Weekly Intelligence — Week of August 31–September 6, 2026

1. Staking. How healthy is Ethereum staking?

Saturated, and setting records while paying less. 42.92M ETH is staked — 35.55% of supply, an all-time high, up from 41.89M and 34.7% on Aug 13. 904,541 validators are active. The network pays 2.61% APR against a 3.92% three-month bill. And the plumbing is extraordinary: 2,074,270 ETH are queued to enter with a 36-day wait, while the exit queue is empty. Underneath, the fee engine has stopped contributing — 1,160.53 ETH burned against ~87,700 issued over 30 days (1.3%), with the average base fee at 0.179 gwei against the 13.384 gwei needed to hold supply flat. Verdict: structurally healthy, economically strained. More capital wants in than ever; the network itself is paying for it almost entirely out of issuance.

2. stETH. Supply, rate and anomalies

Supply reads 9,661,935 stETH (Sep 5), up 0.4% from the 9,626,252 we carried on Aug 27 — the first supply delta this column has been willing to publish. The conflict with lido.fi's own counter (8,890,883 ETH) remains open, but the balance of evidence now favours 9.66M: it is the only read that reconciles supply × price ($23.97B) with DefiLlama TVL ($23.73B). The user-net rate sits in a 2.21%–2.24% band against a 2.40% frontend APR — the gap is a basis ambiguity we flag rather than paper over. On anomalies: none dated inside the window. The peg printed $2,480.62 against ETH's $2,479.92, a ratio of 1.0003.

3. DeFi/LSDfi. What changed inside the window?

Very little, and that is the finding. Ethereum held $48.95B of DeFi TVL — 56.3% of the ~$86.9B cross-chain total — down 1.81% on the week. Lido $23.730B (−1.97%), Aave $18.074B (−1.37%), Morpho $9.658B (+0.18%), Binance Staked ETH $9.148B (−2.12%), EigenCloud $6.436B (−3.27%), Sky $5.672B (−3.19%). Only Spark, at −21.24%, moved idiosyncratically. A board that declines together in a week when ETH rose 2.6% is a board being marked down on something other than flows — most likely collateral values on the lending side. For LSDfi strategies the live fact is unchanged: net yields of ~2.2% do not fund leverage, so looping activity stays subdued, and a subdued loop count is one plausible reason the exit queue has nothing in it.

Worth naming the mechanism, because it is arithmetic rather than sentiment. A leveraged stETH loop borrows at roughly 4–5% against collateral yielding ~2.2%, so every turn of the loop is cash-flow negative before any price move. That is the cleanest available explanation for the empty exit queue: the structures that generate redemption pressure are not economical to run at this corridor. It also means the queue is a live monitor rather than a static fact — if the corridor lifts, or if borrowing costs fall, looping becomes viable again and exit demand reappears. We would see it in the exit queue first, which is why that number now sits at the top of our watchlist.

4. Competition. Where does Lido stand?

Position intact, measurement repaired. Lido remains the largest DeFi protocol in any category at $23.73B, 2.6× its nearest liquid-staking competitor (Binance Staked ETH, $9.148B), and 72% of that $32.88B pair. Share of all staked ETH is 22.5% — the first clean read we've published, because both inputs are now dated within two days. On rates, the week's real result: parity with rETH is restored (stETH net 2.21–2.24% vs rETH 2.18–2.20%), which closes a question this column has carried unresolved since the founding edition. The strategic read is unchanged from last week and worth repeating: in a regime where short-dated cash pays ~1.5–1.7 points more than staking, Lido's moat is integration depth and exit capacity — and exit capacity is currently abundant.

5. Outlook. What to watch next week?

• The entry queue's composition. 2.07M ETH is the number that decides whether this is organic demand or treasury/ETF plumbing. Any disclosure that breaks it down would be the most valuable data point of the month.
• Same-day reads on both stETH supply sources — our own checklist item, to close the last major conflict on the page.
• Spark's −21.24% — the only idiosyncratic move on the board and therefore the only one that might actually be a story.
• Sept 15–16 FOMC still unresolved, still the event every LST valuation keys to.

One caveat on the list above: two of these four items depend on disclosures that may never come. The queue's composition in particular is something no source currently publishes, and if it stays unpublished this column will keep flagging it as unresolved rather than filling the gap with a plausible guess. Of the other two, the same-day supply re-read is on us and the FOMC date is fixed. A watchlist that admits which of its items are unobservable is worth more than one that sounds complete.

6. Scorecard: grading edition two's calls

Edition two made four trackable calls. Grading them against this window:

Scorecard: 2 held, 1 delivered by an unexpected route, 1 converged but unresolved. Nothing contradicted. One note against ourselves: we did not anticipate that our own TVL figure was the mispriced side of the failed cross-check. That is a miss in the data column, and it is logged here rather than quietly fixed.

Edition two's scorecard carries the useful lesson of the month: direction is easier to hold than mechanism. The calls that held — the custodial tier taking share while the board fell, and ETF flows turning staking-positive — arrived for the predicted reasons. The corridor call is the one that converged without resolving, because the two rates it compares are not yet on one basis. The call that scored direction-right arrived by an unexpected route: the TVL "surge" turned out to be our own mispriced baseline, not capital inflow, which is exactly the failure mode a corrections ledger exists to catch. The takeaway for future outlooks is procedural: predict levels, not narratives; commit to the source and date that will settle each call; and treat an unexplained right answer as unresolved rather than vindicated. A scorecard that only counts hits teaches nothing; one that asks how each hit happened is a research tool.

7. How to use this column

Three layers, in reading order: the five answers give you state; the outlook gives you triggers; the scorecard gives you accountability. This week's edition is a good demonstration of why the third layer exists — the cross-check got repaired, but not for the reason we predicted, and the scorecard is where that discrepancy goes on the record. The five questions never change, so any two editions read side by side; the scorecard makes sure our misses cost us something.

FAQ

Related reading

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