Why
Today's coverage joins three supply numbers into one sentence: ETH is queuing to be staked, exchange balances are falling, OTC desks are nearly empty, therefore a supply crunch, therefore price. Each number is real. The inference is not, because each of the three measures supply and none of them measures demand. A shelf with fewer coins on it makes any buyer move the price more, and it makes the absence of a buyer let any seller move it more. The queue adds a second error: it is read as a one-way valve, ETH going in and staying in, when it is a rate limit, and a rate limit is symmetric. The constant that makes new stake slow to arrive is the constant that makes staked ETH slow to leave. This is the galaxy-brain test from Tech #125 applied to a chart: which of these numbers would look the same if the conclusion were the opposite. All three would.
How it works
One constant, two queues
Since Pectra (EIP-7251) the churn limit is counted in ETH rather than in validators. Per epoch the chain can activate, and separately can exit, at most min(256 ETH, max(128 ETH, total active balance / 65,536)). With about 41 million ETH staked the uncapped term is roughly 630 ETH, so the 256 ETH ceiling is the binding number. An epoch is 6.4 minutes, 225 epochs a day, so 57,600 ETH a day in each direction; 2.48 million divided by 57,600 is 43 days, which is the wait the trackers print. Before Pectra the same ceiling existed in a different unit: EIP-7514 capped activations at eight validators per epoch, eight times 32 ETH, the same 256.
The exit side shares the limit and adds two delays the entry side does not have. After a validator's exit epoch it waits a fixed 256 epochs, about 27 hours, before it is withdrawable, and then the withdrawal sweep has to reach it, which at 900,000 validators takes about eight days (ValidatorQueue shows 7.9). A full round trip for one unit of stake, entered today, is therefore roughly 43 days in, plus whatever the exit queue holds when it wants out, plus nine days.
The ratio is pressure at the gate, not flow through it
Entry demand at 13.6 times exit demand does not mean 13.6 times as much ETH is being staked as unstaked. Both sides are capped at the same 256 ETH per epoch, so when both queues are non-empty the flows are equal and the ratio only says which waiting room is longer. It is the sign of the imbalance and a measure of how long it will persist, not a volume. The same gate ran the other way in September 2025, when the exit queue reached 2.67 million ETH (CCN). One recording a week of entry queue, exit queue and their ratio is enough; the week the ratio crosses one is the candidate for a regime change, and it will be visible weeks before it clears, because clearing takes weeks by construction. Direction is a demand question and comes from the ETF flow print (Tuesday morning for Monday), not from the queue.
Supply says how much, demand says whether
Exchange balances, Binance's reserve and the OTC desk figure are all the same kind of number: coins sitting somewhere they can be sold quickly. They are stocks, not flows, and a smaller stock is an amplifier. The August Investing.com piece makes the argument in its correct form: float down from staking and exchange withdrawals, plus ETF inflows, mechanically amplifies moves. That sentence has a demand term. Today's posts keep the first half and drop the second, which turns an amplifier into a forecast. The OTC number cuts the same way: if large buyers must now fill in the public market because desks have no inventory, the same money prints more on the tape, and so does the same selling.
What the queue costs a holder
A validator entering now earns nothing for about 43 days and about 2.6 percent a year after that (ValidatorQueue's APR at time of writing). Leaving costs the exit queue, today about two days, plus the 27 hours and the sweep. Liquid staking tokens do not remove the wait, they move it: the protocol's withdrawal queue faces the same 256 ETH constant, so a secondary-market discount on an LST is the price of skipping the queue, and it widens exactly when the exit queue is long.
Where it lands in Jayverse
- Invest: one weekly row. Entry queue, exit queue, their ratio and the week's ETF net flow, side by side. The number to watch is the crossing, not the level, and the flow column is the only one that speaks to direction.
- Auditor: "a supply metric is not a demand claim". When a briefing moves from "sellable coins fell" to "price rises", the rule asks for the demand term. Same shape as "an empty result is not a negative answer" (Tech #131): the inference is missing a premise, and the check is to name which one.
- Wallet (#5) and Rabbit: two dates before signing. If a smart account ever stakes, the confirmation screen states the activation wait and the current exit wait, computed from the live queue, because the constant is public and the wait is the product the user is buying.
- Devnet and Chains: not simulable locally. An Anvil fork has no beacon chain, so queue behaviour can only be observed on Sepolia or mainnet; the Chains page should say so in the capability matrix.
