit work, I apply a simple rule: when a data source aligns with the incentives of the person reporting it, I discount its precision. SBI reported its own hashrate. A pool that is preparing to close has no incentive to inflate its hashrate, but it also has no incentive to be perfectly precise. The decline from 16.222 EH/s to 0.452 EH/s is so steep that even a small measurement margin does not change the conclusion. SBI is gone. The top three pools dominate. The question is whether that dominance is growing and whether the dominant pools can act on it.
That question cannot be answered by a single metric. It requires looking at the metadata inside the blocks, the fee structures on the pool websites, and the regulatory climate around the mining companies. In a bear market, survival matters more than gains. The pools that survive are the ones with the deepest pockets and the most compliant operations. The miners who survive are the ones who can switch pools without letting their pride get in the way. This is not a story about technology. It is a story about industrial consolidation.
Takeaway
The quiet death of SBI's mining pool is not a red flag for Bitcoin's consensus security. It is a red flag for the illusion that the network remains meaningfully decentralized at the service layer. The 60 percent threshold was already there. The closure did not cause it. The closure merely made it more visible to the lazy dashboard.
The next crisis will not begin with a pool closure announcement. It will begin with a pool that stays open and changes its policy. It will begin with a pool that quietly updates its block template with a compliance filter, then convinces itself that the filter is ethical. It will begin with a pool whose miners have forgotten how easy it is to change a Stratum URL. When that day comes, the metadata will not save us. The only question is whether the miners still have the agility to move. SBI's miners moved. That is the only comforting sentence in this entire autopsy.