Staked ETH Is Concentrating in Owners, Not Yet in Operators

Executive Summary

Institutional capital now sits at the center of Ethereum staking, but the evidence supports a narrower claim than "centralization." Bitmine Immersion Technologies (NYSE: BMNR) held 6,016,414 ETH on October 5, 2026, equal to 4.9% of supply, and had staked 5,067,309 ETH. That stake is about 11.6% of all staked ETH. This is a capital-concentration story. It does not show that one operator controls that share of validators.

The report separates three layers that headlines tend to merge. Layer one is who owns staked ETH: Bitmine, ETFs and corporate treasuries. Layer two is who operates validators: Lido's curated operators, Bitmine's MAVAN platform and its unnamed partners, and exchanges. Layer three is how the validator set is structured: post-Pectra 0x02 validators are letting stake collapse onto fewer, larger validators. Lido's Curated Module v2 alone is expected to cut the validator count by about one-third.

Three findings matter most. Bitmine's reported seven-day annualized yield of 2.63% is close to the network staking APR, so the economics look like balance-sheet scale, not operator alpha. Lido's share of all staked ETH is about 22–23%, down from a peak of about 32% in 2023. And the evidence on MEV and builder or relay distribution is thin. This report does not assert any MEV concentration figure, and it says where the gap lies.

Market Context

ETH traded at $2,566.19 on October 8, 2026, with a market capitalization of $313.29 billion and a circulating supply of 122,084,755 ETH (DefiLlama, CoinGecko id ethereum). At that price Bitmine's holdings are worth about $15.4 billion, and its staked balance about $13.0 billion (computed). Bitmine's own release valued the staked ETH at $13.8 billion using $2,726 per ETH on October 5. The gap is price, not a change in holdings. Against live circulating supply, Bitmine's share computes to 4.93%.

The staking base itself is large and still growing. The Validator Queue dashboard, which sources beaconcha.in data, showed 43.7M ETH staked, equal to 35.78% of supply, across 857,234 active validators when retrieved on October 8. The network staking APR was 2.59%. In mid-June 2026, a Bitcoin.com News summary of Datawallet data put staked ETH at about 39.6M. That implies roughly 4.1M ETH was added in about 16 weeks, though the two sources use different methods.

Entry is rate-limited, which shapes everything below. The entry queue held 1,389,996 ETH with a 24-day wait at retrieval, down from 3,589,414 ETH and a 62-day wait on May 20 (KuCoin Blog, secondary). The exit queue held 831,614 ETH with a 14.4-day wait. The cause of that exit queue is not established in the sources reviewed, so this report does not attribute it to any event.

On the DeFi side, DefiLlama showed the Ethereum chain at $52.18 billion TVL on October 8. Lido's TVL was $24.96 billion, Binance Staked ETH $9.55 billion, EigenLayer $6.71 billion, ether.fi $5.32 billion, Rocket Pool $1.34 billion and Coinbase Wrapped Staked ETH $487 million.

Three Layers, Not One: Separating Ownership, Operation and Structure

Most commentary treats "who stakes" as one question. It is three, and each has different evidence and different risks. Ownership determines who earns the rewards and who bears price risk. Operation determines who can be slashed, who can fail to attest and who makes block-building choices. Validator-set structure determines how much work the network does and how much stake each key controls.

A single entity can be large in one layer and small in another. Bitmine is the clearest example. Its ownership footprint is documented to the token. Its operating footprint is not documented in detail, because it stakes through the MAVAN platform and partners that the sources do not name.

Layer 1: Who owns staked ETH

Bitmine launched its ETH treasury strategy on June 30, 2025 and has bought ETH every week since. Its holdings rose from about 5.74M ETH in early July 2026 to 6.02M in early October, so buying has slowed relative to its earlier pace. The most recent weekly purchase was 15,112 ETH. About 84% of the treasury was staked as of October 5. Its release puts total crypto, cash and marketable securities holdings at $17.4 billion, including $643 million of cash and marketable securities.

