ARK Invest and Glassnode have published a joint study that compares the degree of decentralization of Bitcoin, Ethereum, and Solana from dimensions such as block production, holding distribution, verification costs, and infrastructure distribution. After considering six indicators, the study ranks Bitcoin first, Ethereum in the middle, and Solana in third place.
The thresholds for both Bitcoin and Ethereum are 3.
The report measures the block production concentration of three chains using different thresholds. Bitcoin is calculated based on 51% of the computing power, while Ethereum and Solana are calculated based on a 33% stake ratio, as one-third of the staked shares is sufficient to interfere with the final confirmation.
According to this metric, Bitcoin would only need 3 mining pools to exceed 51% of the computing power, which are Foundry USA, AntPool, and F2Pool, accounting for over 61% in total. As for Ethereum, Lido, Binance, and Kraken together account for about 38.8% of the staked ETH, also surpassing the 33% threshold.
However, the report emphasizes that this does not mean that the three institutions directly control Bitcoin or Ethereum. Mining pools aggregate the computing power of independent miners, and staking platforms may also bring together multiple node operators; therefore, the underlying resources are not entirely owned by a single entity.
Migration speed affects actual control power.
Research suggests that Bitcoin miners switch mining pools relatively quickly, which weakens the actual control represented by the concentration of mining pools. The report estimates that it would take about 29 seconds to transfer 1% of Bitcoin's computing power. If there is any form of review or coordinated behavior among mining pools, miners could theoretically leave promptly.
In contrast, the exit process for Ethereum validators is slower. Reports estimate that it takes about 14.6 days to exit 1% of staked positions, and in times of network congestion, it can take up to 55.6 days. This means that it is more difficult to quickly resolve the concentration of staking on Ethereum through migration in the short term.
The report also points out that the risks associated with Ethereum cannot be assessed solely based on the share of validators. Different clients independently implementing the protocol rules can, to some extent, mitigate the impact of a single software failure.
Solana Higher concentration of computer rooms
In terms of the single indicator of block production threshold, Solana has the highest Nakamoto coefficient. Research shows that 19 validators are required to control over 33% of the delegated staking, which is higher than that of Bitcoin and Ethereum.
However, the report also points out that Solana has a higher concentration of physical infrastructure. In the study sample, almost all of the relevant infrastructure is deployed in commercial data centers, with about 68% located in Europe and 21% in North America. The report mentions that an issue with TeraSwitch routing in August caused 102 out of 699 Solana validators to stop voting. Although the network continued to process transactions, the incident demonstrated that a failure in shared infrastructure could affect multiple independent validators simultaneously.
Significant verification cost differences observed
The study also compared the verification and reconstruction costs of three chains. The estimated hardware cost for a full Bitcoin node is about $289, which is lower than Ethereum’s $730 and far lower than Solana’s $21,478.
- The complete Bitcoin blockchain data is approximately 753 GB.
- The complete archiving of Ethereum requires approximately 2 TB.
- Solana The scale of historical reconstruction is approximately 480 TB
In terms of hosting distribution, Bitcoin is also more decentralized. Studies indicate that only 16% of the relevant infrastructure is located in data centers; 63% of the nodes operate anonymously over a Tor network, while another 15% are in residential or self-hosted environments. For Ethereum, approximately 49% of its execution layer nodes are located in cloud environments, of which AWS accounts for about 20%.
The report concludes that a single indicator is not sufficient to draw a definitive conclusion about the decentralization of blockchain. Mining pools, exchanges, staking protocols, and custody platforms can all exacerbate apparent centralization, and they can also lead to different risks in various dimensions across different networks. Overall, Bitcoin performs best in terms of distribution of holdings, auditability, and geographical resilience.
