ETH Supply in 2026: Scarce by Default, Concentrated by Design

The question that dominated Ethereum through most of 2026 was whether the network still produced a scarce asset. By the final week of October, that argument was effectively over. Ether trades at roughly $2,247, about 55% below its August 2025 record of $4,946, and yet almost none of it is available to sell. The scarcity is real. What has quietly replaced the scarcity debate is a harder question about control.

More than 43.9 million ETH is locked in staking. Public companies hold close to 7.9 million on their balance sheets. US spot exchange-traded funds hold billions of dollars more. And the ether sitting on exchanges, the slice that anyone can actually trade on short notice, has fallen to a multi-year low. Supply that leaves an exchange wallet for a validator deposit or a treasury vault does not come back quickly, and most of it is not coming back at all.

So the interesting question for the next twelve months is not how much ETH exists. It is who holds the part that is locked away, and what it means that the list of holders keeps getting shorter. One treasury company is days from owning 5% of every ether in existence. One staking protocol routes about a fifth of all staked coins. One exchange custodies most of the ether inside American ETFs. Ethereum solved its scarcity problem and walked straight into a concentration problem, and the two are the same story told from opposite ends.

The supply picture in one table

Before the question of who controls ETH, it helps to fix how much there is and where it sits. The numbers below are the late-October 2026 snapshot that every argument in this piece rests on. Two things stand out: issuance is low but no longer negative, and the tradable float has shrunk far faster than the headline supply figure suggests.

The supply line itself is almost boring. Ethereum’s monetary policy has been stuck in neutral all year. According to ultrasound.money, the canonical tracker, total supply sits at about 122.18 million ETH, and the annualized net issuance rate hovers near zero on any given day while trending slightly positive over longer windows. TokenPost puts the staked total at 43.94 million ETH, about 35.2% of supply, with gross yields down near 2.47%. Those are not the numbers of an asset that is being inflated away. They are the numbers of an asset that is being absorbed.

Scarcity is settled; concentration is the new story

For two years the fight over ETH supply was a fight over protocol policy. The “ultrasound money” thesis held that the EIP-1559 fee burn plus the Merge’s issuance cut would make ETH deflationary. Then the Dencun upgrade in March 2024 pushed most activity onto layer-2 rollups, the mainnet fee burn collapsed from thousands of ether a day to a few dozen, and the network tipped back into mild inflation. That broke the automatic-scarcity narrative and set off a governance war over whether to restore it by hand.

That war ended in October. EIP-8363, the “tapered issuance burn” proposal that would have driven consensus-layer issuance to zero as staking climbed past half the supply, was withdrawn from the Hegota upgrade on 1 October. Its co-author, Jerome de Tychey, argued the idea was too consequential to bundle into a routine hard fork and deserved its own dedicated process, now pushed into 2027. The practical result is that Ethereum’s issuance schedule is frozen at “slightly inflationary,” and the one lever that could have re-engineered scarcity has been set aside. The market, not the protocol, now decides how scarce ETH feels.

Which is exactly why the story moves to concentration. If scarcity is a market outcome rather than a protocol guarantee, then it depends entirely on the behavior of the handful of entities doing the absorbing. A thin float held by a few large, patient holders looks scarce right up until one of them changes its mind. Understanding who those holders are, and how much latitude each of them has, is now the whole game.

How the float drained: a two-year story

The vanishing float did not happen overnight. It is the sum of three waves that each pulled ether out of circulation for different reasons, and they stacked on top of one another through 2025 and 2026.

The first wave was staking. When the Merge switched Ethereum to proof of stake in September 2022, and the Shapella upgrade enabled withdrawals the following April, staking stopped being a one-way door. Deposits climbed anyway, because the yield was real and the lock-up risk was gone. The staked total crossed a third of all supply in 2026 and has hovered there since, around 35% to 36%, a level that would have looked implausible when validators could not yet withdraw.

The second wave was the corporate treasury. Following the template that MicroStrategy wrote for Bitcoin, a cluster of public companies decided ether belonged on the balance sheet. The difference is that staked ETH pays a yield, so an ether treasury is a productive one. By late October, CoinGecko tracked 34 public companies holding roughly 7.9 million ETH, about 6.49% of supply, worth close to $18.9 billion. The third wave was the exchange-traded fund. US spot ETH ETFs launched in 2024, and once a 2026 joint interpretation from the SEC and the Commodity Futures Trading Commission clarified that staking rewards are not securities, several funds began staking the ether they held, turning a passive vehicle into another source of permanent lock-up.