Verified and unverified
Verified on 2026-09-21: the churn mechanism (EIP-7251 in Pectra sets the per-epoch activation and exit churn in ETH with a 256 ETH ceiling; EIP-7514 had capped activations at eight validators per epoch before it), 225 epochs a day, the 256-epoch withdrawability delay, and the sweep. ValidatorQueue at time of writing showed 1,753,909 ETH in the entry queue with a 30-day wait, 131,040 ETH exiting with a two-day wait, 906,351 active validators, 43.2 million ETH staked (35.4 percent) and churn 256 per epoch, which is a smaller queue and a larger stake than the report's figures from one day earlier; either the queue drained unusually fast or the two sources count differently, and this item keeps the report's numbers as the day's headline and flags the gap rather than reconciling it. Taken from the report and its sources and not independently checked: 2.48 million ETH, 13.6 times, 43 to 45 days, 41 million and 33.5 to 34 percent, 15.65 million on exchanges, the Binance three-month low, 123,000 BTC on OTC desks (CryptoQuant, analyst Darkfost), and the 2.67 million ETH exit queue of September 2025. Not seen: any source on staking concentration or Lido's share; today's numbers are totals only. Sources: CCN "Ethereum staking demand explodes" (2026-09-20), Blockonomi "Ethereum staking demand outpaces exits", Investing.com "Ethereum supply tightens" (Itai Smidt, 2026-08-26), PANews on OTC reserves, validatorqueue.com; related items: Tech #16 (stake concentration), #7 (DVT), #125 (galaxy-brain test), #130 (Ethlabs week 13), #131 (empty result), Invest 813 (which ETF window).
Key expressions
| Expression | 뜻 · 쓰이는 자리 |
|---|---|
| rate limit | 레이트 리밋(단위 시간당 처리량의 상한) · 큐의 본질을 한 단어로. "a rate limit that runs both ways" |
| churn limit | 천 리밋(에폭당 활성화·이탈할 수 있는 검증자 잔고의 상한) · 이더리움 합의층 용어. "the churn limit is counted in ETH" |
| entry queue / exit queue | 진입 큐 / 이탈 큐 · 스테이킹 대기열 양쪽. "which waiting room is longer" |
| epoch | 에폭(32슬롯, 6.4분; 검증자 상태가 바뀌는 단위) · 처리량의 분모. "225 epochs a day" |
| binding (constraint) | 실제로 걸리는 (제약) · 여러 상한 중 지금 작동하는 것. "the 256 ETH ceiling is the binding number" |
| withdrawal sweep | 인출 스윕(검증자를 순서대로 돌며 잔고를 내보내는 절차) · 이탈 뒤의 마지막 지연. "the withdrawal sweep has to reach it" |
| round trip | 왕복 · 들어갔다 나오는 총 시간. "A full round trip for one unit of stake" |
| pressure at the gate | 문 앞의 압력 · 비율이 재는 것. "pressure at the gate, not flow through it" |
| regime change | 국면 전환 · 시장의 성격이 바뀌는 시점. "the candidate for a regime change" |
| by construction | 구조상, 만들어진 방식 때문에 · 결과가 설계에서 따라 나올 때. "clearing takes weeks by construction" |
| stock vs. flow | 재고 대 흐름(어느 시점의 양 대 기간 동안의 양) · 경제 지표를 읽는 기본 구분. "They are stocks, not flows" |
| liquid float | 유동 물량(즉시 팔 수 있는 코인의 양) · 공급 지표의 이름. "float down from staking and exchange withdrawals" |
| amplifier | 증폭기 · 공급 감소가 하는 일. "turns an amplifier into a forecast" |
| print (on the tape) | (체결 기록에) 찍히다 · 거래가 시세에 반영되는 것. "the same money prints more on the tape" |
| OTC desk | 장외 거래 데스크(Over-The-Counter; 공개 호가창 밖에서 큰 물량을 중개하는 곳) · 큰손의 매매 경로. "known OTC desk addresses" |
| LST | 유동화 스테이킹 토큰(Liquid Staking Token; 스테이킹된 ETH를 대신하는 거래 가능한 토큰) · 큐를 옮기는 상품. "a secondary-market discount on an LST" |
| APR | 연 이율(Annual Percentage Rate; 복리 없이 계산한 연 수익률) · 스테이킹 보상 표기. "about 2.6 percent a year" |
| ETF flow | ETF 자금 흐름(Exchange-Traded Fund의 순유입·유출) · 수요 쪽 지표. "comes from the ETF flow print" |
| demand term | 수요 항(추론에 있어야 할 수요 쪽 요소) · 빠진 전제를 가리킬 때. "That sentence has a demand term" |
| capability matrix | 기능 표(체인별로 되는 것과 안 되는 것의 표) · Chains 페이지. "say so in the capability matrix" |