ARK Invest and Glassnode’s latest report puts Bitcoin ahead of Ethereum and Solana on the decentralization measures it studied, and, annoyingly for the “everything is a scam except my bag” crowd, the numbers are hard to dismiss.
- Bitcoin ranked most decentralized overall across the measures studied
- Ethereum came in the middle, with real resilience and real choke points
- Solana showed strong validator spread but heavier infrastructure concentration
- Concentration is not ownership, but it can still be a serious risk
The 32-page report, The Decentralization Spectrum: Design Tradeoffs in Digital Assets, was published on Sept. 1 by ARK Invest and Glassnode. It compares Bitcoin, Ethereum, and Solana across ownership, exit fluidity, verification costs, critical resilience, reconstruction costs, and infrastructure distribution, with a broader framing echoed in Bitcoin leads Ethereum and Solana in decentralization, ARK.
The core message is simple. Decentralization is not one number and it is not one cute chart. A network can look broad on one layer and brittle on another. Mining pools, staking platforms, cloud hosts, and client software all create different kinds of pressure points.
That distinction matters because concentration metrics are easy to misuse. A mining pool is not automatically a miner. A staking platform is not the same thing as every person who delegates through it. And a hosting provider is not the protocol itself. Still, coordination points can become choke points fast if the wrong incentives show up.
Why the report matters
Before getting into the chains themselves, the methodology deserves a plain-English pass.
Verification costs refer to what it takes to run and verify the network. If node operation becomes expensive or cumbersome, fewer independent operators will bother.
Exit fluidity measures how quickly participants can move away from a coordinator or platform. In practice, that means how easy it is to switch pools, withdraw stake, or change infrastructure.
Critical resilience is about how much pressure a network can take at key thresholds before the system starts to wobble in a meaningful way.
Reconstruction costs are the resources needed to rebuild or archive chain history. If those costs are extreme, the network becomes more dependent on well-funded operators.
Those are the right questions to ask. Decentralization is a stack of tradeoffs, not a marketing slogan, and Understanding the Impact of Climate Change on Global is not the only kind of systems-level risk that deserves a sober look.
Bitcoin: messy at the pool layer, durable in practice
The report found that three Bitcoin mining pools together exceeded the 51% hash-rate threshold. Foundry USA accounted for 27.27%, AntPool for 17.06%, and F2Pool for 16.96%. Together, those three represented more than 61% of measured hash rate. ViaBTC came in at 9.50%, and SpiderPool at 5.82%.
That looks ugly if you stop at the headline. But the report’s point is that mining pools are coordination points, not necessarily the actual owners of the hash power behind them. Miners can move their work elsewhere, and the report estimated that a Bitcoin miner could switch a 1% hash-rate position in about 29 seconds.
That mobility matters. Pool concentration is real, but it does not translate cleanly into hard control over the network. Censorship risk and protocol ownership are not the same thing. Mixing them up is how people end up making loud, stupid arguments with fake confidence. If you want a practical primer on the mechanics, How to Choose a Bitcoin Mining Pool is a better starting point than most Telegram brain fog.
Bitcoin also remains relatively cheap to verify. The report estimated full node hardware at $289, with full-chain storage at 753 gigabytes. It found that 63% of measured infrastructure used Tor, the anonymity network that helps hide node location. Another 15% was residential or self-hosted, while 16% operated in data centers.
That is the sort of operational profile that still makes Bitcoin the cleanest example of rough-and-ready decentralization at scale. Running a node is not trivial, but it is still within reach for ordinary people. That matters more than abstract purity tests and internet warrior theater.
Ethereum: still decentralized, but with obvious coordination layers
Ethereum’s picture is more layered. ARK and Glassnode used a 33% stake threshold rather than Bitcoin’s 51% framing because one-third of staked ETH can disrupt finality. Finality is the point at which transactions are considered settled enough that rolling them back becomes a serious consensus problem.
That is the right lens for proof-of-stake. Copy-pasting proof-of-stake language onto Ethereum is lazy and usually wrong.
On staking concentration, the report found Lido at 23.04%, Binance at 8.88%, and Kraken at 6.91%. Combined, those three represented roughly 38.8% of the stake concentration measured in the report’s framework. That does not mean those platforms “own” all that ETH. It does mean they sit at important coordination layers, and those layers can matter a great deal when pressure hits.
The report also examined client diversity. Geth led execution clients at 34.88%, followed by Nethermind at 26.96% and Reth at 18.98%. Lighthouse accounted for 54.16% on the consensus side.
That client mix is better than many chains can claim, but software monoculture is still a quiet risk. If too much of the network depends on one implementation, bugs and exploit paths can spread fast. Ethereum is not broken. It is just not magically immune to “too much of the network runs the same damn code” syndrome.