The arithmetic of those three waves shows up most clearly on exchanges. Ether reserves across all venues have fallen to about 14.7 million ETH, a multi-year low, according to CryptoQuant data reported by Finbold, down from a 2021 peak above 35 million. Binance alone shed around 430,000 ETH between late August and mid-October. The headline supply barely moved, but the part of it that can be sold without warning shrank by more than half over the cycle.

One caution on the arithmetic: these buckets overlap, so they cannot simply be added. BitMine’s ether counts as a treasury holding and, because most of it is staked, as part of the staking total too. The same coin can sit inside an ETF and be staked by the fund that holds it. Stacking staking, treasuries, and ETF balances on top of one another would double-count heavily and overstate how much supply is truly locked. The honest way to read these figures is as overlapping lenses on one shrinking float, not as separate piles to be summed.

BitMine and the 5% ceiling

No single entity embodies the concentration story like BitMine Immersion Technologies. The company, chaired by Fundstrat’s Tom Lee, has bought ether in each of the past 65 consecutive weeks. Its 19 October filing with the SEC disclosed a position of 6,041,481 ETH, about 4.9% of circulating supply, as reported by The Block. That makes BitMine the largest single-entity holder of ether in the world outside the Ethereum Foundation itself.

What makes the position remarkable is not just its size but its declared ceiling. Speaking at TOKEN2049 in Singapore in early October, Lee said BitMine would stop buying once it reached 5% of supply, roughly 6.105 million ETH. “We only need to get another 100,000 ETH to get to 5%,” he told CoinDesk. “Now we’re going to stop.” He added, with evident surprise at the pace: “We thought this would take five years. It took us a little over a year.” By late October the gap to the cap had narrowed to under 60,000 ETH, a few more weekly buys away.

The cap is a deliberate signal. By naming 5% as a hard stop, Lee is trying to reassure the market that BitMine is an accumulator, not an acquirer trying to corner the asset. But the signal cuts both ways. A buyer that has absorbed roughly 5% of all ether in a little over a year, and that is about to switch off, is itself a supply event. Much of the ETH that cleared exchange shelves in 2026 did so because BitMine was buying. The scarcity that holders are celebrating was partly manufactured by one balance sheet, and that balance sheet has just announced it is nearly full.

A treasury with a validator’s vote

BitMine’s 5% cap is measured against total supply, but there is a second number that matters more for how Ethereum actually runs. The company stakes the large majority of its ether. Of its roughly six million ETH, well over five million is staked through its MAVAN platform and partner operators. Measured against the 43.9 million ETH staked network-wide, that single treasury now backs more than a tenth of all staked ether, worth well over ten billion dollars.

In a proof-of-stake system, staked ether is voting weight. Validators attest to blocks and occasionally propose them, and their influence scales with the stake behind them. The classic danger zone is one third of all stake, the point at which an actor can stall finality, and one half, the point at which it can do far worse. BitMine is nowhere near those thresholds on its own. But concentration does not have to reach a mathematical attack threshold to matter. It changes who the network has to trust, and it narrows the number of parties whose operational failure, legal exposure, or change of strategy could ripple through consensus.

There is a mitigating detail worth stating plainly: BitMine does not run all of that stake as one validator. MAVAN spreads the ether across multiple professional operators, so the economic ownership is concentrated even though the validator operation is distributed. That distinction softens the technical risk without erasing the governance one. A single owner directing a tenth of the stake still has a single owner’s incentives, and those incentives answer to a public company’s shareholders and regulators, not to Ethereum’s rough consensus.

Lido, stETH, and the one-third line

If BitMine is the loudest concentration story, Lido is the oldest one. The liquid-staking protocol lets holders stake any amount, receive a tradable stETH token in return, and avoid running hardware. That convenience made it dominant. As of late October, Lido held about 9.47 million ETH, roughly 21.5% of all staked ether, the single largest share of any provider, per data compiled by CoinCentral. Its stETH accounts for around 62% of the entire liquid-staking market.

The good news is that Lido’s share is falling, not rising. It has slipped from a 2023 peak near 32%, and from about 23.9% at the start of 2026 to roughly 21% by autumn, as institutional stakers, treasuries, and exchange-run validators took a larger slice of net growth. The community has long debated a self-imposed limit at one third of all stake, the reasoning being that no single staking entity should ever be able to threaten finality. Lido has never formally adopted that cap, but its drift back toward a fifth of stake has taken some of the urgency out of the argument.