Bitmine's staked ETH grew from roughly 4.9M when quarterly results were reported in July to 5,067,309 by October 5. That is an increase of roughly 0.17M ETH. Over a longer window, network-wide staked ETH rose by about 4.1M. The windows differ, so this is a rough comparison, but it suggests that most inflows came from other entrants, though the windows do not match. The sources reviewed do not identify them.

Bitmine also has a stated ceiling. Chairman Tom Lee said at Token2049 Singapore on October 7 that the company will not hold more than 5% of Ethereum, calling it a hard cap (reported by Cryptoast). Bitmine was within about 88,000 ETH of that level on October 5 (computed from 5% of 122.1M supply). Bitmine's stated cap implies its weekly purchases would end within roughly six weeks at the latest pace, which would remove a steady buyer from the market. That is our inference.

Ownership is not limited to treasuries. Grayscale activated staking in its Ethereum products in October 2025, the first US issuer to do so. Its Ethereum Staking ETF (ETHE) had 84.6% of its ETH staked in September 2026, per Cointelegraph citing the fund's webpage. Total ETF and peer-treasury holdings and their staked shares were not retrieved, so the sum of institutional ownership cannot be stated.

Layer 2: Who operates validators

Ownership share is not operation share. Bitmine describes MAVAN (Made in America VAlidator Network) as a staking destination for itself and institutional investors. The Block reports the ETH is staked "through the company's MAVAN platform and staking partners." The partners are not named in the sources reviewed. A $280 million preferred stock raise funded ETH accumulation and the build-out of the platform, according to Yahoo Finance's summary of company disclosures. MAVAN was originally built for Bitmine's own ETH and is now pitched to institutions.

This leaves a real gap. Bitmine's 11.6% share of staked ETH could be run by a handful of operators, by many, or partly by Bitmine itself. None of that is observable from the sources. The correct statement is that ownership concentration is documented and operator concentration behind it cannot be assessed.

Lido is the best-documented operator layer, and it is a curated one. Its Curated Module handles well over 90% of Lido's staked ETH and relies on DAO-approved professional node operators. Lido held 8.89M ETH in mid-June 2026, a 61.66% share of the liquid staking market. Binance Staked ETH was second at 3.66M ETH (25.37%) and Rocket Pool held 529,406 ETH. The 61.66% figure is a share of liquid staking, not of all staked ETH. Against 39.6M ETH staked at that time, Lido's share of all staked ETH was about 22.4% (computed). Datawallet puts its peak at about 32% in 2023.

That trend matters for the thesis. Lido's share has fallen from a peak of about 32% (Datawallet, 2023) to about 22–23%, though the methods differ, while total staking grew. Yet 22–23% held by one protocol, operated by a curated set, is still a large concentration. Entity-level shares for Coinbase, Kiln, Figment, Kraken and other operators were not retrieved, so this report does not rank them.

Exchanges form the third group of operators, but the evidence is partial. Binance Staked ETH and Coinbase Wrapped Staked ETH appear in the TVL data, which shows their liquid-token footprint. It does not show how many validators sit behind them or how many of their users' deposits are staked natively.

Layer 3: How the validator set is structured

The third layer is the least intuitive and the most misread. Pectra, activated on Ethereum mainnet in May 2025, raised the maximum effective balance per validator from 32 ETH to 2,048 ETH through 0x02 validators (EIP-7251). That change lets stake sit on fewer validators. It does not change who owns the stake or who operates it.

Lido is the largest single user of that option in this dataset. On July 27, 2026 it launched Curated Module v2 (CMv2), which begins consolidating more than 8M ETH onto 0x02 validators. At the time that was about $16 billion, roughly a fifth of all staked ETH. Lido said the share of its stake on 0x02 validators rises from about 32% to about 52%. It also said the move cuts Ethereum's total validator count by about one-third.

A secondary report of Lido's own blog gives more precise numbers. CMv2 enables migration of more than 265,000 existing validators from 0x01 credentials. The validator set would fall from about 880,000 to an estimated 628,000, about 29% fewer attestation messages per epoch. The "one-third" validator-count figure (The Block) and the 880,000-to-628,000 estimate, about 29% (Cryptonews), differ, and the sources reviewed do not explain why. Both point the same way.

