Ledgers do not lie, but liquidity always flees.
On July 31, SBI Crypto's 24-hour average hashrate printed 0.452 EH/s. On June 30, the same telemetry line printed 16.222 EH/s. In thirty-one days, the Japanese mining pool lost roughly 97% of its working hashpower. It produced no blocks after July 29. The obvious headline writes itself: another mining pool is gone, the remaining top three pools capture more than 60% of block production, and Bitcoin's decentralization scorecard takes another hit.
But the ledger does not support that story. Hashrate Index's attributed-block data shows the three largest pools held 64.80% on July 20 and 60.78% on July 27. Both readings predated SBI's formal shutdown. The 60% threshold was not crossed after the exit. It was already crossed before the exit. SBI's closure is a symptom of the existing market structure, not the cause of it.
This is the information gap in every 'SBI exits, Bitcoin centralizes' take. The concentration was in the data long before the event, and the event itself changed almost nothing at the protocol level.
The Pool Is an Accountant, Not a Consensus Node
Mining pools are not validators. They do not produce consensus; they aggregate work. A pool operates a Stratum server that distributes job parameters to miners, collects their proof-of-work shares, and builds block templates on their behalf. When a miner connects to a pool, it is not delegating custody or signing authority. It is choosing which block template proposer to work with.
The pool can decide which transactions enter the template. It can set fee policy. It can exclude a transaction if it wants. But it cannot finalize a state, confiscate mining rewards, or rewrite Bitcoin's consensus rules. Its business model is a service fee, typically 1% to 4% of the block reward.
SBI Crypto was a small but visible pool under SBI Holdings, one of Japan's largest financial conglomerates. It entered Bitcoin mining during the institutional expansion of the sector, when traditional capital wanted exposure to the asset without direct setup risk. Its infrastructure was standard, its client base unremarkable, and its share of block production rarely crossed into the top tier.
The pool's exit is not a novel event in Bitcoin's history. Pools close when miners leave, and miners leave when the economics stop working. The 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC. For a pool operator, that cut halves the gross revenue from every block the pool mines. If Japan's industrial electricity costs are added to the formula, the pool's operating margin is squeezed from two sides. SBI's decision to shut down was a business calculation, not a security alarm.
There is a deeper point that most commentary skips. The pool service layer matters because it influences which transactions get confirmed. But Bitcoin's actual consensus mechanism โ Proof-of-Work, difficulty adjustment, UTXO bookkeeping โ remained untouched while SBI was shutting down. In the audit, we find the truth that price hides: this event belongs to the mining industry, not to the protocol.
The Numbers, Read Slowly
Let's walk through the timeline carefully.
SBI's seven-day moving average hashrate fell from 16.222 EH/s on June 30 to 5.817 EH/s by July 30. By July 31, the 24-hour average had collapsed to 0.452 EH/s. In one month, the pool lost more than 97% of its hashpower.
The shape of the decline matters. A market-driven hashrate loss is usually a slow bleed: miners gradually point machines elsewhere while they search for better rates. SBI's decline was a managed extraction. The telemetry shows a coordinated wind-down, not a sudden failure. The pool's attributed block share dropped to 0.72%, which corresponds to roughly 6.8 EH/s, before the pool stopped appearing in block attribution altogether.
There is a discrepancy worth noting. The official telemetry showed 0.452 EH/s on July 31, while the attributed-block share implied 6.8 EH/s. That gap is not a data error. It is the lag between when a pool's miners disconnect from its Stratum server and when the block attribution tables stop counting its legacy templates. The official SBI telemetry only counts miners still connected to SBI's endpoint; it cannot see miners that quietly moved to another pool a day before the shutdown. The real churn is therefore slightly larger than the headline numbers, and the destination is invisible to the public data.
At peak, the global impact of SBI's exit was close to 2.5% of estimated network hashrate. That is meaningful for a single pool, but it is not a systemic shock. Bitcoin's difficulty adjustment absorbs such changes in two weeks. The network did not need an emergency rule change. It did not even need a conversation.
The Concentration Was Already There
Now consider the concentration reading that triggered the alarm.
Foundry USA, AntPool, and F2Pool held 26.67%, 17.13%, and 16.21% of attributed blocks at the latest observation. The simple sum is 60.01%.
That number is real, but the story around it is incomplete. The same three pools held 64.80% on July 20 and 60.78% on July 27. They had already crossed 60% before SBI stopped submitting blocks. SBI's exit narrowed a distribution that was already concentrated. It did not create that distribution.
The causality matters for anyone making a risk decision. If you believe SBI's exit pushed the network over a dangerous threshold, you are placing the event at the wrong point in time. The threshold was crossed while SBI was still active. The first response should have been to watch pool concentration weeks ago, not to react to the shutdown announcement.
There is also a problem with the unit of measurement. Attributed blocks are not the same as hashrate control. Attribution measures which pool template solved the block. A pool that solves 26.67% of blocks in a window likely has a large share of hashpower, but the correlation is not exact. Latency, block propagation, and temporary changes in payout schemes can create short-term distortions. A 60.01% reading is a point-in-time figure, not a permanent state of control. The threshold only matters if it persists over a mining epoch or several difficulty adjustments.