Even a shrinking Lido is a governance question, because stETH is not just a staking receipt; it is collateral across much of decentralized finance. A protocol that controls a fifth of stake and whose token underpins a large chunk of on-chain lending is a systemic node by any definition. Lido has responded by hardwiring decentralization into its newer modules, which is the subject of a later section, but the basic shape holds: Ethereum’s staking layer is led by one protocol, one treasury, and a short list of exchanges, and the gaps between them are the network’s real margin of safety.

The custody chokepoint nobody voted for

Staking concentration is visible on-chain, so it gets debated loudly. The custody chokepoint is quieter and arguably more acute. Of the nine US spot Ethereum ETFs, eight use Coinbase as their custodian, which means a single company safeguards roughly 80% of all the ether held inside American funds. Gabor Gurbacs, an adviser to VanEck, has publicly flagged the concentration, noting that so many competing products relying on one custodian creates a risk that Coinbase might not resource its operations well enough to support all of them.

A handful of issuers took the warning seriously. VanEck chose Gemini rather than Coinbase, and Fidelity’s fund is widely cited as the only large US spot ETH product with zero Coinbase exposure at both the custody and validation layers. But those are the exceptions. The default architecture of the ETF wrapper, the vehicle that was supposed to make ETH safe for conservative money, routes the overwhelming majority of that ether into one set of cold wallets.

This is the same category of risk that has bitten crypto repeatedly, where the failure is not a cracked key or a bug in a contract but an operational or process breakdown at a single trusted party. The year’s largest exchange loss, examined in our report on the $387 million Bitget hack, turned on exactly that kind of weak link rather than broken cryptography. Concentrating ETF custody in one provider does not create a hack, but it does create a single surface that an attacker, a court order, or an internal mistake could all push against at once.

There is a subtler point for fund investors specifically. An ETF holder does not hold ether; they hold a share in a trust that holds ether through a custodian. That is three layers of counterparty between the investor and the asset, and all three narrow toward the same short list of names. For a product sold on the safety of a regulated wrapper, the ultimate security still comes down to how well one custodian guards a set of private keys, the same problem a self-custodial holder faces, only moved out of sight behind a ticker symbol.

Who builds Ethereum’s blocks

There is a third chokepoint that most holders never see: the production of blocks themselves. Under the MEV-Boost market that dominates block building, validators outsource the job of assembling a block to specialized builders who compete to extract the most value from transaction ordering. That market is efficient, and it is also concentrated. A small number of builders assemble the large majority of Ethereum blocks, and the relays that connect them to validators have at times filtered transactions to comply with sanctions lists, which is the polite phrase for censorship.

This is not a fringe worry. Vitalik Buterin put it at the center of his roadmap. In his 2024 “Scourge” essay, he called proof-of-stake centralization “one of the biggest risks to the Ethereum L1,” and identified two places where it bites: block construction and staking capital provision. He noted that two entities were building roughly 88% of Ethereum blocks at the time, and warned that left unchecked, worsening censorship could force users to wait up to about two minutes for a transaction to be included rather than a few seconds. The mechanics of how value gets extracted from that process, and how it changed after the year’s biggest exchange hack, are the subject of our piece on MEV strategies in 2026.

Put the three vectors side by side and the shape of the problem is clear. Scarcity concentrates ownership, ownership concentrates stake, and stake concentrates block production, with custody sitting across the top of all of it. None of these is near a catastrophic threshold on its own. The concern is that they rhyme.

The table is a snapshot, not a verdict. Shares move, and several of these entities are actively trying to decentralize the very positions they occupy. But a reader scanning that column of percentages should notice how few names it takes to describe where most of Ethereum’s locked value lives.

The decentralization toolkit: DVT, FOCIL, solo stakers

Ethereum’s developers are not blind to any of this, and the counter-measures are further along than the headlines suggest. The most mature is distributed validator technology, or DVT, which splits a single validator’s duties across several independent operators so that no one machine or company holds the keys. Two networks, Obol and SSV, provide the plumbing. Lido has leaned into it hardest through its Simple DVT module, which by 2026 spread a low but steadily growing single-digit share of all staked ether across clusters of independent operators rather than a single company. Lido also shipped its V3 architecture, with configurable staking vaults, at the start of the year.

