The 60% That Was Already There: SBI Crypto and the Architecture of Mining Concentration

In the span of a single month, a Japanese mining pool dissolved into the arithmetic of Bitcoin's history. SBI Crypto's seven-day average hash rate fell from 16.222 EH/s to 5.817 EH/s, then collapsed to 0.452 EH/s — losing roughly ninety-seven percent of its computational presence before its Stratum servers went dark for good. The exit was orderly, almost polite: a phased disconnection of miners, a graceful withdrawal from the block race. Bitcoin's price did not flinch. The market barely registered the event. And yet the week before SBI stopped producing blocks, the three largest pools already controlled more than sixty percent of attributed block share. The threshold had been crossed before anyone noticed. SBI's exit was not the cause of the concentration; it was the confirmation of it. SBI Crypto was never a typical mining operation. It was the digital asset arm of SBI Group, one of Japan's largest financial conglomerates, a name that carried institutional weight and regulatory access. Its presence in mining was a statement about legitimacy — a traditional finance giant willing to point machines at Bitcoin's proof-of-work. Its exit is a statement of a different kind. The timing matters. We are fifteen months past the fourth halving, when the block subsidy fell from 6.25 to 3.125 BTC. Miner revenue per unit of hash power was cut in half overnight, and the industry has been adjusting ever since. Mining pools take one to four percent of a miner's output as a service fee, so their revenue contracts with the subsidy: when blocks deliver less, the cut delivers less. SBI's steady decline — from 16.222 EH/s at the end of June to roughly 5.8 EH/s by mid-July, then to 0.452 EH/s on July 31 — follows the arc of a business whose unit economics no longer justified its existence. Add Japan's persistently high electricity costs to a post-halving margin squeeze, and the decision begins to look less like strategy and more like gravity. But here is the data point that reframes the story: on July 20, the three largest pools — Foundry USA, AntPool, and F2Pool — held a combined 64.8 percent of attributed block share. On July 27, they still held 60.8 percent. SBI, meanwhile, had fallen to 0.72 percent. The concentration existed before the exit, during the exit, and after the exit. SBI's departure removed roughly 0.07 percent of network hash power from the pool's own attribution — a rounding error in security terms. What it removed from the narrative, however, was our excuse for pretending the concentration was someone else's doing. Based on my own work auditing protocol consensus mechanisms during the last bear market, I have learned to distrust the clean line between the network layer and the service layer. The Bitcoin core protocol — proof-of-work consensus, difficulty adjustment, the UTXO model — changed not at all during SBI's shutdown. The network's difficulty did not lurch; block times did not stretch; settlement continued its indifferent rhythm. What changed was entirely within the Stratum service layer, the protocol that connects miners to pools. And that, I would argue, is both the problem and the opportunity. Let me be precise about the measurement issue, because it is where most analysis goes astray. Pool share statistics are calculated from attributed blocks — when a pool's coinbase template finds a block, the block is credited to the pool. This is not a measurement of computation; it is a label applied after the fact. The 60 percent readings that circulate in industry dashboards capture a single instant, not a persistent state of control. A miner can switch pools by editing one connection parameter in a config file; it takes seconds. The concentration we observe is not a structural cage; it is a photographic frame. Miners stay in the frame, but they can leave it at any moment. That is the insight the headline numbers obscure. SBI's remaining hash power almost certainly did not vanish into the void. It migrated. The pool's official telemetry reflected only the computation still attributed to SBI's service; the machines themselves were likely reconnecting to other pools in the days before the shutdown, redirecting their work to Foundry, AntPool, F2Pool, or one of the smaller operations. The aggregated data cannot tell us where the flow went, and that blind spot is precisely where the real story lives. We are measuring labels, not loyalty. The economic analysis points in the same direction. SBI's 16.222 EH/s represented roughly 2.5 percent of network hash rate at its recent peak; by the end, its share was negligible. The token-level impact is minimal: Bitcoin's supply schedule, its hard cap of twenty-one million, and its halving rhythm are entirely indifferent to which pool finds a block. What shifts is the balance of fee revenue among pool operators. Foundry held 26.67 percent of attributed blocks, AntPool 17.13 percent, F2Pool 16.21 percent. Every displaced hash from SBI's collapse flows into someone else's fee stream. Yet the market for that stream is growing more competitive, not less. The mid-tier rankings show a quiet reshuffling: Luxor has risen on data services and hash rate derivatives; Braiins, the open-source pioneer, has declined; NeoPool has vanished from the rankings. This is the texture of an industry in consolidation, but it is also evidence that the doors have not fully closed. A pool retains miners through service quality, fee structure, and the policies embedded in its block templates — whether it runs Bitcoin Core's default templates, whether it accommodates Ordinals and BRC-20 traffic, whether it pays out promptly. Miners are not prisoners of their aggregation; they are customers of it. Here is the contrarian truth, and it will satisfy neither the panic camp nor the apologists. SBI's exit did not make Bitcoin meaningfully less secure. Removing under three percent of network hash power does not move the needle on a 51 percent attack cost measured in hundreds of millions of dollars. The doomsday readings — "three pools control 60 percent, the network is at risk" — mistake a service-layer statistic for a consensus-layer vulnerability. At the same time, the deflationary comfort — "the protocol is immutable, nothing has changed" — mistakes protocol stability for health. Both framings share the same flaw: they look at the network layer and ignore the layer where human decisions are actually made. The uncomfortable location of power is the block template. A pool's operator chooses which transactions enter the candidate block, which policies apply, which version of the node software runs beneath the service. This is not theoretical; it is the everyday texture of mining coordination. When three pools control sixty percent of attributed blocks, the question is not whether they could collude to rewrite history — such collusion would be economically self-destructive and technically detectable — but whether convergent policies create a de facto standard smaller operators must adopt to survive. Standardization, not conspiracy, is how oligopolies begin. If Foundry's template choices become the industry default because the math of the market demands it, the diversity of the ecosystem erodes one quiet decision at a time. No attack is required for that erosion. Only adherence. SBI's exit is a reminder that this layer has real-world consequences. The pool was not driven out by malice; it was driven out by margins. In a bear market, survival is the only metric that matters, and survival creates its own architecture: hash power seeks the most reliable aggregator, the lowest fee, the deepest balance sheet. The result is an industry that centralizes precisely because each actor makes the rational choice not to be the exception. We chart the code, but the market charts the behavior. What remains is not a question about SBI, which has already answered it. What remains is a question about measurement. We track attributed blocks, but not the actual flow of computation across the network's changing geography. We quote percentages that freeze a single moment, then treat them as permanent structures. The tools we use to see Bitcoin's ecology are older than the ecology itself, and they miss the migration that happens in seconds. Watch whether the top-three share hardens above sixty percent across consecutive weekly buckets, or fragments back toward fifty under fee competition. Watch whether mid-tier pools like Luxor convert their data-service offerings into hash rate share. And watch the block templates, not just the block finders. The network remembers what the market forgets, but only if someone is looking at the right ledger. We chart the code, but the soul chooses the path — and in mining, as in everything else, the path is chosen by whoever is willing to look at the thing everyone else has agreed not to see.