Bitcoin Mining Pool Centralization: How Five Pools Control Block Production

Bitcoin mining pool centralization has reached a point where just four pools produce roughly 73% of all blocks. The top six pools mine over 95%. This concentration creates systemic risks that extend far beyond mining itself: every Bitcoin layer, including Layer 2 protocols, depends on a decentralized base layer that resists censorship and manipulation. When a handful of operators control which transactions enter blocks, the censorship resistance that makes Bitcoin valuable is no longer a guarantee.

Current Hashrate Distribution

Bitcoin's total network hashrate fluctuates between 850 and 930 EH/s as of mid-2026, with mining difficulty at approximately 124.9 trillion. That enormous computational power is concentrated in a remarkably small number of mining pools.

The Nakamoto Coefficient for Bitcoin mining (the minimum number of entities needed to reach 51% of hashrate) currently stands at just three pools. By any standard concentration metric, this is concerning for a system whose security model assumes no single party or small coalition controls block production.

Measuring Concentration: The Herfindahl-Hirschman Index

The Herfindahl-Hirschman Index (HHI) is a standard measure of market concentration used by regulators including the U.S. Department of Justice. It is calculated by summing the squares of each participant's market share percentage. A perfectly competitive market with 100 equal participants would score 100. A monopoly scores 10,000.

As of mid-2026, Bitcoin mining's reported HHI sits at approximately 1,492 according to D-Central's H1 2026 mining report: right at the boundary between moderate and high concentration. But this number understates the problem significantly, because it treats each pool's reported hashrate as independent.

The Hidden Centralization: Bitmain's Proxy Pool Network

The most alarming finding from recent research is that the pool distribution table overstates decentralization. Research by developer 0xb10c and Bitcoin Core contributor Matt Corallo revealed that AntPool (operated by Bitmain) functions as a "pool of pools." Multiple ostensibly independent mining pools share identical block templates and transaction prioritization with AntPool, indicating they are proxies rather than independent operators.

Confirmed proxy pools include BTC.com Pool, Binance Pool, Poolin, EMCD, and Rawpool. Additional pools have shown suspicious template coordination patterns. When these proxy operations are combined, AntPool's effective control over block template construction has been estimated at 37–50% of total network hashrate: far above its stated 18–19%. As Corallo noted in his analysis: "There's one pool that has near 50% of hash power."

Why proxy pools matter: Even though individual miners can theoretically switch pools, the critical function at stake is block template construction: which transactions are selected, ordered, and potentially excluded. If multiple pools use the same template provider, the real number of independent transaction selectors is much smaller than the number of pool brands suggests.

Bitmain's influence extends further. Its Antminer ASIC hardware ships with AntPool pre-configured as the default pool. This creates a pipeline where new mining equipment funnels hashrate to AntPool unless the operator actively changes the configuration. In July 2025, AntPool mined seven consecutive blocks, an event that, while statistically possible, underscored the scale of its hashrate concentration.

Why This Happened: Post-Halving Economics

The April 2024 halving reduced the block reward from 6.25 to 3.125 BTC, compressing miner margins and accelerating pool consolidation. Smaller pools could no longer offer competitive variance reduction or fee structures. Miners gravitated toward larger pools offering Full Pay-Per-Share (FPPS) payout models with 0% fees, a business model only viable at massive scale.

The trend has been directional. In May 2017, the two largest pools combined held less than 30% of hashrate. By December 2023, the top two (Foundry and AntPool) exceeded 55%. Post-halving consolidation pushed mining economics toward a structure where scale advantages compound: larger pools attract more hashrate, which enables lower fees, which attracts still more hashrate.

Concrete Risks of Pool Concentration

Transaction Censorship

Transaction censorship by mining pools is not a theoretical concern. It has happened multiple times.

In May 2021, Marathon Digital launched an "OFAC-compliant" mining pool that filtered transactions involving addresses on the U.S. Treasury Department's Specially Designated Nationals (SDN) list. This was a voluntary decision, not required by any regulation. Marathon reversed the policy after intense community backlash.

F2Pool has repeatedly implemented OFAC-based transaction filtering. In November 2023, developer 0xb10c detected six transactions from OFAC-sanctioned addresses excluded from F2Pool blocks. F2Pool co-founder Chun Wang acknowledged the filter and said it would be removed. In January 2025, research showed F2Pool was censoring again, with 15 transactions from OFAC-listed addresses excluded. This pattern of filter, remove, reinstall demonstrates that voluntary censorship resistance commitments are unreliable.