The second tool targets censorship directly. Fork-choice-enforced inclusion lists, or FOCIL, would let a committee of validators force transactions into blocks even if the dominant builders would rather leave them out, taking the censorship lever out of a few hands. FOCIL was slated for the Hegota upgrade, now expected in 2027, after being shuffled out of the nearer-term roadmap. A third idea, a two-tier “risk-bearing” and “risk-free” staking model that Buterin has championed, would make it easier for small and solo stakers to participate without running the gauntlet of slashing risk, pulling stake back toward the edges of the network.

Underneath all of it sits the solo staker, the home validator running a single node for the base reward. That cohort is the ground truth of decentralization, and it is under pressure from exactly the economics this article describes: a three-year-low yield, rising hardware and operational demands, and the pull of professional pools. We traced that squeeze in detail in our look at Ethereum solo staking in 2026. The toolkit is real, but most of it ships later, and in the meantime concentration compounds.

Glamsterdam slipped, and the burn stayed broken

The next upgrade on the calendar, Glamsterdam, is sometimes misread as a chance to revive ETH scarcity. It is not. Glamsterdam is a scaling fork. Its marquee features are enshrined proposer-builder separation (ePBS), which bakes the builder relationship into the protocol and reduces reliance on today’s external MEV-Boost relays, block-level access lists for parallel validation, and a set of gas-repricing changes aimed at letting the network raise its gas target toward 200 million, more than triple the current mainnet limit.

More blockspace is good for users and good for the decentralization fight, because ePBS is a direct answer to the block-building concentration Buterin flagged. But it works against the burn. Ethereum’s fee mechanism targets blocks that are half full, so adding capacity lowers the base fee, and a lower base fee means less ether burned. The scaling roadmap and the deflation narrative point in opposite directions, which is why net issuance is likely to stay modestly positive even after Glamsterdam ships.

And shipping is the operative uncertainty. Glamsterdam activated on the Sepolia testnet on 6 October, but a planned Hoodi testnet activation on 26 October was postponed after the security firm ChainSecurity flagged how one of the gas-repricing changes, EIP-8037, could interact with existing mainnet contracts. Developers floated 1 December and 8 December as candidate mainnet dates, conditional on a clean Hoodi run, but nothing is confirmed. The takeaway for supply is simple: the one upgrade that could have changed Ethereum’s fee dynamics this year is late, and even on time it would loosen scarcity, not tighten it. With EIP-8363 shelved and Glamsterdam delayed, the protocol is doing nothing to the float. Everything that happens to ETH scarcity from here happens on the demand side.

Scarcity by decree versus scarcity by protocol

The withdrawal of EIP-8363 sharpened a contrast that ETH investors should sit with. Bitcoin’s scarcity is written into the protocol: a 21 million hard cap, a halving schedule, no discretion, no committee that can change the issuance curve without effectively forking into a different asset. Ethereum’s scarcity, after a year of debate, is the opposite. There is no cap, issuance is governed by the choices of core developers and the social layer, and the fact that it is currently “slightly inflationary” is a decision that was actively defended rather than a law of nature. The one mechanism that would have hard-coded tighter issuance was pulled, and it may return through a dedicated process in 2027, or it may not.

Call it scarcity by decree. It is not worse than scarcity by protocol in every respect; a discretionary policy can respond to circumstances a fixed one cannot. But it means ETH’s monetary story is only as credible as the governance that backs it, and that governance is the same concentrated set of stakeholders this article has been describing. The irony is neat: Bitcoin fixes issuance but concentrates its security in a few mining pools, a dynamic we covered when Binance Pool shut down; Ethereum fixes nothing about issuance but concentrates its ownership and its stake. Both networks face a concentration problem. They just locate it on different floors of the building.

For valuation, the practical consequence is that ETH behaves less like digital gold and more like a claim on how much the network actually gets used. When fees fall, as they have since activity moved to layer 2, the burn falls with them unless usage rises to compensate. Scarcity, on that view, is necessary but not sufficient. A scarce asset that nobody needs to use is just illiquid, and illiquidity is not the same thing as value.

A Fed on hold, a float still locked

The macro backdrop in late October did the concentration thesis no favors. On 28 October the Federal Reserve held its target range at 3.75% to 4.00% in a 10-2 vote, with Governor Lisa Cook and Cleveland Fed President Beth Hammack dissenting in favor of a hike. Chair Kevin Warsh framed the hold as an observational pause to judge how September’s increase was feeding through, reiterated that 2% is a hard target, declined to guide toward December, and insisted that politics played no role so close to the midterm elections. For a crypto holder, the signal is a risk-free rate parked near 4% with the door open to going higher.

