Bitcoin’s blockchain briefly split in two on September 11, 2026, when SpiderPool and AntPool each mined a valid block at height 966,500 within seconds of each other. The network settled the dispute the way it always does: the chain with more accumulated proof-of-work won, AntPool’s block became canonical, and SpiderPool’s version was discarded along with roughly 3.1389 BTC in block reward, worth about $240,000 at current prices. On its own, a one-block reorg is a footnote. This was the third one in four weeks, following similar events on August 16 and August 24, and it has put a number back in front of the industry that a lot of people would rather not look at too closely: two mining pools now produce nearly 45% of every Bitcoin block mined.
That number is the real story behind the reorg headlines. Bitcoin’s security model has always rested on the assumption that no single actor controls enough hashrate to rewrite history or censor transactions at will. Researchers who track pool concentration say that assumption is looking thinner than it has in years, and the September reorg cluster is the most visible symptom yet.
What Happened: The Sept. 11 SpiderPool-AntPool Reorg
According to Crypto Economy’s reporting, the event unfolded at block height 966,500 on September 11, 2026, when SpiderPool and AntPool each found a valid block pointing to the same parent block at nearly the same moment. Both blocks propagated across different parts of the network before either miner had a chance to confirm the other’s existence. Galaxy Research, the crypto markets research firm that first flagged the event, said its own node never saw the winning AntPool block until it had already been confirmed and extended on top of the chain, a detail that illustrates how fast these races resolve in practice.
Bitcoin’s consensus rule is simple in theory: the chain with the most cumulative proof-of-work wins, full stop. In this case, once a miner extended the AntPool block with a follow-up block, that side of the fork carried more total work, and every node on the network reorganized to it. SpiderPool’s block, valid in every other respect, became an orphan. The transactions inside it went back into the mempool for the next miner to pick up, and the 3.1389 BTC subsidy SpiderPool would have earned simply evaporated, according to Bitcoin.com News‘ account of the incident.
Reorgs like this are not new, and Bitcoin’s difficulty adjustment and ten-minute block target exist precisely so the network can absorb the occasional simultaneous discovery. What has changed is the frequency, and that frequency is a direct function of how few entities are producing most of Bitcoin’s blocks.
Inside a One-Block Reorg: How Bitcoin’s Consensus Resolves Competing Blocks
A one-block reorg happens when two miners solve a valid block at roughly the same height within seconds of each other, before either can propagate fully across the global peer-to-peer network. For a brief window, the network genuinely disagrees about which block is the tip of the chain. Nodes closer to SpiderPool’s mining infrastructure saw its block first; nodes closer to AntPool’s infrastructure saw that one first. Both are legitimate, fully valid blocks under Bitcoin’s rules. The fork only resolves once someone mines the next block on top of one of the two candidates, at which point that branch has strictly more accumulated work and every honest node switches to it.
This is proof-of-work functioning exactly as designed, not a bug or an attack. Satoshi Nakamoto’s original design anticipated that simultaneous block discovery would happen occasionally and built the heaviest-chain rule specifically to resolve it without a central arbiter. The economic cost falls entirely on the losing miner, who forfeits the block subsidy and transaction fees for work that turned out not to count. That is the mechanism working as intended.
What makes September’s cluster notable is not the mechanism, it’s the cadence. Three reorgs in a four-week span, each involving pools that individually or in combination command a meaningful share of total hashrate, is a higher rate than Bitcoin has seen in years of relatively quiet block production. Faster average hashrate growth and larger individual pools both increase the odds that two of them find a block within the same propagation window.
A Pattern, Not an Anomaly: Three Reorgs in Four Weeks
The September 11 event was not isolated. Bitcoin recorded a prior one-block reorg at block 962,722 on August 16, 2026, and another at block 963,853 on August 24, 2026, according to the same Galaxy Research monitoring that flagged the SpiderPool-AntPool clash. That puts three reorgs inside a roughly four-week window, a pace that analysts tracking the chain describe as unusual even accounting for Bitcoin’s growing hashrate.
None of the three has been characterized as an attack. Each looks like the byproduct of two large pools discovering blocks within moments of one another and losing the propagation race, exactly the scenario proof-of-work is built to survive. But security researchers who model Bitcoin’s threat surface point out that the same concentration that makes these races more frequent also lowers the bar for something more deliberate. A pool holding a large enough share of hashrate does not need a network-breaking bug to reorganize the chain; it just needs enough of an edge in block production to consistently win these short races, or in a more adversarial scenario, to intentionally withhold and release blocks to reorg out a rival’s work.