The compounding risk: A single pool censoring transactions causes roughly a one-block (10-minute) delay: the next non-censoring pool includes them. The danger escalates if a majority of hashrate refuses to build on blocks containing sanctioned transactions. In that scenario, compliant pools would orphan non-compliant blocks, making censored transactions permanently unconfirmable. With three pools controlling 51% of hashrate, this is operationally achievable.

Empty Block Mining

Empty blocks (blocks containing only the coinbase transaction) waste block space and slow transaction confirmation. Currently, 0.15–0.5% of blocks are mined empty: a dramatic improvement from Bitcoin's early years, but still meaningful when network congestion is high.

Empty blocks occur because pools begin hashing on an empty template immediately after receiving a new block header, before the full block has propagated and they can construct a template referencing it. The incentive is rational: seconds of hashing on an empty template beats waiting for full block propagation when the block subsidy and transaction fees are at stake.

Pool-level differences are notable. Foundry USA and MARA Pool reported zero empty blocks over a 24-month period. OCEAN, by contrast, mines empty blocks at 17.5x the network average, likely a consequence of its DATUM protocol requiring individual miners to construct templates using their own full nodes, which adds latency. SpiderPool shows a similarly elevated rate at 13.4x the average.

Selfish Mining Feasibility

Selfish mining is a strategy where a miner withholds discovered blocks to gain a disproportionate share of rewards. The original 2014 research by Eyal and Sirer established a 33% hashrate threshold for profitability without propagation advantage, dropping to 25% with moderate network advantages.

Subsequent research has significantly lowered these thresholds. A May 2026 paper published on IACR ePrint ("The Joint Channel Threshold") demonstrated that a sophisticated adversary combining network-layer eclipse capabilities with transaction fee exploitation could profitably selfish-mine with less than 1% of hashrate: a 15x reduction from single-channel analyses. While this requires state-level resources, it makes any large pool a plausible selfish mining threat.

At Foundry's 30% share, classical selfish mining is unambiguously profitable. The question is not whether it's possible but whether reputational and legal consequences deter it.

51% Attack Surface

A 51% attack allows an attacker to rewrite recent transaction history through chain reorganizations. With AntPool's proxy network potentially controlling 37–50% of hashrate, the gap to 51% is smaller than it appears. Even without reaching 51%, significant hashrate concentration enables probabilistic attacks: a pool with 30% hashrate can execute short reorganizations with meaningful success probability.

Hashrate Ownership vs. Pool Operation

A common counterargument to centralization concerns is that miners can switch pools at any time. This is technically true: miners own the ASIC hardware and merely point it at a pool's Stratum endpoint. Redirecting hashrate takes minutes.

In practice, switching is less fluid than it appears. Miners consider payout models (FPPS, PPS+, PPLNS), fee structures, payout thresholds, geographic latency, and track record. Foundry's 0% FPPS model is difficult for competitors to match. Bitmain's pre-configured defaults create friction against switching away from AntPool. Most importantly, individual miners have no visibility into whether their pool is censoring transactions or sharing templates with other pools.

The distinction matters most for understanding the threat model. Hashrate ownership is decentralized: thousands of independent mining operations exist worldwide. But block template construction is centralized: roughly six operators (and possibly fewer, accounting for proxy pools) decide what goes into 95% of blocks.

Block Template Construction: The Real Bottleneck

Under Stratum V1 (the protocol used by the vast majority of mining operations), the pool operator constructs the block template. This means the pool decides which transactions from the mempool are included, their ordering, and which are excluded. Individual miners contributing hashrate receive only a block header to hash. They have no influence over transaction selection.

This architecture means that even though tens of thousands of ASIC machines secure the network, transaction inclusion policy is set by a handful of pool operators. A pool operator could censor specific transactions, extract MEV through transaction ordering, or comply with government directives to exclude addresses: all without the knowledge of the miners contributing hashrate.

Decentralization Efforts

Stratum V2: Miner-Driven Template Construction

Stratum V2 is the most significant protocol-level response to mining centralization. Originally developed by Braiins (formerly Slush Pool, the first Bitcoin mining pool), V2 introduces a Job Declaration feature that allows individual miners to construct their own block templates. The pool coordinates reward distribution, but transaction selection moves from pool operator to miner.

In October 2025, Bitcoin Core v30 shipped with experimental Stratum V2 support. In November 2025, DEMAND (DMND) launched as the first Stratum V2-native pool. In 2026, GoMining mined the first Bitcoin block where the miner (not the pool) selected transactions, using V2's Job Declaration feature.