Attributed Blocks Are Not Control
This distinction is not academic. A pool that controls 26% of attributed blocks can, in theory, censor a transaction by omitting it from its template. But if it does, the response is instant: miners can point their rigs elsewhere within minutes. The pool's 'control' is the privilege of serving the best offer, not the power to compel.
The technical exit cost for a miner is close to zero. Changing pools is a one-line edit in a Stratum URL. The ASIC firmware, the mining software, and the wallet address remain the same. The pool can set its own payout frequency and fee, but it cannot lock a miner in. This is the critical difference between pool concentration and sequencer centralization in a Layer 2 system. A centralized sequencer controls ordering and can impose state-finality assumptions on users. A mining pool only proposes a block template, and miners can reject it by disconnecting.
This is also why the aggregate attribution data is incomplete. The top-three chart looks stable because the chart only reflects blocks that were actually solved. The flows that are about to change where those blocks get solved will only appear in the next several attribution windows. The next two difficulty adjustment periods will tell us far more than the last one.
The protocol layer did not change. The coinbase reward is still 3.125 BTC per block. The difficulty adjustment still targets a two-week block issuance. The UTXO set remains the global source of truth. If a pool disappears, its miners migrate, and the network keeps producing blocks. That resilience is the real story under the SBI headline. Trust the protocol, verify the exit.
The Economics of a Pool That Did Not Survive
Why did SBI fail where Foundry, AntPool, and F2Pool continue to grow? The answer is unit economics.
A pool with, say, 1% of global hashrate can expect about 1.4 blocks per day at the network's typical pace. At 3.125 BTC in subsidy plus transaction fees, each block might pay roughly 3.2 BTC. That pool's daily revenue before fees is about 4.5 BTC. With a 2% pool fee, gross revenue is about 0.09 BTC per day. At current prices, that is a few thousand dollars a day.
For an independent operator, that is a living. For a Japanese financial conglomerate with compliance overhead, litigation risk, and shareholder expectations, it is noise. SBI probably did not close because the pool was losing five dollars. It closed because the opportunity cost of running a marginal pool inside a listed financial group was too high. Mining pools are becoming a scale business. The small pool's fee revenue is too thin to support the engineering and regulatory cost structure that institutional ownership requires.
This explains why the top-three pool share keeps growing. Foundry USA has a compliance-first, institution-friendly brand. AntPool and F2Pool have deep ties to Asian hardware supply chains and financing. Their cost of capital is different, their geographic base is different, and their client relationships are different. They are not a cartel. A cartel requires coordination. These pools compete on fees, payout stability, and infrastructure quality. They share the same broad software stack, but they do not share a voting mechanism. The combined 60% is an arithmetic coincidence, not a governance agreement.
Keep an eye on the second tier as well. Luxor is rising on data services and hashprice derivatives. Braiins continues to sell open-source pool software to operators who want independence. The top-three chart looks static, but the service layer beneath it is changing. Pool ranking is not a permanent feature of Bitcoin; it is a snapshot of who offers the most attractive deal this quarter.
The Wrong Threat Model
The conventional interpretation of top-three concentration is that Bitcoin is becoming fragile. The contrarian interpretation is that the fragility is overstated because pool share measures choices, not control. Miners are independent actors who sell hashrate to the best service provider. A pool can only claim 'control' if the miners contributing to that pool willingly accept its template policy. That is not the control of a central bank. It is a negotiated service relationship.
I hold a specific bias. In 2017, I audited the 0x v1 smart contracts during the ICO boom. I found a re-entrancy vulnerability in the exchange proxy contract, submitted a fix, and watched the market treat that single bug as a failure of the entire protocol. It was not. An implementation bug in one contract is not a consensus failure. The same discipline applies to SBI. A vendor-level shutdown is not a protocol-level failure. It is a balance-sheet event.
The threat model that keeps me awake is not the pool count. It is geography. Foundry USA operates under American institutional and regulatory pressure. AntPool and F2Pool operate in Asia with different political constraints. If a regulatory storm targets one jurisdiction, the concentration metric could reshuffle faster than any governance forum can react. A pool can be sanctioned, pressured, or forced to alter its template policy. The miners of that pool will then move, and the 60% chart will be rewritten in a week.
We trade the code, not the culture. Pool names are not consensus rules. The code does not care whether the winning pool is Japanese, American, or Chinese. The ledger only records which proof-of-work was valid.
I watched the ape sell; the code still audits. The ape sells because the chart moves. The code audits because the structure is still sound.
The Verdict Is Still Open
The 60.01% reading is a dashboard warning, not a death sentence. The next four to eight weeks will reveal whether SBI's miners settled into the big three or diversified into smaller pools with better fees. If the top-three share stays above 60% for months, the conversation about pool governance becomes serious. If it falls below 55% as migration patterns settle, we will have evidence that the concentration was a snapshot, not a law.
The question for every market participant is not whether the top three pools are large. They are. The question is whether the miners who made them large can still walk away when the service no longer serves them. SBI's miners walked away. The network did not flinch. Exit liquidity is a courtesy, not a right. For the miners who moved, the courtesy was exercised without drama. The ledger will record where they went. The next difficulty adjustment will tell us what the market truly believes.