Who Controls Bitcoin’s Hashrate in September 2026
Bitcoin’s total network hashrate sat at roughly 886 to 934 EH/s across various mid-September 2026 snapshots, with mining difficulty at 127.45 trillion following a 1.31% upward adjustment at block 965,664 on September 6, 2026. That difficulty level is near an all-time high, meaning more computing power than ever is competing to produce Bitcoin’s roughly 900 blocks a week. The problem is how unevenly that power is distributed. Data from Hashrate Index, pulled from the trailing block window as of September 18, 2026, shows the following breakdown.
| Rank | Pool | Est. Hashrate | Network Share |
|---|---|---|---|
| 1 | Foundry USA | ~209 EH/s | 25.5% |
| 2 | AntPool | ~157 EH/s | 19.2% |
| 3 | F2Pool | ~119 EH/s | 14.5% |
| 4 | ViaBTC | ~87 EH/s | 10.6% |
| 5 | SpiderPool | ~66 EH/s | 8.1% |
| 6 | MARA Pool | ~46 EH/s | 5.6% |
| 7 | Binance Pool | ~15 EH/s | 1.8% |
| – | All others (Luxor, Braiins, SecPool, etc.) | ~186 EH/s | ~14.7% |
Add up just the top two names and they account for 44.6% of everything the network mines. The top four, Foundry USA, AntPool, F2Pool and ViaBTC, control nearly 70% between them. Research published by Spark puts Bitcoin’s Nakamoto coefficient, the minimum number of entities that would need to collude to control the majority of hashrate, at just three pools. That figure has been sliding for close to a decade. In May 2017, the two largest pools together held less than 30% of hashrate; by December 2023 that combined share had already crossed 55%.
The Hidden Centralization: AntPool’s “Pool of Pools” Problem
The headline pool-share numbers likely understate the real concentration, and that’s the part researchers have been raising alarms about for over a year. Bitcoin Core contributor 0xb10c has documented that several pools that appear independent in block explorers are actually relaying jobs built by AntPool’s parent company, Bitmain. In a detailed technical writeup, 0xb10c lists Poolin, CloverPool, Ultimus Pool, Binance Pool, SecPool, SigmaPool, Rawpool, Luxor and Mining Squared as pools that have relayed AntPool-constructed block templates rather than building their own.
When that proxy network is counted as a single economic actor rather than nine separate pools, the combined share climbs well past what the public leaderboards show. Spark’s research estimates that AntPool and its affiliated proxies have at various points controlled somewhere between 37% and 50% of total network hashrate, a level that, if concentrated in one operator’s hands with intent, would put a single entity within reach of the 50% threshold long associated with double-spend and censorship risk. One researcher quoted in that coverage put it bluntly: there is effectively one pool with hashrate approaching half the network, even though the public dashboards list nine.
To be clear, controlling block template construction is not the same as controlling mining hardware. The physical ASICs doing the hashing are still owned and operated by many separate businesses and individuals around the world. What the proxy-pool structure concentrates is a narrower but still important lever: the ability to decide which transactions go into a block, and in the case of a reorg race, the software and infrastructure that determines how quickly a block propagates.
Measuring the Risk: HHI, Empty Blocks and Censorship History
Economists have a standard tool for quantifying market concentration: the Herfindahl-Hirschman Index, calculated by squaring and summing each competitor’s market share. A market below 1,500 is considered unconcentrated; above 2,500 is considered highly concentrated. Spark’s analysis puts Bitcoin mining’s current HHI at roughly 1,492, sitting right on the boundary between moderate and high concentration, and trending upward as the top pools have grown.
Empty blocks as a centralization tell
One practical symptom of concentrated, fast-propagating infrastructure is the empty block rate, blocks mined with no transactions at all because a pool started hashing before it finished building a template. Industry estimates put the empty block rate at 0.15% to 0.5% of all blocks mined, a figure that has stayed roughly stable, though decentralized pools like OCEAN reportedly mine empty blocks at a rate several times higher than the network average because their miners build templates independently rather than relying on a centralized operator’s optimized pipeline.
