Bitcoin leads Ethereum and Solana in decentralization, ARK finds

ARK Invest and Glassnode published a joint study on Sept. 1 that found three entities could cross the measured block-production thresholds for both Bitcoin and Ethereum, while Solana required 19.

The 32-page report, titled The Decentralization Spectrum: Design Tradeoffs in Digital Assets, compares the networks across ownership, exit fluidity, verification costs, critical resilience, reconstruction costs and infrastructure distribution.

The findings do not mean three companies control Bitcoin or Ethereum. The metric counts mining pools and staking platforms as entities, even when the underlying hardware, stake or node operators belong to separate participants who may withdraw or redirect their resources.

Bitcoin’s three-pool threshold does not equal ownership

The report applied a 51% hash-rate threshold to Bitcoin. Foundry USA represented 27.27% of the measured hash rate, followed by AntPool at 17.06% and F2Pool at 16.96%. Together, the three pools exceeded 61%.

This produced a Nakamoto coefficient of three, defined as the minimum number of measured entities needed to cross a network’s critical production threshold. ViaBTC controlled another 9.50%, while SpiderPool represented 5.82%.

Mining pools coordinate block construction and distribute rewards, but they do not necessarily own the machines producing their hash rate. Independent miners connect to pools to receive steadier income and can redirect their computing power elsewhere.

That mobility limits how closely pool concentration can be equated with permanent control. The report estimated a Bitcoin miner could switch a 1% hash-rate position in approximately 29 seconds. A coordinated attack or censorship attempt could prompt participants to leave the responsible pools.

Pools still influence transaction inclusion and ordering because they usually provide the block templates miners use. Pool concentration therefore represents an operational risk, even if it overstates the concentration of underlying mining ownership.

The issue is not new. Earlier crypto.news reporting found that two mining pools produced a majority of sampled Bitcoin blocks in late 2022. Pool shares have changed since then, but production continues to be concentrated among several large coordinators.

Ethereum crosses a lower threshold through pooled stake

ARK and Glassnode applied a 33% stake threshold to Ethereum because participants controlling one-third of staked $ETH can disrupt finality. This differs from Bitcoin’s 51% majority threshold, so the two coefficients do not describe identical powers.

Lido represented 23.04% of staked $ETH in the report’s July data. Binance controlled 8.88%, and Kraken held 6.91%. Those three entities collectively represented approximately 38.8%, taking Ethereum above the selected threshold.

Lido is not a single validator. It distributes stake among multiple node operators, although those operators participate through a common protocol and governance framework. The report therefore treats Lido as shared infrastructure that aggregates economic weight rather than one machine or company directly controlling every validator.

Ethereum’s exit mechanics also restrict validator mobility. The report estimated that exiting a 1% position would take around 14.6 days under current conditions and as long as 55.6 days under heavy congestion. That is much slower than redirecting Bitcoin hash rate.

Client diversity provides another layer of resilience. The study placed Geth’s execution-client share at 34.88%, followed by Nethermind at 26.96% and Reth at 18.98%. Lighthouse represented 54.16% of consensus clients.

Different clients independently implement Ethereum’s rules, reducing the portion of the network exposed to one software defect. The relationship between Ethereum nodes and their software clients means validator concentration alone cannot describe the network’s full failure risk.

Solana’s 19-validator result comes with infrastructure costs

Solana recorded the highest Nakamoto coefficient for the selected block-production threshold. The report found that 19 validators were needed to control more than 33% of delegated stake.

Figment was the largest individual validator at 3.78%, followed by Helius at 3.69%, Jupiter at 2.91%, Binance Staking at 2.81% and Ledger by Figment at 2.16%. The remaining 84.65% was spread across other validators.

One passage in the report says Solana requires 20 entities, but its chart, comparison table and published Glassnode summary all report a coefficient of 19. The table also says the figure increased from 18 in March 2026.

Solana’s validator distribution performed well on this particular measure, but its physical infrastructure was more concentrated. Approximately 100% of the infrastructure measured by the researchers operated in commercial data centers. About 68% was in Europe, while 21% was in North America.

TeraSwitch hosted 30.23% of measured stake, and the top two hosting companies served around 35.7%. Common infrastructure can create correlated failures even when the validator set contains many separate operators.

That risk became visible in August when 102 of 699 Solana validators stopped voting during a TeraSwitch routing problem. Solana continued processing transactions, but the episode showed how one infrastructure failure can affect multiple otherwise independent validators.

The report used Solana geographic data from November 2024, while most Bitcoin and Ethereum infrastructure data came from July 2026. That timing difference limits direct comparisons and leaves room for Solana’s distribution to have changed.

Bitcoin leads infrastructure resilience and auditability

Bitcoin had the least expensive verification requirements in the study. The researchers estimated hardware for a full node at $289, compared with $730 for Ethereum and $21,478 for a Solana RPC node or validator-class configuration.

Its measured full-chain storage requirement was 753 gigabytes. Ethereum required approximately two terabytes for a full archive setup, while reconstructing Solana’s history was estimated at 480 terabytes because historical data is commonly offloaded to external providers.

Bitcoin also had the most distributed hosting profile. Only 16% of measured infrastructure operated in data centers, while 63% of nodes used Tor. Another 15% was residential or self-hosted.

Ethereum placed approximately 49% of execution-layer nodes in cloud environments and 45% in self-hosted settings. AWS alone hosted around 20%, while the top two providers accounted for approximately 27%.

Solana’s higher hardware and bandwidth demands reflect its focus on throughput. The tradeoff is that fewer ordinary users can independently recreate or verify the full network history using consumer equipment.

No single score settles blockchain decentralization

The report ultimately ranked Bitcoin as the most decentralized of the three networks overall, followed by Ethereum and Solana. Bitcoin led in ownership distribution, auditability and geographic resilience.

Ethereum generally occupied the middle across the six dimensions. Solana scored strongly for its critical resilience threshold and validator participation but ranked lower for ownership distribution, verification accessibility and infrastructure diversity.

The methodology remains sensitive to how entities are grouped. Exchanges can hold tokens for many customers, mining pools aggregate independent miners, and staking protocols coordinate multiple operators. Wallet-size bands can likewise combine custodial assets belonging to thousands of users.

The comparison is therefore more useful as a map of separate concentration risks than as a definitive ranking. A network may distribute block production broadly while relying heavily on several hosting companies, software clients or governance organizations.

Future editions could improve comparability by using synchronized data dates, separating pools from underlying resource owners and distinguishing censorship thresholds from thresholds capable of rewriting finalized history.

FAQs