That matters because the entire locked-float story rests on a behavioral assumption. Staking yields around 2.47% gross. A Treasury bill pays more than that with no price risk. On paper, every staker is accepting a negative spread versus cash for the privilege of holding ether. They do it because they want price exposure and because the structures holding the ether are not yield-sensitive in the normal sense: a treasury company is not going to unwind 6 million ETH because T-bills look attractive this quarter, and an ETF holder is making a directional bet, not a carry trade. The float stays locked because of who holds it, not because the math favors staking.

None of these structures would exist without a regulatory green light. The 2026 SEC and CFTC interpretation that staking rewards are not securities is what let funds stake and let treasuries market themselves as yield-bearing, a shift we put in context alongside the agency’s broader posture in our coverage of SEC crypto enforcement in 2026. The same clarity that unlocked demand also deepened concentration, by channeling institutional money through a small number of compliant custodians and issuers. Policy giveth the float, and policy shapeth who holds it.

Three scenarios and five dials for 2027

If the protocol is out of the scarcity business and the demand side now sets the price, the useful thing is not a point forecast but a map of how the concentrated holders behave. A thin float is a two-way amplifier: the same structure that turned modest buying into a sharp 2026 rally can turn modest selling into a sharp drawdown. Three broad paths cover the likely range into 2027.

The base case is the dull middle: the lock holds, nobody panics, and ETH trades on macro and flows rather than any supply drama. The tails are where concentration earns its keep as a risk factor. The squeeze is the bull story everyone on crypto social media is telling, and it is plausible. The unwind is the story nobody is pricing, and it is the one that follows directly from everything above, because a market whose scarcity depends on a short list of holders is only as stable as the most fragile name on that list.

Rather than guess which path wins, watch the dials that decide it. Five of them tell you most of what you need to know:

• BitMine’s cap. Does it hit 5% and genuinely stop, hold the position, and let the market find a new marginal buyer, or does the pace of accumulation mask a later change of heart?
• Lido’s share. Does it keep drifting down toward a fifth of stake, or climb back toward the one-third line that reopens the finality debate?
• Custody spread. Do more issuers follow VanEck and Fidelity away from a single custodian, or does Coinbase’s roughly 80% grip on ETF ether hold?
• Decentralization tooling. Do DVT adoption, FOCIL, and friendlier solo-staking economics actually ship and scale, or do they keep slipping to the next upgrade?
• Exchange reserves. Do they keep falling, confirming the lock, or rebuild, signaling that the patient holders are quietly becoming sellers?

Ethereum spent 2026 winning the argument it set out to win. Supply is tight, the inflationary scare is contained, and the asset is scarcer in practice than its issuance schedule implies. The prize for winning that argument is a new one. Scarcity delivered by a handful of treasuries, one dominant staking protocol, one majority custodian, and a few block builders is scarcity with a counterparty, and counterparties have interests, regulators, and exit doors. For the next year, the number that matters is not how many ether exist. It is how many hands they are in.

Frequently Asked Questions

How much ETH is staked in 2026?

As of late October 2026, about 43.9 million ETH is staked, roughly 35% of total supply, according to network trackers. Gross staking yields have fallen to around 2.47%, a three-year low, as more ether competes for the same validator rewards.

Who owns the most Ethereum?

Outside the Ethereum Foundation, the largest single holder is the treasury company BitMine, with roughly 6 million ETH, about 4.9% of supply. In total, 34 public companies hold close to 7.9 million ETH on their balance sheets, about 6.49% of all ether.

Is Ethereum deflationary in 2026?

No. Ethereum turned mildly inflationary after the March 2024 Dencun upgrade pushed most activity onto layer-2 rollups and collapsed the mainnet fee burn. Net issuance now runs modestly positive, roughly 0.2% to 0.8% a year.

Why are Ethereum exchange reserves falling?

Staking, corporate treasuries, and exchange-traded funds are absorbing ether faster than it flows back to exchanges. That has pushed reserves across all venues down to about 14.7 million ETH, a multi-year low, shrinking the supply available to trade.

What is EIP-8363 and did it pass?

EIP-8363 was a proposal to burn a rising share of validator rewards, driving ETH issuance toward zero as staking climbed past half the supply. It was withdrawn from the Hegota upgrade in October 2026 and pushed into a separate issuance-policy process, so it has not taken effect.

Priya Reddy is a markets correspondent at HOGE Wire, covering Ethereum supply, staking, and on-chain market structure.