Weekly Intelligence — Week of September 7–13, 2026

1. Staking health: is the network absorbing capital or stalling?

Both, on this week's evidence. The stock of staked ETH rose to 43.1M (Sep 12, validatorqueue) from 42.92M a week earlier — +0.4% — with 910,898 active validators, up 0.7% on the Sep 2 count. But the rate of new commitment fell sharply: the entry queue dropped from 2.07M ETH / 36 days to 725,000 ETH / 12.6 days, and the network pays 2.59% (Sep 10) against a carried 3.92% three-month bill — a 133bp gap (DERIVED), the widest we have printed. For a year-old reference point that keeps the panic honest: the exit queue peaked above 2.6M ETH in September 2025, in an actual flight — this week's 2.656M sits at that same level, but with a named, orderly cause.

The queue's own arithmetic sizes the event. 2.656M ETH spread over a 46-day wait implies roughly 57,700 ETH leaving the active set per day at steady flow (DERIVED: 2,656,000 / 46) — about 1,800 validators a day at 32 ETH each, the same daily figure last week's deep dive computed for the entry side when the pipe ran the other way. That symmetry is the point: the machinery that admitted tens of thousands of ETH a day in August can discharge the same in September without any part failing. The number to watch is therefore not the queue's length but whether the daily drain matches this arithmetic — a faster drain means exits are clearing early, a slower one means new exits are joining behind Kiln's.

Source: validatorqueue.com reads (Sep 10 via Fixygen/Coinliva; Sep 12 via BlockBeats); Sep 2025 peak per Coinliva's retrospective.

2. stETH: supply, peg, and what the whale flows actually did

Supply: 9,659,594 stETH (Sep 9, CryptoCompare) — down 2,341 tokens, −0.02%, from the Sep 5 print. In a week when a whale redeemed 10,000 ETH and a provider exited 1.6M ETH of validators, the token's supply moved by less than a rounding error: redemptions through Lido's queue are being offset by new stakes, and Kiln's validator churn does not change the underlying deposit balance at all. Peg: 0.9987 on the Sep 15 same-source pair (stETH $2,514.60 vs ETH $2,517.76) — the first discount print in this column's series, after two weeks at a 1.0003 premium. The magnitude is 13 basis points; we flag the direction change, not a crisis. Lido TVL: $24.35B (Sep 15, DefiLlama direct read), +2.6% on the week, and the supply×price cross-check passes at a 0.2% gap.

The absorption arithmetic makes the calm legible. The week's redemption flow — 10,000 ETH via the tracked whale — is 0.10% of the 9,659,594 stETH supply (DERIVED), and even Kiln's 1.6M ETH of validator churn is a scheduling event, not a deposit event: the tokens never left the staked balance at all. On one ruler: whale redemptions 10,000 ETH (0.10% of supply), provider churn 1.6M ETH (3.7% of the 43.1M staked, all of it announced in advance), and ETF demand of +$196.9M in four sessions — on the order of 78,000 ETH of buy-side at the Sep 12 price (DERIVED: Farside window net / $2,513.56). Demand of that size is the backdrop against which a 0.02% supply dip reads as noise, not signal.

Source: DefiLlama coingecko feed (Sep 15); CryptoCompare via TickerReport (Sep 9); DefiLlama API (Sep 15); redemption flows per Ai Yi via GetChainNews (Sep 8).

3. DeFi / LSDfi: what changed inside the window?

The LST sector board is carried this week (Sep 7 print; no fresh full-board read we could verify), so no row-level moves are claimed. Three dated items instead. Glamsterdam slipped: Devnet-9 failed to finalize, pushing Sepolia — printed as Sep 28 in our September 7 edition from a different source, now dated Oct 6 — into conditional territory, and putting the early-December mainnet date we carried there at risk with a 2027 path opening; EIP-8037's state-gas dimension is the specific source of complexity (P2P.org's September dispatch). The same dispatch quantifies the demand case: Layer 2s processed ~25.01M daily transactions in the Sep 7 UTC session vs 1.86M on the base layer. And Frgmnt shipped an institutional custody integration on Sep 11 — adjacent to stETH's corridor, tracked, not yet load-bearing for it.