A documented history of transaction filtering
Concentration risk is not purely theoretical. Marathon Digital ran an OFAC-compliant filtering policy on the blocks it mined starting in May 2021 before reversing course. F2Pool filtered six sanctioned transactions in November 2023 and an additional fifteen in January 2025, according to on-chain analysis cited in Spark’s research. Neither incident amounted to a chain-level censorship event, since excluded transactions were simply picked up by the next pool to mine a block, but both demonstrate that individual pools operating at scale can and do make unilateral decisions about which transactions they process.
Historical Context: From the 2010 Overflow Bug to the 2014 GHash.io Scare
Bitcoin has been here before, in different forms. The chain has experienced exactly two deep reorgs in its entire sixteen-year history, both caused by software bugs rather than hashrate concentration, and neither remotely resembles what happened on September 11.
| Event | Date | Depth | Cause |
|---|---|---|---|
| Value overflow incident | Aug. 15, 2010 | ~53 blocks | Integer overflow bug (CVE-2010-5139) created 184B+ BTC; emergency patch forced a chain rewrite |
| v0.7/v0.8 database fork | Mar. 11-12, 2013 | 24 blocks | BerkeleyDB vs. LevelDB incompatibility (BIP-50); miners coordinated a downgrade, one double-spend occurred |
| Minor consensus split | 2015 | ~6 blocks | Isolated consensus disagreement, resolved via normal heaviest-chain rules |
| GHash.io hashrate scare | Jun.-Jul. 2014 | No deep reorg | Single pool briefly exceeded 50% of network hashrate, sparking public pressure to self-limit |
| SpiderPool/AntPool reorg cluster | Aug. 16 – Sept. 11, 2026 | 1 block (x3) | Simultaneous block discovery between large pools amid rising concentration |
The 2014 episode is the closer historical parallel to today’s concerns. GHash.io, a pool that no longer exists in its original form, briefly crossed the 50% hashrate threshold that year, prompting enough public alarm that the pool publicly committed to voluntarily capping its own share. No attack was ever attempted or proven. What that episode demonstrated is that Bitcoin’s security model has always depended partly on informal, social pressure on dominant pools rather than a hard protocol-level cap, because no such cap exists or could easily be enforced. Twelve years later, the industry is relying on essentially the same informal backstop, just with the concentration now split, on paper, across nine or ten separate-looking pool brands instead of one.
Why Concentration Keeps Rising: Post-Halving Mining Economics
The economic backdrop matters here. Bitcoin’s April 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC, permanently compressing the revenue every miner earns per block. That squeeze rewards scale. Large mining operations can negotiate better electricity rates, run newer and more efficient ASICs, and spread fixed infrastructure costs over more hashrate than smaller, independent miners can. The practical result is that solo and small-pool mining has become progressively less viable since the halving, pushing more hashrate toward the handful of pools that can operate at the lowest marginal cost.
Hashprice, the revenue a miner earns per unit of hashrate, has been volatile through 2026. One mid-September tracking report put daily hashprice at roughly $37.96 per petahash, while an earlier September update described a roughly 22% jump in hashprice alongside the difficulty increase, evidence that miner economics remain sensitive to both Bitcoin’s price and network-wide competition for the same fixed block reward. That volatility itself favors well-capitalized operators who can weather thinner margins, reinforcing the consolidation trend rather than reversing it.
Industry Response: Stratum V2 and the Push for Miner-Chosen Blocks
The mining industry’s main answer to pool-level centralization is a protocol upgrade called Stratum V2, and it has picked up real momentum in 2026. Bitcoin Core v30 shipped experimental Stratum V2 support in October 2025. In May 2026, seven major pools, Foundry, AntPool, F2Pool, SpiderPool, MARA Pool, Block Inc. and DMND, representing a combined 75% of global hashrate, joined a formal Stratum V2 Working Group, according to CoinDesk’s reporting on the announcement.
The feature that matters most for centralization is called Job Declaration, a Stratum V2 component that lets individual miners, not the pool operator, choose which transactions go into the block they are hashing. That would directly undercut the pool-of-pools template problem, since a proxy relaying someone else’s block template becomes far less meaningful if the miner at the hardware level can override that template’s contents. GoMining mined the first Stratum V2 block using Job Declaration in production in May 2026, a genuine milestone.
The gap between announcement and adoption is still wide, though. As of mid-2026, only Braiins Pool and DMND actually run Job Declaration in production, representing a combined 3% to 5% of network hashrate. Signing a working-group pledge and shipping the software that changes how trillions of dollars in mining infrastructure operates are two very different timelines, and the pools with the most to gain from the status quo, the largest ones, have the least commercial incentive to move quickly.