The most important development came in May 2026, when seven major pools joined the Stratum V2 Working Group: Foundry, AntPool, F2Pool, SpiderPool, MARA Pool, Block Inc., and DMND. These pools collectively represent approximately 75% of global hashrate.

However, signaling support and running V2 in production are different things. As of mid-2026, only Braiins Pool and DMND run Stratum V2 with Job Declaration in production, representing roughly 3–5% of hashrate. The gap between signaling and deployment is the critical metric to watch.

OCEAN and the DATUM Protocol

OCEAN is a mining pool founded by Bitcoin Core developer Luke Dashjr in November 2023, backed by a $6.2 million seed round led by Jack Dorsey. Its approach to decentralization centers on the DATUM protocol (Decentralized Alternative Templates for Universal Mining), launched in September 2024.

Under DATUM, miners build their own block templates using their own full nodes. The pool receives only merkle branches sufficient to verify proof of work, never seeing the actual transaction list. This makes it structurally impossible for the pool operator to censor transactions. OCEAN also uses non-custodial payouts: miners are paid directly from block coinbase rewards, so the pool never holds miner funds.

OCEAN currently holds approximately 2% of hashrate. In April 2025, Tether announced it would deploy hashrate to OCEAN, providing a meaningful endorsement for the decentralized mining model. The pool charges 2% standard fees, reduced to 1% for miners using DATUM.

P2Pool and Other Decentralized Alternatives

The original P2Pool is effectively defunct for Bitcoin mining due to three structural problems: variance was too high for small miners, coinbase outputs paying thousands of participants made blocks uncompetitively large, and it could not keep pace with ASIC hashrate growth. A Rust-based rewrite using libp2p is under development but not yet production-ready.

In practice, Stratum V2 with Job Declaration has emerged as the pragmatic successor to P2Pool's decentralization goals. It preserves the economic benefits of pooled mining (variance reduction, predictable income) while returning transaction selection power to individual miners. Other options include public-pool.io, a zero-fee solo mining pool for miners willing to accept high variance.

Comparing Decentralization Approaches

What Mining Centralization Means for Layer 2s

Every Bitcoin Layer 2 protocol inherits the censorship resistance properties of the base layer. If the base layer can be censored, Layer 2 exit transactions can be censored. This has direct implications for the security guarantees of systems like Lightning, Spark, sidechains, and rollups.

Consider a scenario where a user needs to broadcast a unilateral exit transaction from a Layer 2 to settle on Bitcoin L1. If pools controlling a majority of hashrate refuse to include that transaction (whether due to regulatory pressure, commercial interests, or technical filtering), the user's funds become effectively trapped. This is not a vulnerability in the Layer 2 protocol itself: it is a failure of the base layer's censorship resistance that propagates upward.

For Spark, which enables instant, self-custodial Bitcoin transfers off-chain, the guarantee of unilateral exit is a core security property. Users hold pre-signed exit transactions that can be broadcast to Bitcoin L1 at any time, but this guarantee is only as strong as the base layer's willingness to include those transactions in blocks. Mining decentralization is not an abstract concern for Layer 2 users: it is a prerequisite for the exit guarantees they depend on.

The Path Forward

Mining pool centralization is one of Bitcoin's most pressing structural risks, but it is not irreversible. Several dynamics work against permanent consolidation.

• Stratum V2 adoption is accelerating. With 75% of hashrate signaling support through the working group, the protocol infrastructure is being built. The question is deployment pace, not directional intent.
• OCEAN and DATUM demonstrate that miner-driven template construction works in production, even if market share remains small.
• Transparency tools like 0xb10c's miningpool.observer make censorship detectable. F2Pool's repeated OFAC filtering was caught precisely because independent researchers monitor block templates.
• Miners retain the option to switch pools. If a pool is caught censoring, economic self-interest (preserving Bitcoin's value) incentivizes miners to leave for non-censoring alternatives.
• Bitcoin Core's integration of Stratum V2 support reduces the barrier for miners to run their own template construction.

The most effective action individual miners can take today is to choose pools that support miner-driven template construction. For developers building on Bitcoin, understanding mining centralization dynamics is essential. Spark's architecture and the broader Bitcoin Layer 2 ecosystem depend on a base layer where no small group can selectively exclude transactions. Tools like the Spark SDK make it straightforward to build applications that leverage Layer 2 capabilities, but the long-term viability of those applications depends on continued progress toward mining decentralization.

This article is for educational purposes only. It does not constitute financial or investment advice. Bitcoin and Layer 2 protocols involve technical and financial risk. Always do your own research and understand the tradeoffs before using any protocol.