Lido & stETH Weekly Data Report — Week of August 31–September 6, 2026

1. Executive summary

The third snapshot is the first where the network moved more than the market. Five findings, one of which is a correction to our own published work:

• The cross-check passes again. Since we corrected our ETH price the supply×price test had been failing. It now reconciles: 9.66M stETH × $2,480.62 (its own Sep 5 print) = $23.97B against a DefiLlama TVL of $23.73B — inside 1%.
• Which means our own Aug 31 TVL was wrong, and we are correcting it. The $18.7B we carried came from a table whose own USD column priced ETH near $2,105. Corrected to the Aug 31 close, the same supply reads ~$23.3B. Any "+27% TVL week" you see computed off the old pair is an artifact of our error, not a capital inflow. Our derived move on the corrected basis is +1.8%; the source's own weekly change prints −1.97%, which implies a prior-week value about 4% above ours. Both fit inside a ±2% band we cannot yet tighten. Neither is +27%.
• Staking saturation set a new record: 42.92M ETH staked, 35.55% of supply — up from 41.89M and 34.7% on Aug 13. The entry queue holds 2.07M ETH with a 36-day wait; the exit queue is empty.
• The parity call is resolved. stETH's net user rate (2.21–2.24%) and rETH's (2.18–2.20%) now sit within a few basis points of each other — the founding edition's corridor-parity prediction, finally testable on a same-basis read, and it holds.
• LDO reversed: +11.1% to $0.3917, against a flat ETH tape — beta and narrative, not fundamentals, in both directions.

2. The WoW table

Rows marked "withheld" or "not comparable" are the column's contract: we refuse to manufacture a delta from mismatched bases. Every other row reconciles on the stated basis.

3. The yield corridor, on a wider source base

This week we can finally put the whole stack on one page, because each layer now has a dated source:

One honest complication, because it matters for anyone quoting a single stETH yield: these rates do not sit on one basis. The 2.60% seven-day figure is a protocol/gross rate; the 2.21–2.24% band is described as the published user rate already net of Lido's 10% fee; the frontend's 2.40% sits between them and states no basis. Net of a 10% fee, a 2.60% gross would read ~2.34% — above the observed band, which suggests the observed band is measured over a slightly weaker rewards window. Rather than pick the number that flatters a narrative, we publish the 2.21%–2.40% corridor as the defensible range for what a stETH holder actually earns, and flag the ambiguity. This is the third consecutive week that cross-source basis differences, not market moves, have been the main obstacle to a clean yield series.

4. Competitive board — the custodial bloc keeps growing

From the DefiLlama weekly wrap covering the window:

Ethereum held $48.95B of DeFi TVL, 56.3% of the cross-chain total (implying ~$86.9B industry-wide), down 1.81% on the week. Three reads. One: Lido remains the largest DeFi protocol of any category, and its nearest liquid-staking competitor is 2.6× smaller ($23.73B vs $9.15B) — together the two hold $32.88B, with Lido taking 72% of that pair. Two: the entire board declined together, which tells you this was a price week, not a flows week — and it sits inside the ±2% band our corrected basis produces rather than the +27% artifact. Three: Spark's −21.24% is the only idiosyncratic move on the board and the only one worth investigating next issue.

5. New section — the queue, and what it says about the network

This is the first week we can publish the network's plumbing alongside the protocol's, and the two numbers do not agree with each other:

• Entry queue: 2,074,270 ETH (~$4.92B at the Sep 2 spot of $2,373; the close that day was $2,417.92), 36-day wait. If it fully processes, the staked ratio rises to ~36.6% — a figure the source computes on the Sep 2 read of 34.94%; against the 35.55% print below, the same queue implies ~37.3%.
• Exit queue: 0 ETH. Exit wait under three hours.
• Staking ratio: 35.55% of supply, a record; 42.92M ETH staked.
• Fee engine: 30-day burn 1,160.53 ETH against ~87,700 ETH issued — 1.3%. Average base fee 0.179 gwei against the 13.384 gwei needed to hold supply flat.
• Net supply growth: +0.86% annualised.

A one-way door with a 36-day line to get in and nobody waiting to get out, in a week when staking paid 131bp less than Treasury bills, is not a yield trade. Our deep dive this week takes it apart properly; here we only record the figures and note that they are the most consequential new data on the page.

