stETH Yield Update — Week of September 21–27, 2026

1. The rate board

The board's two stETH rows refreshed downward this week, and everything else held its date. That combination — fresh rows falling, carried rows frozen — is why the corridor's shape changed more than at any time since this column began.

One disclosure, the same one the deep dive builds on: the 39bp step sits on a single provider's weekly print, and the same print's other series (the stETH 7-day average) stepped just 8bp. We print the board as sourced rather than smoothing the step into a trend, and we identify the marker that settles it — the next weekly read.

Source: Ebunker Ethereum staking weekly (Sep 28, 2026), with the Sep 21 print of the same series for the prior row; validatorqueue series as previously cited; US Treasury 3M print (Sep 1, carried); 2Y per the Aug 28 keynote-day read.

2. The carry trade, on this week's price

Position value uses the Sep 27 same-source read: 10 ETH = $27,092.58 ($2,709.26). Rates are the board above. Every line is DERIVED, on one price and one set of rates:

The gap is $461 a year against bills and $575 against the two-year — 168–171bp and 210–213bp across the 2.21–2.24% user-net range (DERIVED). Measured against the fresh 2.18% network print instead of the carried user-net mid, the shortfalls become 174bp against the carried bill (DERIVED: 3.92 − 2.18), 172bp against the rate the Fed pays on reserves, 182bp against the policy rate and 192bp against the SEP median. Two caveats ride with those four numbers: the bill is a market price not re-read since the hike, and the 2.18% is a single print — so each gap is honest at its own leg's confidence, which is why we print the pairs rather than one flattering number.

Source: position value from the DefiLlama coingecko feed, Sep 27 print; rates as tabled in section 1; FOMC statement and implementation note, Sep 16, 2026; all conversions our own arithmetic.

3. Peg watch — a week of discounts

The full daily record, on this snapshot's re-read basis (the Sep 20 row is the re-read of the date edition five printed as a 5bp premium — the ledger in the data report carries the note):

Seven of eight sessions at or below par, a floor at −13.4bp, a single +2bp touch, a close at −5.3bp. Read against the same week's tape — ETH +5.1%, ETFs +$689.8M, LDO +22.3% — the discount is better described as constant than as widening: the biggest prints came early in the week and the spread narrowed into the close. This is the first window in our daily series where the ratio spent every session but one below par; the pre-committed test stays the same. A discount of this size with open redemptions is the exit door working. A discount that widens past the round numbers with redemptions impaired is a different article.

Source: DefiLlama coingecko feed, stETH and ETH daily prints at the same fixed UTC hour, same-session re-read.

4. LDO's best week in our records

LDO rose 22.3% in-window close-to-close ($0.39973 on the re-read Sep 20 basis → $0.48893 on Sep 27), the largest weekly gain since this site began publishing. The path was violent: −11.0% in a single session to $0.38510 on Sep 24, then +27.0% off that low into the close — a second consecutive week with a double-digit round trip inside it. The market capitalization we print is $408M (DERIVED on the 834.151M supply basis resolved last week); against the re-read prior of $333M, the weekly move is exactly the price move.

What is dated inside the window that could bear on it: Vote #205 executed Sep 21, the committee's reserve authority activated Sep 25, and $689.8M of sector-wide ETF inflows. What is not dated: any Lido disclosure explaining the move. We describe the sequence and stop there — the last two times this token moved double digits, the nearest narrative took the credit and the tape reversed within a week both times.

Source: LDO daily series from the DefiLlama coingecko feed, same-session re-read; circulating supply per Analytics Insight market data (Sep 17, 2026).

5. Flows that touch the corridor

ETF flows returned. The Farside daily table reads +$270.0M (Sep 21), +$162.2M (Sep 22), +$104.5M (Sep 23), +$66.1M (Sep 24), +$87.0M (Sep 25) — a window net of +$689.8M (DERIVED), an $830.4M swing from the prior week's −$140.6M. Every session of this window out-raised every session of the last one.

The queue, on the fresh series. Entry wait 27.7 days, a fourth consecutive weekly decline (34.6 → 31.0 → 30.3 → 27.7, Ebunker series); exit wait 2.8 days, holding where it stepped two prints ago. The ETH-count series (validatorqueue) stays carried at Sep 20 — 2,480,000 ETH in, 182,352 out — and no ratio crosses the two sources. For the corridor the structural point survives the source change: a multi-week activation queue means staking supply cannot chase the rate quickly in either direction, which cuts both ways this week — it shields stakers from competition when yield falls, and it delays the response if it rises.

Concentration, updated. Bitmine's Sep 21 disclosure puts its treasury at 5.98M ETH ($17.1B), with more than 5.06M staked and $359M in projected annual staking revenue per the Sep 21 report (DERIVED). Against 43.68M staked network-wide, one corporate balance sheet sits behind roughly 13.7% of all staked ETH (DERIVED) — and it kept accumulating through the week the network yield printed 2.18%.

Source: Farside Investors Ethereum ETF daily flow table; Ebunker weekly (Sep 28) for queue waits; validatorqueue.com via Tremplin (Sep 20, carried); Bitmine disclosure (Sep 21, 2026) via Cryptopolitan and PR Newswire.

6. What would move this update next week

Four things, each with a number attached. The next Ebunker print — if the network APR reads near 2.18% again, the step becomes a regime and the carried rows of section 1 become unsustainable; if it reverts toward 2.5%, this week was a print error and the record gaps retrace. The 0x02 CSM launch — the committee expects October, the documentation says Q4; it is the trigger for the deposit-reserve decision and the first genuinely new deposit channel in the corridor. The reserve motion — if the committee zeroes the 1,500 ETH target, watch stressed-case withdrawal finalization against the modeled 6.3 days. A fresh dated stETH supply print — twenty days stale now, and the 1.2% gap cannot narrow honestly forever on a frozen input.