Decentralized Alternatives: OCEAN, DATUM and P2Pool
A smaller set of pools has tried to route around the centralization problem entirely rather than wait for the big players to adopt Job Declaration. OCEAN, founded in November 2023 with $6.2 million in seed funding, built its own protocol called DATUM that gives individual miners direct control over block template construction from day one, without depending on Stratum V2 adoption elsewhere. Stablecoin issuer Tether announced in April 2025 that it would direct a portion of its own hashrate to OCEAN specifically to support that decentralization goal.
Despite the attention, OCEAN’s market share sits at roughly 2% of total network hashrate, a reminder of how difficult it is for a philosophically-motivated alternative to compete with pools offering lower variance payouts and tighter integration with large-scale mining operations. Older decentralized options like P2Pool, which removes the pool operator entirely by having miners share a peer-to-peer chain of shares, occupy an even smaller sliver of the network. The economics of mining reward scale and low-variance payout structures, and both currently favor consolidation over decentralization, regardless of which option is more resilient in theory.
Bitcoin Mining vs. Ethereum Staking: Comparing Centralization Risk
Bitcoin is not the only major chain wrestling with the gap between decentralized in theory and concentrated in practice. Ethereum moved from proof-of-work to proof-of-stake in 2022, trading mining pools for staking providers, and the concentration conversation there looks structurally similar even though the mechanics differ.
| Metric | Bitcoin (mining) | Ethereum (staking) |
|---|---|---|
| Largest single entity | Foundry USA, ~25.5% | Lido, ~23% of staked ETH |
| Top-2 concentration | ~44.6% (Foundry + AntPool) | Lido + Binance combined, well above 25% |
| Peak historical concentration | GHash.io briefly >50% (2014) | Lido peaked near 32% of staked ETH (late 2023) |
| Primary decentralization fix | Stratum V2 / Job Declaration | Distributed validator technology, operator diversification |
Lido’s share of all staked ETH has fallen from a peak near 32% in late 2023 to roughly 23% in 2026, a genuine decline driven by a widening field of competitors and compressed liquid-staking yields, according to staking trackers that follow the sector. Coinbase holds close to 2.9 million ETH in its staking program, and Binance separately manages roughly 3.7 million ETH in pooled staking, meaning the three largest providers combined still command a share of Ethereum’s validator set that would raise the same governance and censorship questions Bitcoin’s mining pools face if it were concentrated under one roof. The core difference is that Ethereum’s concentration lives in custody of staked capital, while Bitcoin’s lives in control of computing hardware and, increasingly, the software that decides what that hardware works on.
Market Impact: Price, Hashprice and Miner Economics
Bitcoin’s price has not moved sharply on the reorg news itself, consistent with how markets have generally treated one-block reorgs as routine rather than alarming. BTC traded at roughly $76,431 on September 17, 2026, down about 2.7% for the month after opening September near $78,550. That’s a market shrugging off a mining-infrastructure story rather than pricing in a security failure, and it reflects the fact that none of the three September reorgs involved a double-spend, a stolen transaction, or any loss to an end user. The only party that lost money was SpiderPool itself, forfeiting one block’s worth of subsidy and fees.
Where the story does matter economically is for mining-sector investors and the exchanges that rely on confirmation counts to guarantee finality. Publicly traded miners like MARA Holdings, whose MARA Pool controls roughly 5.6% of network hashrate, and Foundry’s parent Digital Currency Group both have a direct commercial stake in how the centralization narrative develops, since regulatory or exchange-level responses that push large pools to decentralize could raise their operating costs. Hashprice remaining volatile, tracked at close to $38 per petahash per day in mid-September, keeps pressure on smaller miners to either sell out to larger pools or exit entirely, a dynamic that, left unaddressed, points toward more concentration rather than less.
What This Means for Exchanges and Confirmation Policies
Exchanges and custodians set confirmation thresholds, the number of blocks that must be mined on top of a deposit before crediting a customer’s account, specifically to protect against reorgs. A one-block reorg like September 11’s is well within the margin that a standard six-confirmation policy is designed to absorb; no legitimate deposit or withdrawal was at risk during the event. But a cluster of three reorgs in a month is the kind of data point that risk teams at major exchanges tend to review, since it can inform whether existing confirmation thresholds remain adequate as pool concentration rises.
