stETH Yield Update — Week of September 28–October 4, 2026

1. The week for yield, in one line

The yield story this week is that there is no new yield story — and that is the disclosure. Ebunker's October weekly did not land inside the window, so the staking register stands exactly where edition six left it: a 2.18% network base APR, a 2.52% stETH 7-day average, and a Lido user-net band of 2.21–2.24% (mid-point 2.22%). No week-over-week change is claimed for any of them, because no fresh dated read exists inside September 28 – October 4 to support one. What the week did add was a stress test of the yield's continuity, not its level: a 523,000 ETH security exit ran through the protocol's redemption machinery, and the question the next edition answers is whether that exit shows up as a dip in the network or stETH 7-day print. This update sets the pre-event baseline and works the estimate.

Source: Ebunker Ethereum staking weekly (Sep 28, 2026), carried; user-net and provider APR bands per their dated sources in the data report; the 2.18%/2.52% pair printed in edition six and unchanged this edition.

2. What a stETH holder actually earns

The gap between the headline network APR and the yield a holder keeps is the whole point of a yield update, so it is worth stating on one basis. Lido's protocol fee (10% of staking rewards) and MEV-smoothing take the 2.52% stETH 7-day average down to a user-net band of 2.21–2.24% (mid 2.22%) — the figure this site has carried since edition five, and the one the carry table uses. A holder of 10 ETH at the Oct 4 print ($2,694.13535, position $26,941, DERIVED) earns about $598 a year at that mid-point. Set that against cash: the 3-month T-bill at 3.92% (Sep 1 print) returns $1,056; the 2-year at 4.34% (Aug 28) returns $1,169; the 10-year at 5.11% (Sep 23) returns $1,377. The carry gap is −$458, −$571 and −$779 respectively — all three still negative, all three computed on the same 2.22% mid-point.

The thing that changed this week is not the gap but its direction of travel. The October 2 jobs report collapsed October-hike odds from roughly two-thirds to a hold probability near 72–77% at the October 28 meeting (FedWatch, via China Securities Journal). Every probability point that migrates from “hike” to “hold” narrows the cash leg without a single basis point of staking yield moving — so for the first time since the September 16 hike, the carry tables are pointing the right way even though they are still red.

Source: position value DERIVED on the DefiLlama coingecko feed Oct 4 print; staking and cash rate rows carried with their print dates; FedWatch per China Securities Journal (Oct 2, 2026).

3. The 39bp step, and why it is carried not resolved

Last week's defining print — the network APR's −39bp step to 2.18% — is examined in the 39-Basis-Point deep dive, and it frames everything this update can and cannot say. The step was real on the Ebunker series (2.62% → 2.57% → 2.57% → 2.18% across September), it was the largest single-week move in this site's records, and it has now been carried for two editions because the Ebunker October weekly did not land. One more read at or near 2.18% would turn a step into a regime; one read back up would make it a one-week aberration. This edition takes neither side: the board is carried, the question is open, and the next obtainable Ebunker print — post-window, so in the next edition — is the one that settles it. A carried register that says it is carried is more informative than a fresh-looking number with an undisclosed basis, and the 2.18% print is precisely the kind of print that should not be quietly “refreshed” with a source that does not grade the same series.

Source: Ebunker weeklies (Sep 7, 14, 28) and the carried Sep 28 print; the 39bp deep dive (Sep 29 edition) for the step's mechanics.

4. Bitmine's own 2.62% — a different basis, printed side by side

The treasury bid reports its own number, and the two should not be averaged. Bitmine's September 28 SEC exhibit listed 6,001,302 ETH in treasury and 5,067,309 ETH staked, with a 7-day yield print of 2.62%, flat on the week. That 2.62% is Bitmine's staking revenue on its own accounting — ahead of the protocol's fee split and before the timing effects that show up in any 7-day window — so it is not comparable with the network's 2.18% base APR or Lido's 2.52% stETH 7-day. The useful comparison is directional: Bitmine's print held at 2.62% across its September disclosures (2.62% → 2.57% → 2.62%) while the network rate fell 44bp, which is what a large, sticky staker looks like when the marginal validator's yield is moving — its average barely flinches. The staked balance, 5,067,309 ETH, is roughly 11.6% of all staked ETH (DERIVED against the 43,681,037 ETH Sep 28 board), so Bitmine's yield behavior is now a non-trivial slice of the network average's next move.

Source: Bitmine SEC exhibit (Sep 28, 2026) via MEXC research; network staked total per the Ebunker weekly (Sep 28, 2026).