What we would not read as a signal: the peg touching par intraday. One +2bp print in eight sessions is noise around a working redemption mechanism; five consecutive premium sessions after a discount month would be a signal, and that is not what printed.

Source: supply and TVL reads as cited in section 1 and the data report; committee statement (Sep 2) via CryptoSlate; peg series from the DefiLlama coingecko feed.

7. Gross versus net: the spread is about to be tested

The rate board prints three different stETH yields, and this week they disagree by more than usual. The network staking APR is 2.18% — what the chain pays, fresh and sharply lower. The stETH 7-day average is 2.52% — Lido's smoothed headline, down only 8bp. The user-net band is 2.21–2.24% (carried since Sep 2) — what a holder keeps after the 10% fee and the smoothing buffer. The 7-day headline now sits 30bp above the carried net mid (DERIVED: 2.52 − 2.22), the widest that wedge has been in our records, and the reason is mechanical: smoothing is designed to slow the pass-through of exactly the kind of step the network print just took.

The honest question this column has been deferring for seven carried weeks is now live: if the 2.18% network print is real, the user-net band is the next number to move — down. We have no fresh provider print of it (the frontend endpoint was unreachable this week), so we carry the Sep 2 band, labeled, and we flag the asymmetry: every number in section 2's table that uses 2.22% is, if anything, flattering to staking. A staker reading the 2.52% headline should subtract the wrapper's cost before comparing with a Treasury bill — that arithmetic has not changed; what changed is the size of the pie it divides.

Source: network APR and stETH 7-day average per Ebunker Ethereum staking weekly (Sep 28, 2026); user-net band as previously cited (Sep 2); fee share per Lido documentation.

8. The LSDfi shelf competes for the same ETH

The alternatives re-priced upward this week, on fresh reads. EtherFi carries $5.19B in TVL (DefiLlama, Sep 27, +5.6% on the week — the fastest grower on the board) and Rocket Pool $1.41B (+2.5%), with rETH's APY mid-point carried at 2.19% (Sep 2). Note what that carry means right now: the carried rETH rate sits 1bp above the fresh 2.18% network print — the first time in this column's run that the mid-competitor's paper rate has reached the network's fresh print. We print both with their dates; we do not conclude parity from rates nine weeks apart in freshness.

The structural thread from last week holds and tightens. CMv2 concentrates Lido's operator set into 34 bonded firms; Rocket Pool's pitch is the opposite shape. If the network yield is genuinely stepping down, the decentralisation premium becomes a harder sell at the margin — buyers pay for shape only when the yield is comparable, and this week moved the comparison against every staker. The next queue and TVL reads will say whether any of this mattered.

Source: EtherFi and Rocket Pool TVL per DefiLlama protocol-history reads (Sep 29, 2026); rETH APY per the section 1 board (carried Sep 2).

9. Where the record gaps sit in the series

Every gap this column tracks set a record this week, and the mechanism matters as much as the level. Last week's records were set by the Fed — the policy rate rose 25bp while the network APR fell 2bp. This week's were set by the yield alone: the Fed did nothing, and the shortfall against the carried bill still moved from 135bp to 174bp (DERIVED: 3.92 − 2.18), against the policy rate from 143bp to 182bp, against the SEP median from 153bp to 192bp. The squeeze on staked ETH is now two-sided in the record books — the reference side rose in September, and the yield side may have started falling in the same month.

The reason we print the pairs rather than the biggest number: the four gaps measure different things. The bill is a market price that predates the hike; the policy rate is administered and current; the SEP median is a projection through 2027; the network APR is a single unconfirmed print. A reader who wants one line should take the pair whose confidence they share — and should notice that the least confident leg, the 2.18%, is the one doing all the work this week. The next print decides whether this section's records stand or retrace.

Source: FOMC decision and SEP (Sep 16, 2026); network APR per Ebunker weekly (Sep 28); bill rate carried (Sep 1); all gaps DERIVED.

10. The 39bp in dollars

A basis-point step is abstract until it is a balance sheet. On the edition's own position — 10 ETH = $27,092.58 at $2,709.26 — the drop from the prior 2.57% print to 2.18% is $105.66 a year of staking income (DERIVED: $27,092.58 × 0.0039). Scale the same 39bp to the network: staked ETH of 43.68M at the same price is a $118.4B base, so the step removes roughly $461M of annual yield from stakers versus the prior print (DERIVED, illustrative: $118.4B × 0.0039).

The reason this column prints pairs rather than one total is visible in that $461M: it is a DERIVED figure on a single weekly print, and it inherits every fragility section 1 attached to the 2.18% — one provider, unconfirmed by a second source this week, with the same print's smoothed series down only 8bp. If the next print reverts toward 2.5%, the $461M was a measurement artifact; if it holds, it is a real annual transfer out of stakers' pockets and into ETH's price-supported scarcity. The number is honest at its own confidence, which is exactly why the whole question rests on one more weekly read.

Source: position value and staked total as cited in sections 1–2 and the data report; all dollar conversions our own arithmetic on those inputs.

FAQ

Related reading

• Lido & stETH Weekly Data Report — Week of September 21–27, 2026
• The 39-Basis-Point Week: What Just Happened to Staking Yield
• stETH Yield Update — Week of September 14–20, 2026