Weekly Intelligence — Week of September 21–27, 2026

1. Staking health

Answer: participation rose again, the entry queue kept draining, and the yield the network pays stepped down to its lowest print in our records.

Staked ETH reached 43,681,037 on the Sep 28 Ebunker weekly — about 36.19% of a 120.70M implied circulating denominator, +0.42pp against the prior print of 35.77%, with the Sep 21 print (35.91%) between them on the same denominator family. The series is now four prints long on comparable terms: 35.55 → 35.77 → 35.91 → 36.19. The entry queue's wait fell for a fourth consecutive weekly print: 34.6 → 31.0 → 30.3 → 27.7 days (Ebunker series). The exit wait, which jumped from under an hour to 2.4 days at the Sep 21 print, held near 2.8 days.

The week's hard number is the yield: the network staking APR printed 2.18%, down 39bp from three stable prints at 2.57–2.62%, with the stETH 7-day average down 8bp to 2.52%. Against a 4.00% policy rate that is a 182bp shortfall (DERIVED), against the 4.1% end-2026 SEP median 192bp. The validator count could not be refreshed this week — both aggregator endpoints returned access errors — so 906,000 stands as a carried read, and the number that tests CMv2's central projection is temporarily frozen.

Source: Ebunker Ethereum staking weeklies (Sep 21 and Sep 28, 2026); validator count per validatorqueue.com via Tremplin (Sep 20, carried); FOMC decision and SEP (Sep 16, 2026).

2. stETH: supply and peg

Answer: the peg spent its first full week at or below par in our daily records, and closed 5bp down; the supply print is now the oldest dated number on the page.

The daily same-source path, all on this snapshot's re-read basis: 0.99865 (Sep 20, −13.5bp) → 0.99866 (Sep 22, the week's low print) → a brief 1.00020 (Sep 25, +2bp, the week's only premium) → 0.99947 (Sep 27, −5.3bp). The discount never widened past 13.4bp even on the week's one down session, and the Sep 25 touch of par shows the redemption path still pulls the ratio back. What is new is persistence: previous discounts this quarter lasted a session or two; this one held all week. We print that as what it is — a small, stable exit-door price in a rising market — and we hold the pre-committed line: a discount that widens past the round numbers while redemptions stay open would change the story; 13bp does not.

Supply remains 9,659,594 stETH, dated September 9 and now twenty days old — the oldest print we publish. Against a $26.47B TVL the implied supply is 9,776,114, a 1.2% gap that narrowed this week for the mechanical reason the data report gives: the price leg rose under a frozen supply figure. Net issuance of about 2,830 stETH a day covers barely a fifth of the gap over eighteen days. We are not going to paper over a stale print with an aggregator's number we cannot date.

Source: stETH/ETH ratio from the DefiLlama coingecko feed, same-session prints; supply per CryptoCompare via TickerReport (Sep 9); TVL from the DefiLlama protocol history (Sep 27 point).

3. DeFi / LSDfi

Answer: the board is fully fresh for the second week running — and it got smaller in ETH terms while getting bigger in dollars.

Ethereum's chain TVL stands at $53.54B (DefiLlama, Sep 27), up 1.8% from the $52.58B Sep 20 board. ETH rose 5.1% in the same window, so the chain's TVL in ETH terms shrank slightly — the first time this column has been able to make that statement from two fresh same-series reads rather than a carried print. Inside the chain, Lido is $26.47B (+2.9%), Aave $19.24B (+1.3%), Morpho $11.04B (+2.8%); Lido outgrew the chain it lives on. The two carried rows from the Sep 7 board finally refreshed: EigenCloud $7.06B, Sky $6.07B. The one flow reading that is not price-denominated is the queue: an entry wait that has fallen four prints straight says deposit demand is real and, at the margin, easing — not because ETH is leaving, but because the backlog is being processed faster than it is refilled.

Source: DefiLlama protocol-history reads (api.llama.fi/protocol/<slug> and /v2/historicalChainTvl/Ethereum), same session; Ebunker weeklies for the queue series.

4. Competition

Answer: Lido extended its lead in dollars, started the visible part of its CMv2 migration, and spent a governance week narrowing its own discretion.

Lido is 2.6× Binance Staked ETH and 1.4× Aave (both DERIVED); the four liquid-staking protocols hold $43.11B between them, 85% of it in the top two. Lido's share of all staked ETH reads 22.1% (DERIVED: 9,659,594 ÷ 43.681M), on inputs eighteen days apart — if the frozen supply print is understating reality, the true share is higher, not lower. The week's structural events: the first CMv2 consolidation gained a detailed second account (Sep 21), Vote #205 put execution delegation and the deposit-reserve framework on-chain (Sep 21), and the committee received its bounded dial over the reserve (Sep 25, ceiling 9,600 ETH). The 0x02 CSM launch — October per the committee, Q4 per the documentation — is the next dated thing on the calendar.

Source: DefiLlama protocol-history reads, Sep 27; CMv2 account per Crypto Briefing (Sep 21); governance per Bitcoinist and CryptoSlate; committee statement (Sep 2) via the same coverage.