5. The MetaMask exit: a worked estimate of next week's yield drag

This is the section the rest of the edition points at. MetaMask's precautionary exit removes roughly 523,000 ETH from earning position — about 1.2% of the 43,681,037 ETH staked network-wide (DERIVED), or 5.3% of the stake Lido itself backs (DERIVED on the in-window TVL-implied 9.84M ETH). Exited validators earn nothing until they re-enter. So the question is how much that idle stake drags the network average.

Upper bound, stated as an illustration: if 1.197% of stake earned 0% for a full week instead of the 2.18% base, the blended network APR for that week would be 2.18% × (1 − 0.01197) ≈ 2.154% — a drag of about 2.6bp (DERIVED). That is the most the incident could cost the network average in a single week if the entire exited stake sat idle the whole week. In practice the drag is smaller and spread out: the exit queue processes at ~57,600 ETH/day, so the 523,000 ETH leaves earning position gradually over roughly nine days, and the stake re-enters through a 27.7-day entry queue after the exit completes — meaning the idle window for any given ETH is closer to the 45-day cycle Lido cites than to seven days. The honest read: a measurable but small, gradually-applied drag on the network 7-day average, most visible in the stETH print (which is more sensitive to Lido-node downtime than the raw network figure), and exactly the kind of dip the next edition's yield update will be able to attribute rather than estimate.

Source: exited stake and queue mechanics per the news review and the exit-queue deep dive; network staked total per the Ebunker Sep 28 board; the 2.6bp drag DERIVED on the 1.197% idle share and the 2.18% base; the smoothing caveats per the churn arithmetic in the deep dive.

6. Issuance and burn: the small negative under the yield

Staking yield is not the whole return — ETH's own supply change sits underneath it. The carried Ebunker Sep 28 window showed net issuance of +19,815.93 ETH over seven days, about +0.016% of that series' supply denominator, annualizing to 0.86% (DERIVED). Against that, the burn side keeps the carried pair that frames it: a 30-day average base fee of 0.179 gwei against the 13.384 gwei needed to hold supply flat — a ratio of roughly 1/75. The implication for a yield holder is the same every week: staking pays ~2.2% net while ETH supply grows ~0.86%/year, so the real carry on staked ETH is the staking yield minus the dilution, and the cash leg still wins on the nominal comparison this site runs. None of those figures refreshed inside the window; they are carried with their dates, and the burn side in particular is a threshold (13.384 gwei) rather than a weekly print.

Source: Ebunker weekly (Sep 28, 2026) for the issuance window and the carried base-fee and breakeven figures (printed Sep 2); annualization DERIVED.

7. Method and what was withheld

Two editorial decisions belong in the body, not the footnotes. First, the supply cross-check is withheld. The stETH supply input behind the supply×price comparison is dated September 9 — 26 days before the window — so rather than run a stale input as current, the data report suspends that check and derives the same quantities (ETH under Lido management, share of stake) from two in-window prints instead. This update uses those derived figures and labels them as derived. Second, no post-window data. Network reads, treasury updates and flow sessions dated after October 4 exist and are deliberately not used — not in tables, not in prose. The yield numbers below all trace to the carried Sep 28 board or to Oct 4 price prints; the next edition's yield update is where the Ebunker October print and Bitmine's next weekly land.

Source: site_data.py (Snapshot #7, as of Oct 4, 2026) and the edition archive; supply-withhold note per the data-report ledger.

8. What to watch for yield

Four markers, all checkable on dated prints. (1) The Ebunker resume — the next obtainable weekly grades whether 2.18% is a step or a regime; until it lands, the board is carried. (2) The MetaMask drag — the next network and stETH 7-day prints, watched for the ~2.6bp upper-bound dip this update estimates, with the caveat that smoothing spreads it across the 45-day cycle. (3) The Bitmine share — whether its staked balance moves off 5,067,309; accumulation has flowed into treasury, not new validator positions, so a staked-number move would be the signal. (4) The October 28 FOMC — a hold narrows every cash leg and, for the first time since the September hike, points the carry tables the right way without any staking yield changing. The yield story next week is almost entirely about whether the carried board finally refreshes and whether the exit shows up in the print.

Source: markers constructed from the figures in sections 1–6; the watchlist is the yield half of the intelligence scorecard.

FAQ

Related reading

• Lido & stETH Weekly Data Report — Week of September 28–October 4, 2026
• The 39-Basis-Point Week: What Just Happened to Staking Yield
• The Exit Queue Shock: What 850,000 ETH Exiting Actually Means