Two cautions apply. First, the dashboard showed 857,234 active validators on October 8, well above the 628,000 projection. Either the migration is incomplete or it has not yet shown up in the network-wide figure. The CMv2 timeline to completion was not verified. Second, a falling validator count is a load metric, not a decentralization metric. Fewer, larger validators reduce beacon-chain messaging. They also mean each key, and each operator running it, controls more stake. A fault hits more ETH per validator. CMv2 requires Lido's node operators to post their own ETH as a bond that can be seized for slashing, improper execution-layer rewards and other failures. Lido says the bond is smaller than a fully permissionless module would require. Isidoros Passadis of Lido called the release the biggest change to Lido Core since Lido V2.

The same upgrade shipped Community Staking Module v3 (CSM v3). It adds a Distributed Validator Technology route that splits validator keys and duties across four independent operators. The same bonded capital can support more staked ETH, which Lido presents as making the module more accessible to smaller community stakers. Whether smaller stakers will use it at scale is not yet shown in the data.

The structural conclusion is that consolidation and decentralization are separate axes. The network can become lighter to run while ownership stays concentrated. It can also become more concentrated by operator while individual validators look more numerous. Counting validators alone cannot settle either question.

Why Institutional Yield Is Roughly the Network Base Rate

Bitmine said its own staking operations produced a seven-day annualized yield of 2.63%. The network-wide staking APR on the Validator Queue dashboard was 2.59% when retrieved on October 8. The gap is 0.04 percentage points. The two measures use different windows and methods, so the comparison is indicative, not exact. Still, the headline yield shows no large operator premium.

That fits how Ethereum's reward system works. Consensus-layer base rewards depend on protocol rules and total stake, so a larger balance does not by itself raise the percentage earned (to be verified on ethereum.org). Execution-layer rewards may differ by operator, and no data on that was retrieved. Bitmine's advantage is the size of the balance, not the percentage earned on it. At that yield, its staked balance generates about 133,000 ETH per year (computed). At the live price that is about $342 million. Bitmine's own projection of about $363 million used $2,726 per ETH, and the difference is entirely price.

This distinction matters for how to read treasury staking. The ETH yield is stable and protocol-driven. The dollar revenue moves with the ETH price. A treasury that stakes 84% of its holdings earns a slowly accruing yield, but unstaking is subject to the exit queue (14.4 days at retrieval) and sweep delay (7.4 days), so that part of the balance cannot be sold immediately.

For professional operators the pressure runs the other way. Lido's total revenue fell 23% in 2025 as users withdrew funds and yields declined, per The Block's coverage of its annual report. Operator fees are a margin on the base yield, which may be falling. A secondary source (KuCoin Blog) put native staking APR at about 2.78% across roughly 897,000 active validators in May 2026. Different methods make a direct comparison with the dashboard figure unreliable, but the direction is consistent with a lower base rate.

Margin pressure could push the market two ways. Operators may consolidate to cut costs, which would raise concentration. Or bonded, distributed designs such as CSM v3 may lower the capital needed to enter, which would lower it. The evidence is too early to say which dominates. The bond requirement in CMv2 raises the capital cost of being a curated operator, and the DVT route is built to spread that cost across more participants.

The same logic applies to MAVAN. If yield is the base rate, a platform selling staking to institutions competes on custody, compliance, reporting and trust. Those are service differentiators, not return differentiators. That is our reading, not a statement from Bitmine, and it suggests the unnamed partners matter more than the yield.

Queues Make Ownership Sticky

The entry and exit queues are the mechanism that translates institutional demand into network structure. The churn limit processes about 256 ETH per epoch for both entries and exits, and an epoch lasts about 6.4 minutes. Large inflows therefore turn into multi-week waits.

The queue history shows how fast conditions change. Early-2026 reports described the exit queue falling to zero from a September 2025 peak of about 2.67M ETH, and January 2026 entry queues were near zero. By May 20 the entry queue was 3,589,414 ETH, a 62-day wait. By October 8 it was 1,389,996 ETH, a 24-day wait.