5. Outlook — and edition five's four calls, graded

Edition five published four watchlist items. All four can be settled this week:

Watchlist for the September 28 – October 4 edition, four items: (1) the next Ebunker APR print — one more read at or near 2.18% turns a step into a regime, and the corridor's carried rows would follow; (2) the 0x02 CSM launch — October per the committee, Q4 per the docs, and the deposit-reserve decision leans on it; (3) the reserve motion — whether the committee actually zeroes the 1,500 ETH target, and how fast; (4) the validator count — the first obtainable fresh read after this week's access failure, because CMv2's one-third reduction is a projection until the count moves.

Source: scorecard inputs as cited in sections 1–4 and the data report.

6. The 39bp step: what the record supports

Because the yield drop is the week's hard number, the intelligence column does the arithmetic the headline leaves out. The series: 2.62, 2.57, 2.57 and 2.18 across the last four weekly prints — three stable prints, then one step. What stake growth explains: issuance is effectively fixed per epoch, so per-stake yield falls roughly in proportion to the stake; active stake rose 0.77% between the last two prints, which mechanically dilutes the yield by about 2bp (DERIVED: 2.57 ÷ 1.0077). The observed step is twenty times that. What could account for the rest — a fall in execution income (tips and MEV are part of the reward mix and can move independently of stake), a methodology change at the source, or some of each — cannot be separated from outside a single provider's weekly. What the same print's own second series says: the stETH 7-day average fell only 8bp, to 2.52%. One provider, two series, two different step sizes, no second source. The deep dive takes this apart in full; the falsifiable marker is simply the next print.

Source: Ebunker weeklies (the last four editions); dilution arithmetic our own, on the weekly staked totals.

7. Governance week, mechanically

The two-step governance move is worth laying out because its parts interlock. Vote #205 (Sep 21) created the framework: execution delegation (LIP-37), an upgraded DepositSecurityModule, and a deposit-reserve target whose committee-set value is capped at 9,600 ETH. Four days later (Sep 25) the Curated Module Committee received the Easy Track authority to set that value, against a current setting of 1,500 ETH and a published plan to take it to zero until the 0x02 CSM launches. The reserve exists inside Lido's buffer, which serves withdrawal finalizations first when the two uses compete; the committee's own sizing study prices the trade at 1.6 days of stressed-case withdrawal speed. As of Sep 27 no motion had been filed. The design intent reads clearly in the ceiling: the DAO has bounded the committee's discretion at a level below one day of churn-limit throughput, which is what giving a five-of-nine multisig a real dial is supposed to look like.

Source: Bitcoinist and CoinMarketCap research coverage of Vote #205 (Sep 21); CryptoSlate and WhaleFactor on the Easy Track activation (Sep 25), citing the committee's Sep 2 statement.

8. The flows that came back

The ETF complex took in $689.8M across five sessions (DERIVED from the Farside daily prints: +270.0, +162.2, +104.5, +66.1, +87.0), an $830.4M swing from the prior week. Two things are worth saying precisely. First, the biggest day (Sep 21, +$270.0M) was the window's first day — the reversal was immediate, not gradual. Second, the inflows arrived while the network's yield stepped down 39bp, which means the marginal ETF buyer is not underwriting the staking return at all. Add the treasury companies — Bitmine at 5.98M ETH after the Sep 21 disclosure, roughly 13.7% of all staked ETH (DERIVED) — and the week's demand picture is coherent in one way only: the bid is for scarce supply, and the yield is a coupon on it rather than the reason for it.

Source: Farside Investors Ethereum ETF daily table; Bitmine disclosure (Sep 21) via Cryptopolitan and PR Newswire.

9. Where Lido sits after the week

The summary position, all on this snapshot's bases: $26.47B TVL (+2.9%), 22.1% of staked ETH (on inputs eighteen days apart, labeled), 2.6× the nearest liquid-staking rival, and 85% of the LST board's top two. The migration that will define the next two quarters has one executed consolidation, one detailed account of it, and a governance framework around it — and still no published number for how much stake actually moved. The single most useful number for judging the next quarter is not any of those. It is the validator count, currently frozen at a carried 906,000 by an access failure we have disclosed, because CMv2's one-third reduction is the projection the whole migration is measured against. When a fresh print lands, this column will run it against the 8.4M plan.

Source: figures as cited in sections 1–4 and the data report; CMv2 migration scale per Lido via the Ebunker weekly (Sep 28) and Crypto Briefing (Sep 21).

10. The trend this edition extends

Three series this site has tracked on comparable terms for four straight weekly prints. None is an isolated weekly number; each is a line:

The pattern is two stories pulling in opposite directions. Participation keeps climbing — the ratio has risen 0.64pp across four prints, and the entry queue has shortened in every one of them — while the yield just stepped off the plateau it held for three. The intelligence read of the week is that the structural facts (more stake, shorter waits) and the price fact (lower yield) are now diverging for the first time in the comparable record, and the next print decides whether the divergence is a regime or a print. Every number in the table is carried from the data report of this edition and its three priors, all on the same denominator family and the same weekly source.

Source: the four weekly Ebunker prints cited in section 1 and the data report; ratio, APR and queue as tabled there.

FAQ

Related reading

• Lido & stETH Weekly Data Report — Week of September 21–27, 2026
• Weekly Intelligence — Week of September 14–20, 2026
• stETH Yield Update — Week of September 21–27, 2026