6. Data-quality ledger — including two corrections to ourselves

Corrected this week:

• Our Aug 31 Lido TVL of $18.7B is superseded. It came from an aggregator table whose USD column priced ETH near $2,105 — we flagged the vintage caveat at the time and carried it anyway. Corrected to the Aug 31 close of $2,417.82 on the same supply, the figure is ~$23.3B. The Aug 31 data report and the August monthly report both carry notes pointing here; the wrong number stays visible above them.
• The supply×price cross-check now passes — and its failure last week is explained: the TVL side was the stale one, not the supply side. 9.66M × $2,480.62 (the Sep 5 stETH print) = $23.97B vs TVL $23.73B, a 1.0% gap.
• The 600 vs 900 node-operator conflict is resolved. lido.fi's own scorecard now publishes 900+, settling it in favour of the higher count. Curated 586, CSM 320 (including 182 solo/community), SDVT 35 clusters.

Still unresolved:

• stETH supply: 9,661,935 (crypto.news, Sep 5) vs 8,890,883 (lido.fi counter). The 9.66M read is the one that reconciles with TVL; the 8.89M read misses by 7.1%. Both printed.
• Lido TVL mirrors disagree: one DefiLlama mirror prints $23.73B, another $21.42B. lido.fi's own UI prints $16.05B against its own 8,890,883 ETH counter — which implies ~$1,805/ETH, a price no venue quotes. We take $23.73B and flag the rest.
• The source's own weekly change disagrees with our corrected baseline. The same wrap that prints Lido at $23.730B reports a −1.97% weekly move, implying a prior value of ~$24.2B — about 4% above the $23.3B we derive. We print our derived figure and flag the gap rather than adopt whichever flatters the delta.
• LDO circulating supply: 895.8M vs 836.3M vs whatever DefiLlama's ~$270M market cap implies. Third consecutive week.
• Cumulative rewards paid: lido.fi reads $2.107B since 2020; our Aug 27 edition carried $3.085B. A cumulative counter cannot fall, so one is wrong. Flagged, metric frozen until reconciled.
• Validator count: 904,541 (Sep 2) vs ~789K (Aug 13). A +14.6% jump exceeds what a 36-day queue can deliver — almost certainly a counting-basis difference. Delta withheld.

7. Macro overlay

Three benchmarks now, each with a date: the 3-month bill at 3.92% (Sep 1), the 2-year at 4.34% (Aug 28), and staking at 2.61% (Sep 2). Against the user-net corridor of 2.21–2.40%, the staking trade is short of short-dated risk-free cash by 152 to 171 basis points. September hike probability is carried at ~56% with no new print inside the window. US spot ETH ETFs held $15.61B in net assets — 5.15% of ETH's market cap on our $303B print (DERIVED; the source's own ratio reads 5.23% on a smaller cap basis) — after eleven consecutive inflow sessions through Aug 31, which is the demand side of the queue story in item 5 and the one piece of genuinely staking-positive institutional data we have.

8. Three-snapshot lineage

Four of six rows carry a source conflict or a correction. That is not a defect in the table — it is the honest state of third-party liquid-staking data in September 2026, and it is why this column exists.

The quiet headline of this edition is that the measurement apparatus works in both directions. This column published a correction against its own archive: the $18.7B TVL carried in late August was mispriced, the corrected basis is ~$23.3B, and both the original and the fix remain visible on the archive page. That correction is why this week's supply×price cross-check passes — the derived figure within 1% of the reported TVL — and why a "record TVL growth" print computed elsewhere (+27% in a week) can be identified as an artifact rather than a capital inflow. A data operation is judged less by never erring than by whether its errors are findable, priced and propagated forward consistently. The table above is built to survive that test: same-source, same-date reads, corrections carried, conflicts printed.

9. Conclusion

Three things happened this week, in order of importance. One: the cross-check came back, and in doing so exposed that our own Aug 31 TVL was mispriced — corrected, flagged and left visible on the archived pages. Two: the network, not the market, produced the week's real news: a record 35.55% staking ratio, 2.07M ETH queued for 36 days, and an exit queue with nothing in it. Three: the founding edition's parity call finally became testable and held — stETH and rETH net rates are within a few basis points. Next issue: same-day reads on the two stETH supply sources to close the last big conflict, and a Spark follow-up to explain the only idiosyncratic move on the competitive board.

FAQ

Related reading

• Lido & stETH Weekly News Review — Week of August 31–September 6, 2026
• The Empty Exit Queue: 2.07M ETH Is Waiting 36 Days to Earn Less Than Cash