The more consequential question is what happens if concentration keeps climbing toward the range where a single entity, or a small coordinated group, could realistically sustain a deeper reorg rather than lose a single-block race. Bitcoin has never experienced a reorg attack of that kind driven by pool concentration rather than a software bug, and the 44.6% combined share held by Foundry USA and AntPool today, while high by the network’s own historical standards, remains short of the 50% threshold most security models treat as the meaningful line.
Predictions: Where Bitcoin Mining Centralization Goes From Here
Based on the trajectory of hashrate distribution, protocol development, and mining economics through 2026, a few outcomes look likely over the next twelve to eighteen months.
- Short, one-block reorgs will likely become more frequent, not less, as total network hashrate keeps climbing and a shrinking number of large pools compete for the same blocks, purely as a function of propagation-window math rather than any change in miner behavior.
- Stratum V2 Job Declaration adoption will expand beyond Braiins Pool and DMND, but slowly. Expect one or two mid-sized pools to move to production status within a year, while the largest pools continue treating working-group membership as sufficient for now.
- Scrutiny of AntPool’s proxy-pool network will intensify. Expect further independent research attempting to quantify its true combined share, and possibly renewed public pressure similar to the informal backstop that reined in GHash.io in 2014.
- Exchanges handling high-value Bitcoin transactions will likely revisit confirmation-count policies for large deposits, even without evidence of an actual attack, simply because the reorg frequency data now exists and risk teams tend to act on documented patterns.
- OCEAN and other decentralized-mining alternatives will keep growing in absolute terms but stay a small percentage of total hashrate, since the economics of mining continue to favor pools that can offer low-variance payouts and tight infrastructure integration over philosophically decentralized alternatives.
Frequently Asked Questions
What is a Bitcoin blockchain reorg?
A reorg, short for reorganization, happens when the Bitcoin network switches from one version of the blockchain to a competing version that has more accumulated proof-of-work. It typically occurs when two miners find valid blocks at nearly the same height within seconds of each other, and the network resolves the tie once one branch pulls ahead in total work.
Was the September 11, 2026 reorg an attack?
No. Every account of the event, including Galaxy Research’s on-chain monitoring, describes it as a normal consequence of two large pools discovering blocks nearly simultaneously and losing the propagation race. There is no evidence of a double-spend, stolen funds, or intentional block withholding.
How much hashrate do Foundry USA and AntPool control combined?
As of Hashrate Index data from September 18, 2026, Foundry USA controls roughly 25.5% of network hashrate and AntPool controls roughly 19.2%, for a combined share of about 44.6%.
What is AntPool’s “pool of pools” problem?
Bitcoin Core contributor 0xb10c has documented that several pools listed separately on public dashboards, including Poolin, Binance Pool, SecPool and others, relay block templates constructed by AntPool’s parent company, Bitmain, rather than building their own. When counted together, this proxy network’s combined share has been estimated at 37% to 50% of total hashrate at various points.
What is Stratum V2 and how does it address centralization?
Stratum V2 is a mining protocol upgrade that includes a feature called Job Declaration, which lets individual miners choose which transactions go into the block they are hashing instead of relying entirely on a pool operator’s template. Bitcoin Core added experimental support in October 2025, and seven major pools representing about 75% of hashrate joined a working group around it in May 2026, though only a small fraction of hashrate runs it in production so far.
How does Bitcoin’s mining concentration compare to Ethereum’s staking concentration?
Both show similar patterns of a dominant provider trending down from a higher peak. Bitcoin’s Foundry USA holds about 25.5% of hashrate today. Ethereum’s largest liquid-staking provider, Lido, holds roughly 23% of staked ETH, down from a peak near 32% in late 2023. Neither network has a single entity above the 33% to 50% thresholds most security models flag as concerning, but both remain more concentrated than their respective communities would prefer.
Should Bitcoin holders be worried about a 51% attack?
Not based on current data. No single publicly identified pool controls anywhere near 50% of network hashrate, and Bitcoin’s Nakamoto coefficient of roughly three pools required for a majority coalition, while lower than in past years, still requires coordinated collusion between multiple large, competing businesses rather than the actions of one operator.
How many confirmations are needed to consider a Bitcoin transaction final?
Most exchanges and custodians require six confirmations for large Bitcoin deposits, a threshold designed to make a reorg deep enough to reverse the transaction statistically implausible without majority hashrate control. September’s one-block reorgs fell well within that safety margin.