Two implications follow. First, because entry is throttled, the ownership mix changes slowly, and existing stake is sticky. Second, because the churn limit caps total entries, any large entrant, Bitmine included, adds to a shared queue and slows everyone else's entry. This limits how fast the staked share can change, though it does not stop a large buyer from taking a big share of the queue. It also means a large position, once staked, is expensive to rebuild if exited. An exit queue of 831,614 ETH exists today, and the sources do not explain it. Anyone attributing it to consolidation, selling or institutional rotation is speculating.

The Regulatory On-Ramp Pulling Passive Capital In

Institutional staking is also a regulatory story. Per Astraea Law, the SEC dismissed its Coinbase action in February 2025. Staff statements in May 2025 (protocol staking) and August 2025 (liquid staking receipt tokens) said these are not securities offerings. The SEC approved generic listing standards in September 2025. Grayscale activated ETF staking in October 2025, and a US spot ETH ETF paid its first staking distribution in January 2026. By contrast, Kraken paid $30 million and shut its US staking service in February 2023.

Per Luganodes (secondary) and Astraea Law, the SEC and CFTC issued a joint interpretive release on March 17, 2026 classifying staking rewards as non-securities across 16 digital commodities including ETH. Astraea calls it a Commission-level formally published interpretation, stronger than the 2025 staff statements, but not a binding joint rule. Its title says some matters "only Congress can settle."

The effect on structure is indirect but real. Staking ETFs convert passive allocations into staked ETH, which requires custodians and node operators. Which operators they use is not in the sources. So regulatory clarity adds to ownership concentration in regulated vehicles, and it routes stake to operators whose identities are unknown to this analysis. That is a question for the next round of disclosure, not a conclusion.

What the Evidence Cannot Say: MEV, Builders and Relays

The sources reviewed contain no primary data on MEV-Boost relay market share, builder concentration, proposer-builder dynamics or the split of MEV revenue between institutional and retail validators. This report therefore makes no claim about any of them. The gap is material, because block-building is where operator concentration could translate into influence over transaction ordering.

The yield evidence is also silent on this point. The comparison between Bitmine's reported yield and the network rate suggests no large premium in headline yield. It does not rule out differences in how rewards are composed, because the windows and methods differ. A secondary exchange blog claims MEV adds a meaningful uplift for users of MEV-Boost. That claim is not traced to primary data and is treated here as unverified. We do not rely on it.

What would close the gap is specific. Relay and builder market-share dashboards over a multi-month window would show builder concentration. Operator-level reward breakdowns, split into consensus and execution-layer rewards, would show whether professional operators earn more per ETH. Disclosure of MAVAN's partners and their relay settings would show where Bitmine's stake feeds into block production. Until that data is in hand, MEV centralization is an open question, not a finding.

Data: The Three Layers at a Glance

The table below summarizes the measured values for each layer and the reading they support. Dates differ across rows, as noted.

Sources: Bitmine release via PR Newswire, Validator Queue (beaconcha.in data), Bitcoin.com News and Datawallet, Cryptonews citing Lido's blog, Cointelegraph, KuCoin Blog (secondary). Derived figures are computed by this report.

The second table compares what the evidence supports for each common claim.

Risks

Unknown operator concentration behind MAVAN. Rating: Medium. Bitmine's stake runs through unnamed partners. The risk becomes High if disclosure shows that a few operators run most of it, or if those operators overlap with other large stakers. It falls to Low if the partners are many and disclosed.

Fault concentration on larger validators. Rating: Medium. A 0x02 validator can hold up to 2,048 ETH, so a slashing or outage event affects more ETH per key. The trigger is an operator failure on consolidated validators. Lido's seizable operator bond is designed to limit the damage, though Lido itself says that bond is smaller than a fully permissionless module would require.

Ownership concentration without a binding limit. Rating: Medium. Bitmine's 5% cap is a company commitment, not a protocol rule. The trigger is a change of policy, a corporate transaction, or another treasury building a comparable position.

MEV and builder concentration. Rating: Medium, unquantified. The data is absent, not reassuring. The rating reflects uncertainty. It would rise if relay or builder dashboards showed a few entities handling most blocks, and fall if they showed broad distribution.

Regulatory reversal or ambiguity. Rating: Medium. The March 2026 SEC and CFTC release is an interpretation, not a binding joint rule, and Astraea says some matters need Congress. A shift in agency posture or a court challenge could slow ETF staking. That would reduce the passive inflow described above.