The L2 ratio is worth fixing at one number, because it will be cited for months: ~25.01M daily transactions on Layer 2s against 1.86M on the base layer is a 13-to-1 ratio (DERIVED) — the settlement layer now finalizes an order of magnitude less activity than the layers settling on top of it. For the staking economics this site tracks, the linkage runs through fee markets: a Glamsterdam slip that delays the fee-market changes also delays the yield upside that would narrow the 133bp cash-gap printed in section 1. The upgrade timeline and the yield corridor are the same story on different clocks.

Source: P2P.org, DeFi Dispatch — September 2026, Issue 1 (citing CryptoTimes/Yahoo Finance for the Glamsterdam timeline).

4. Competition: where does Lido stand after the week's turn?

Three observations, each dated. First, the V3 vault ecosystem produced another operator-led launch (Stakely, Sep 9: public vault + non-custodial institutional vaults) — the moat Lido is building is a marketplace of configured vaults, not a single product. Second, the concentration datum: BitMine's 5.067M staked ETH is 11.8% of the network's staked supply (DERIVED on the Sep 12 total) — the largest single treasury behind Ethereum's validator set, delegated across operator infrastructure. Third, LDO fell 8.2% in-window ($0.40398 → $0.37069 close-to-close on our series) and we found no Lido-specific disclosure inside the window to attribute it to — so we don't. Absence of a story is a finding; printing one anyway is how bias gets in.

One cross-check on the LDO move before it hardens into narrative. Three reads on our series: $0.4007 (Aug 23) → $0.3527 (Aug 30, -12.0%) → $0.3917 (Sep 6, +11.1%) → $0.37069 (Sep 13, -8.2%) — a cumulative -7.5% over three weeks (DERIVED), two violent weeks that nearly cancel, ending below where the arc started. The first two legs are Coinbase daily closes and the last is the new fixed-hour series, so the endpoints carry a ~1% basis seam (disclosed in the data report) — but no basis error explains a round trip of this size. A token doing double-digit weekly swings on no protocol news is pricing something outside the protocol, rate expectations mostly. That is consistent with everything else on this page and requires no invented catalyst.

Source: Stakely/Lido announcement (Sep 9) via CoinOKC/CryptoNEWS; BitMine disclosure (Sep 8) via Coinliva; LDO series per DefiLlama coingecko feed.

5. Outlook — and edition three's calls, graded

The scorecard, graded against this week's dated reads:

Watchlist for the Sep 14–20 edition: (1) exit-queue drain rate — if Kiln re-stakes as stated, 2.656M ETH should fall toward zero inside ~2 weeks; a stall flips the story from churn to supply. (2) the peg discount — 0.9987 is 13bp; watch whether it closes (redemption flows find the queue) or widens (leverage unwinds first). (3) FOMC Sep 15–16 — the macro leg of every carry table on this site reprices this week. (4) LDO's supply basis — four weeks conflicted; resolution would let us print a market cap without an asterisk.

A note on how this list is built, since four items is fewer than most weeklies print. Each entry names a number, a direction and a deadline: the drain rate (below ~100k ETH by ~Sep 26), the peg (13bp now; a trend call waits for the second print), the FOMC (Sep 15–16, dated) and the supply basis (four weeks conflicted). Items that cannot be settled by a dated public read — sentiment, institutional appetite, narrative momentum — are excluded by construction, not by oversight. A watchlist you cannot lose on is not a watchlist; it is marketing.

Source: scorecard inputs as cited in sections 1–4; FOMC date per the calendar carried since edition two.

FAQ

Related reading

• stETH Yield Update — Week of September 7–13, 2026
• Weekly Intelligence — Week of August 31–September 6, 2026