Bitcoin experienced three one-block chain reorganizations in four weeks between August 16 and September 11, 2026 — at block heights 962,722, 963,853, and 966,500. Each event involved competing valid blocks discovered near-simultaneously by major mining pools. None resulted in double-spends or con...
Bitcoin experienced three one-block chain reorganizations in four weeks between August 16 and September 11, 2026 — at block heights 962,722, 963,853, and 966,500. Each event involved competing valid blocks discovered near-simultaneously by major mining pools. None resulted in double-spends or consensus failures. Individually, single-block reorgs are routine: they occur roughly once every 45 days under normal conditions and resolve harmlessly as the network converges on the chain with the most accumulated proof-of-work.
Three in four weeks is statistically uncommon. The clustering coincides with a period of acute mining pool concentration: four pools — Foundry USA (~31%), AntPool (~18%), ViaBTC (~13%), and F2Pool (~10%) — control over 70% of Bitcoin's ~940 EH/s network hashrate. The Nakamoto coefficient, the minimum number of pools required to exceed 50% of block production, stands at 3. Research by developer 0xb10c and Bitcoin Core contributor Matt Corallo has documented that AntPool operates as a "pool of pools," with at least five nominally independent pools (BTC.com Pool, Binance Pool, Poolin, EMCD, and Rawpool) sharing identical block templates and transaction prioritization. This hidden centralization means the effective concentration may be higher than headline pool-share numbers suggest.
The industry response is structural: seven pools representing 75% of hashrate joined the Stratum V2 Working Group in May 2026, and OCEAN's DATUM protocol now carries institutional backing after Tether began routing hashrate through it in April 2026. Both protocols shift block-template construction from pool operators to individual miners. Whether adoption reaches critical mass before the next difficulty epoch remains an open question.
| Date | Block Height | Competing Pools | Winning Block | Transactions | |------|-------------|----------------|--------------|-------------| | Aug. 16, 2026 | 962,722 | Not disclosed | — | — | | Aug. 24, 2026 | 963,853 | Not disclosed | — | — | | Sept. 11, 2026 | 966,500 | SpiderPool vs. AntPool | AntPool | 4,359 |
The September 11 event is the best-documented. Galaxy Research detected the anomaly via its private node. At block height 966,500, both SpiderPool and AntPool produced valid blocks referencing the same parent block. AntPool's block, timestamped 11:30:30 UTC and containing 4,359 transactions, accumulated more work when AntPool subsequently mined block 966,501 on top of its own version. SpiderPool's competing block was discarded. No transactions were lost; no double-spends occurred.
An earlier two-block reorganization had occurred in March 2026 at approximately block height 941,880, involving Foundry USA and AntPool/ViaBTC. It also resolved naturally within minutes.
Under standard network conditions, one-block reorgs occur roughly once every 45 days, according to historical chain data. Three within a 26-day window, while not unprecedented, sits outside the normal distribution and warrants examination of underlying conditions.
Bitcoin's proof-of-work consensus operates on the longest-chain rule. When two miners discover valid blocks at near-identical times, different parts of the network temporarily adopt different versions. This creates a brief fork. The conflict resolves when the next block is mined on top of one branch, giving it more cumulative work. The other branch is abandoned, and its transactions return to the mempool for inclusion in a subsequent block.
This is not a bug. It is a designed property of Nakamoto consensus. The mechanism is well-understood and has functioned since Bitcoin's genesis block. The relevant question is not whether reorgs happen, but whether their frequency or depth is changing in ways that correlate with structural shifts in mining.
Deeper reorgs — three or more blocks — are extremely rare on Bitcoin's mainnet and typically indicate either a software bug or a deliberate 51% attack. All recent events have been single-block reorgs, the most benign category.
As of a June 23, 2026 snapshot reported by D-Central, Bitcoin mining pool market share stood at:
| Pool | Hashrate Share | |------|---------------| | Foundry USA | ~31% | | AntPool | ~18% | | ViaBTC | ~13% | | F2Pool | ~10% | | SpiderPool | ~7% | | Others | ~21% |
Four pools control over 70% of the network's approximately 940 EH/s. The Herfindahl-Hirschman Index (HHI), a standard measure of market concentration used by the U.S. Department of Justice, stands near 1,492. The DOJ classifies markets below 1,500 as "unconcentrated," 1,500–2,500 as "moderately concentrated," and above 2,500 as "highly concentrated." Bitcoin mining sits at the boundary between unconcentrated and moderately concentrated — but this does not account for the proxy-pool phenomenon discussed below.
The Nakamoto coefficient — the minimum number of entities required to exceed 50% of block production — is 3. In practical terms, three pools colluding could theoretically censor transactions or attempt selfish-mining strategies.
At Foundry USA's ~31% share, classical selfish mining becomes "unambiguously profitable," according to analysis cited by Spark Money's research division. Selfish mining is a strategy where a pool withholds discovered blocks to gain a probabilistic advantage in future block races. At 25%+ hashrate, the expected return from selfish mining exceeds honest mining under certain propagation-delay assumptions.