Treasury liquidity and price stress. Rating: Medium. With about 84% of Bitmine's treasury staked, the ETH is locked behind the exit queue and sweep delay described earlier. A sharp ETH decline combined with a funding need could force sales at poor timing. The sources reviewed do not give Bitmine's financing terms beyond the $280 million preferred raise, so the severity cannot be sized.

Operator margin compression. Rating: Low to Medium. Lido's 2025 revenue fell 23%. If the base yield keeps falling, smaller operators may exit, and the set of professional operators could narrow. The trigger is a sustained drop in the network APR below current levels.

Outlook

The next three to six months should show whether ownership concentration continues to grow while operator concentration eases. The probabilities below are our estimates, based on the trends documented above. They are not facts.

Base case (55%). Bitmine approaches but does not exceed its 5% cap. Total staked ETH continues to grow, and queue waits stay well below the May peak. Lido's share edges lower while CMv2 migration reduces the network-wide validator count toward the projected range. MAVAN disclosure remains partial. This is the most likely path because every documented trend is moving gradually: buying has slowed, queues have eased, and Lido's consolidation has begun.

Upside case (20%). Bitmine or other large holders disclose their operator partners, and the set is broad. CSM v3's DVT route attracts meaningful community stake. Relay and builder data show wide distribution. In this case the concentration debate narrows to ownership, which is a smaller concern than operation.

Downside case (25%). A slashing or outage event on consolidated validators, a sharp ETH price decline, or evidence of concentrated operators behind large treasuries shifts the debate toward operator risk. Exit queues lengthen, and regulators or Congress revisit staking treatment. We rate this at one in four given the open risks above.

For traders. Watch the entry and exit queues and the active validator count on the dashboard. A rising exit queue with no explained cause is a signal to monitor, not to interpret. Bitmine's stated cap implies its weekly purchases would end within roughly six weeks at the latest pace. That is our inference. Dollar staking revenue moves with price, so treat treasury yield narratives with that in mind.

For builders. The opening is in distributed and bonded designs. CSM v3's four-operator DVT route and CMv2's operator bonds show where demand for tooling lies. Reporting that separates ownership, operation and block-building would fill a visible data gap, as would operator-level reward breakdowns.

For investors. Do not treat staked ETH ownership as a proxy for operator control, and do not treat validator-count declines as decentralization or centralization on their own. The useful diligence questions are who the operators are, how bonds and slashing are handled, and how yield is composed. Until MAVAN's partners and MEV distribution are disclosed, assume the headline yield is the base rate and judge institutional staking platforms on service quality and risk controls.

Sources

• PR Newswire: Bitmine press release on ETH holdings, total holdings and cash, October 5, 2026
• The Block: Bitmine staking through MAVAN and partners, weekly buying pace
• crypto.news: Bitmine 2.63% seven-day yield, projected staking revenue, 84% staked
• Cointelegraph: Bitmine staked ETH growth and September 2026 holdings, ETHE staking share
• Cryptoast: Tom Lee on Bitmine's 5% hard cap at Token2049
• Yahoo Finance: MAVAN launch and $280M preferred raise
• Validator Queue (data: beaconcha.in): staked ETH, validators, APR, queues, churn
• The Block: Lido Curated Module v2, CSM v3, Pectra context, Lido 2025 revenue
• Cryptonews (citing Lido blog): CMv2 validator migration and validator set projection
• Bitcoin.com News (Datawallet): staked ETH in mid-June 2026, Lido and liquid staking shares
• KuCoin Blog (secondary): entry queue history and May 2026 APR
• Crypto Briefing: churn limit and exit queue history
• Grayscale (SEC filing): staking distribution in Ethereum products
• Astraea Law: US regulatory path for ETH staking
• Luganodes (secondary): SEC/CFTC interpretation and BlackRock ETHB
• DefiLlama (CoinGecko id ethereum): live ETH price, market cap and supply, October 8, 2026
• DefiLlama: Lido TVL, October 8, 2026
• DefiLlama: Ethereum chain TVL, October 8, 2026