Infrastructure is another pressure point. Around 49% of execution-layer nodes were in cloud environments, 20% were hosted by Amazon Web Services, and the top two cloud providers accounted for about 27%. The report estimated full node hardware at $730, with a full archive setup requiring around two terabytes.
Ethereum remains meaningfully decentralized. But it has clear coordination layers in staking, software, and hosting. That is not a knock so much as a reality check. Complexity buys capability, but it also buys new failure modes.
Solana: broad validator spread, heavier infrastructure dependence
Solana’s strongest showing came on validator dispersion. The report found that 19 validators were needed to control more than 33% of delegated stake. That is the correct figure to use here, since the chart, comparison table, and Glassnode’s summary all point to 19.
The top Solana validators in the report were Figment at 3.78%, 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. A deeper breakdown is available in Analyzing Solana's Decentralization: Criticisms, Metrics.
That is a respectable distribution by stake metrics. Solana skeptics who pretend the validator set is one giant neon centralization sign are overselling it. But the infrastructure picture is where the heat comes back.
Approximately 100% of measured Solana infrastructure operated in commercial data centers. About 68% was in Europe and 21% in North America. TeraSwitch alone accounted for 30.23% of measured stake, while the top two hosting companies served around 35.7%.
That is the tradeoff in plain English: validator breadth on one side, infrastructure concentration on the other. If a network depends heavily on professional hosting and a narrow set of data-center environments, then outages, policy changes, or regulatory pressure can have outsized effects. No amount of speed-pilled branding changes that.
The report also put real numbers on Solana’s operating burden. A Solana RPC node or validator-class configuration was estimated at $21, 478, and reconstructing Solana’s history at roughly 480 terabytes.
Performance-first design is not free. It usually asks more from operators, and sometimes a lot more. That is not a moral failure, but it is a tradeoff worth stating plainly rather than hand-waving away with meme fuel.
Concentration is not control, but it is still danger
The best part of the report is also the most annoying for lazy takes: concentration metrics can be misleading if read too literally.
A pool can coordinate hash rate without owning the miners. A staking platform can aggregate delegated stake without controlling every end user’s ETH. A hosting provider can concentrate infrastructure without being the network itself.
That is the nuance a lot of “my chain is pure, yours is centralized trash” commentary skips. On the flip side, the existence of nuance does not make concentration harmless. Coordination layers can still become censorship points, failure points, and regulatory choke points.
In other words, both of these things are true at once: concentration is not the same as ownership, and concentration still matters a lot. Pretending otherwise is just intellectually lazy.
What the numbers actually say
Read cleanly, the report’s ranking is straightforward:
Bitcoin scores best on the measures studied, helped by relatively low node costs, Tor usage, and fast miner mobility between pools.
Ethereum sits in the middle, with real decentralization but more obvious coordination layers in staking, client software, and cloud hosting.
Solana shows strong validator spread, but leans more heavily on commercial data centers and concentrated hosting infrastructure.
That is not a morality play. It is an engineering tradeoff. And engineering does not care about your favorite token logo.
That same tradeoff lens shows up in adjacent coverage too, from Russia Legalizes Bitcoin, Ethereum, Solana with Strict rules to market cycles in Bitcoin, Ethereum, XRP, Solana Rebound: 2024 Price Analysis and longer-horizon speculation in Top Cryptos for 2026: Bitcoin, Ethereum, Solana & Hidden.
Key takeaways
- Why did Bitcoin rank highest?
Because, on the measures studied, it combined relatively low node costs, strong Tor usage, and fast miner mobility between pools, which made its decentralization more durable in practice. - Does three Bitcoin pools crossing 51% mean they own the network?
No. Pools are coordination points, not necessarily the owners of the underlying mining hardware. But the concentration is still a real risk. The academic angle on Bitcoin transaction fees and the decentralization of mining incentives gets into why those tradeoffs matter. - Why does Ethereum use a 33% threshold?
Because proof-of-stake has different security dynamics. In Ethereum, one-third of staked ETH can disrupt finality, so the report uses that benchmark instead of Bitcoin’s 51% framing. - What is Ethereum’s biggest centralization pressure?
Staking concentration at large providers, plus client and cloud dependency. The network is decentralized, but it is not carefree. - What is Solana’s strongest point?
Validator/stake distribution. The report found that 19 validators were needed to cross the 33% delegated-stake threshold. - What is Solana’s biggest weakness?
Infrastructure concentration, especially reliance on commercial data centers and a relatively narrow set of hosting environments. - Why does the timing of the data matter?
Because the datasets were not all taken at the same moment. Solana’s geographic data came from November 2024, while most Bitcoin and Ethereum infrastructure data came from July 2026.
The least biased read is also the most useful one: Bitcoin scores best on these decentralization measures, Ethereum sits in the middle, and Solana trades more infrastructure concentration for speed and scale. That tradeoff may be worth it for some users and builders. Just don’t pretend the bill never comes due.
