stETH Yield Update — Week of September 7–13, 2026

1. The rate board — carried, and labelled as carried

Honesty about what did not print this week, first: no provider in our rate board published a fresh dated read inside the window that our pipeline could verify — not Lido's frontend, not the rETH market, not the third-party net-rate trackers. The board below is therefore carried from the Sep 2/Sep 7 prints, exactly as published, with no re-derivation. What is fresh is the network-level anchor: staking APR 2.59% (validatorqueue, Sep 10), down 2bp on the week, against the carried 3.92% three-month bill — a 133bp shortfall (DERIVED).

The gap arithmetic is unchanged in shape: bills out-pay the carried user-net corridor by 168–171bp (DERIVED: 3.92 − 2.24, 3.92 − 2.21) and the two-year out-pays it by 210–213bp (DERIVED). Until a provider prints again, the corridor is a carried object and this update says so on every line.

The carried board still supports a per-unit read, which is the useful way to carry it. Per $10,000 at the carried user-net mid (2.22%), the year pays $222 (DERIVED); the same $10,000 in the carried bill pays $392 — a $170 annual shortfall per $10,000 (DERIVED), and the two-year leg widens it to $212 (DERIVED). Per whole ETH at the Sep 15 print, the user-net mid earns $55.89 a year (DERIVED: $2,517.76 × 2.22%). These are the same gaps as the table, rescaled. Rescaling cannot make negative carry attractive, but it does make it quotable in the units people actually hold.

2. The carry trade, on this week's price — 10 ETH, one year

Position value uses the Sep 15 same-source read: 10 ETH = $25,177.60 ($2,517.76). Rates are the carried board. Every line is DERIVED, on one price and one set of carried rates:

The gaps: −$428 vs bills and −$534 vs the two-year, per 10 ETH staked. Two moving parts, one of which moved: last week's $422/$526 gaps sat on a $24,799 position; this week's are wider by ~$6–8 purely because ETH's price is higher — the rates themselves did not change (they were carried). A wider gap on a carried rate is a price artefact, not a yield event, and this section exists to keep that distinction un-blurred.

3. Peg watch — the first discount print

On the Sep 15 same-source pair (DefiLlama coingecko feed, one timestamp): stETH $2,514.60 vs ETH $2,517.76 → 0.9987 — a 13bp discount, the first negative print in this column's series after editions at 1.0003 (twice) and clean par before that. For scale: the Aug 29 stress test — the worst macro day of this cycle — did not produce a discount. We flag the direction change and withhold the story until a second print confirms or reverts it; a single 13bp fixed-hour read is a data point, not a regime. One cross-source pairing is not used for the peg here: CryptoCompare's Sep 9 stETH print ($2,484.11) against our series' Sep 9 ETH ($2,492.15) reads 0.9968 — but those are two different sources, and this column does not compute pegs across sources.

Source: DefiLlama coingecko feed, stETH and ETH, Sep 15 (same timestamp); CryptoCompare via TickerReport (Sep 9, cross-source, flagged not used).

4. LDO — the give-back, unattributed

LDO closed the window at $0.37069 (Sep 13 print on our series), −8.2% from the Sep 6 equivalent ($0.40398) and −9.4% from the Sep 7 print as of snapshot — giving back most of the prior week's +11%. The derived market cap on the carried CoinLore supply basis is $330M. The attribution section is deliberately short: we found no Lido-specific disclosure inside the window, and the two Lido events we did date (Stakely's vault launch, the queue flip) are not obviously token-relevant in either direction. The price series is the finding; the reason is not yet.

The three-week arc keeps the size of the give-back honest. On our series: $0.4007 (Aug 23) → $0.3527 (Aug 30) → $0.3917 (Sep 6) → $0.37069 (Sep 13). The last two weeks nearly cancel (+11.1%, then -8.2%); the cumulative move is -7.5% over three weeks (DERIVED), with the caveat that the series changed basis at the Sep 7 seam (~1%, disclosed in the data report). Net of everything, LDO ends the arc below where it started, on weeks of double-digit range and no protocol event — which is what a duration asset does while the rate path is unresolved. The market-cap line inherits the shape: $330M on the carried CoinLore basis, fourth week of the supply conflict, printed with the asterisk rather than resolved by picking a side.

Source: DefiLlama coingecko feed (LDO); market cap DERIVED on CoinLore's 895.8M supply basis (conflict with 836.3M basis continues, week four).

5. Flows that touch the corridor

Three flows, all dated. ETFs: +$196.9M net across the window's four sessions (Farside), with the inflow streak breaking on Sep 8 and a $216.4M Thursday — the demand side of every carry table above. Whales: 10,000 ETH redeemed from Lido since Aug 26 (Ai Yi via GetChainNews), against a supply that moved −0.02% — redemptions are being absorbed, not multiplied. Providers: Kiln's 1.6M ETH validator churn, which changes keys, not deposits. None of the three changes the carried rate board; all three shape what the next fresh board will print.

Scale each flow against the corridor it touches. The whale's 10,000 ETH redemption is 0.10% of the 9,659,594 stETH supply (DERIVED) — visible on-chain, invisible in price. Kiln's 1.6M ETH churn is 16.6% of that supply in validator terms (DERIVED) but zero in deposit terms — keys, not tokens. The ETF tape's +$196.9M window against a $24.35B TVL is a flow on the order of 0.8% of protocol scale compressed into four days (DERIVED). None of the three is large enough to move the carried rate board on its own; all three are large enough to move the peg print, which is why the 13bp in the next section gets a second-print rule instead of a narrative.

6. What would move this update next week

One fresh provider print would re-arm the whole corridor section — the carried rows are the weak leg of this edition and we would rather say so than interpolate. The FOMC (Sep 15–16) reprices both Treasury legs within days of this snapshot. And the peg needs a second same-source print: 0.9987 once is a curiosity, twice is a trend. The exit-queue drain rate, tracked daily in the data report, is the plumbing backdrop to all of it.

The FOMC arithmetic can be written down in advance, because the board makes it mechanical. The carried 3.92% bill anchors a 133bp shortfall (DERIVED: 3.92 - 2.59); every 25bp the committee delivers moves that gap 25bp in the staker's favor (DERIVED), and two cuts would still leave cash ahead at roughly 83bp. Symmetrically, a hike widens it. This is not a forecast — the September decision is not ours to make — it is the translation table, so next week's edition can grade the outcome against a number printed before the meeting. That is the scorecard's pre-commitment discipline, applied to the macro leg.

FAQ

Related reading

• Lido & stETH Weekly Data Report — Week of September 7–13, 2026
• stETH Yield Update — Week of August 31–September 6, 2026