No evidence of selfish mining has been detected. The concern is structural, not evidential.
Research by developer 0xb10c and Matt Corallo, discussed on the TFTC podcast and in public GitHub repositories, documented that multiple nominally independent mining pools share identical block templates with AntPool. The confirmed proxy pools are:
According to this research, these pools entered into FPPS (Full Pay Per Share) insurance arrangements with Bitmain, AntPool's parent company and the world's largest ASIC manufacturer. In exchange for Bitmain insuring against variance in pool luck, the pools agreed to use Bitmain's block templates and transaction prioritization, and to route mining rewards through Bitmain's custodial infrastructure.
If AntPool's ~18% headline share is combined with its proxy pools, the effective concentration under Bitmain's template control is substantially higher. The precise aggregate figure is difficult to confirm because pool-share reporting does not distinguish between independent operators and template-sharing proxies. This architectural opacity is itself a risk factor.
The implication: headline pool-share statistics may understate the degree of centralization in Bitcoin's block-construction pipeline. The entity that controls the block template controls which transactions are included, which are excluded, and in what order — a censorship surface that pool-share charts do not capture.
The reorg cluster occurs against a backdrop of squeezed mining economics:
Compressed margins incentivize miners to join larger pools for more predictable payouts, which mechanically increases concentration. This is the economic flywheel driving the Nakamoto coefficient lower: the worse mining economics get, the more miners aggregate into fewer pools, and the more concentrated block production becomes.
Two competing protocols aim to decouple block-template construction from pool operators:
Stratum V2 is an open protocol that allows individual miners to construct their own block templates while still participating in a pool. On May 7, 2026, seven pools representing approximately 75% of global hashrate — AntPool, Block, DMND, F2Pool, Foundry, MARA Foundation, and SpiderPool — joined the Stratum V2 Working Group, according to CoinDesk. Braiins Pool is currently the only pool with full production support including Job Negotiation. The SRI (Stratum Reference Implementation) working group projects that by end of 2026, Stratum V2 will be the default protocol for new ASIC firmware shipments, potentially reaching 40%–60% of network hashrate.
DATUM (Decentralized Alternative Templates for Universal Mining) is OCEAN's open-source protocol, which takes a more aggressive approach to decentralization. Under DATUM, the pool is "mathematically blind" to template contents — it only sees merkle branches, making censorship architecturally impossible rather than policy-dependent. OCEAN mined its first DATUM block on September 30, 2024. In April 2026, Tether announced it would route hashrate to OCEAN using DATUM, lending institutional weight to the protocol.
The key distinction: Stratum V2 allows miners to construct templates; DATUM requires it. Both address the same attack surface — centralized template control — through different trust assumptions.
Neither protocol has reached critical mass. The gap between joining a working group and deploying in production is substantial. Until one or both protocols are running on a majority of hashrate, the block-template construction pipeline remains concentrated among a small number of pool operators.
What the reorgs indicate: The three-in-four-weeks clustering is consistent with increased competition among a small number of large pools operating at similar hashrate levels. When two pools each hold 15–30% of hashrate, the probability of near-simultaneous block discovery is mechanically higher than in a more distributed hashrate landscape. The reorgs are a symptom, not a threat.
What the concentration indicates: A Nakamoto coefficient of 3, combined with evidence of hidden proxy-pool relationships, represents a censorship surface. No censorship has been documented on Bitcoin's mainnet. However, the structural capability exists: three cooperating entities could theoretically enforce transaction-selection policies that exclude specific addresses or transaction types. This is particularly relevant as regulatory pressure on mining operations increases globally.
What is not at risk: Bitcoin's consensus rules remain intact. No double-spends occurred during any of the reorg events. The 21-million supply cap, block interval, and cryptographic security are unaffected by pool concentration. The risk surface is censorship and transaction ordering, not monetary policy.
Three reorgs in four weeks are not a crisis. They are a data point. Single-block reorganizations are a designed feature of Nakamoto consensus, not a failure mode. The events caused no loss of funds and resolved through the protocol's normal longest-chain convergence.
The more consequential signal is what the reorgs illuminate about the underlying structure: a network that processes $78,000-per-unit transactions secured by a block-production pipeline where three entities can theoretically exercise majority control. The economic forces driving this concentration — rising difficulty, compressed hashprice, post-halving margin pressure — are structural and unlikely to reverse absent a sustained Bitcoin price increase.
The Stratum V2 and DATUM protocols represent the most credible path to mitigating the centralization of block-template control. Working-group commitments from 75% of hashrate are a necessary but insufficient condition. The metric that matters is production deployment — how many exahashes are actually constructing their own templates today, not how many have agreed to consider doing so.
Until that number is materially higher than it is today, Bitcoin's censorship resistance rests on the good behavior of a small number of pool operators, not on the protocol's architectural guarantees. The reorgs are harmless. The concentration they expose is not.