ENS Endowment: Independent Risk Teardown

🗳️ Meta-Governance Treasury Management
1

Hi all :waving_hand:, disclosure first: I’m part of the team behind Sentralis, a cryptocurrency portfolio risk- and scenario-analysis tool, which produced this analysis. Nobody at ENS asked for or paid for it, and it takes no position on any of the ongoing governance questions.

Everything below is built from the endowment’s own published positions: the seven open positions in the June 2026 report, covering >99.9% of the open book. Data basis and limitations are spelled out at the end.

The timing for this analysis is deliberate. The DAO adopted a new Investment Policy Statement two weeks ago (6.46, with hard numeric limits), an expansion of the mandate into on-chain options is under discussion, and the June report is the first to cover ETH below $1,600.

The analysis is descriptive and should not be read as advice. If you spot any error in the data basis, please let me know. And if the Meta-Gov working group or the manager would rather see this with their own assumptions, I’m happy to re-run it.

TL;DR (five model estimates; the assumptions and data coverage they depend on are spelled out at the end):

  1. The book tracks its policy target. At the June report the split was ~61% ETH-based / ~39% stablecoins against the 60/40 target adopted June 29. At July-14 prices, after ETH’s small bounce, the ratio drifts to ~64/36. ETH, ETHx and weETH moved at pairwise correlations of 0.89–0.99, so not unexpectedly, the ETH side behaves as a single position.

  2. No protocol breaches the new 30% cap. The largest single-protocol exposure is StakeWise at 25.3%, followed by Stader at 24.5%. A modeled single-protocol incident (temporary freeze, 80% recovery, 20% token haircut) comes out at $6.5–6.7M (~9%) for either of the two big staking venues, and $3.7M (5%) for the weETH position.

  3. A repeat of the Oct-2025→Feb-2026 bear market would take ~$25M (−34%) off the current portfolio. The replay runs on each position’s full price history.

  4. The 1-in-20 bad year lands between −41% and −51%, depending on model choice (two Monte Carlo models, 25,000 paths each, same seed). The probability of ending the year below the IPS’s $49.34M liquidity-floor figure spans 12% to 39% across those two models. That spread is model risk, and it is wide enough to matter for anything sized off a single number.

  5. About $12.4M could be sold through order books within a crisis week, all of it USDC. Under crisis assumptions (5% of daily volume), the DAI position needs ~11 days and everything else is slower. Two caveats work in the treasury’s favor: protocol redemption paths (Sky redemption, LST/LRT withdrawal queues) are not modeled and materially improve real exits, and the Stader ETHx position (~$18M) lacks usable venue-volume data in our source, so it sits outside the liquidity model entirely.

Full analysis below.

Snapshot date: positions per the June 2026 endowment report (period ending 2026-06-30, published to the monthly-reports thread); prices as of 2026-07-14 (ETH $1,784.76); analysis run 2026-07-14

Data basis: the manager’s published report book, i.e. the seven open positions (closed positions excluded); Ethereum mainnet only per the IPS

Covered value: $74.07M across 7 positions at July-14 prices ($68.37M at the report’s own end-June prices, ETH $1,569.51). Not covered: a ~$47k unstaked ETH balance and sub-$300 dust. The two valuations differ only by the price date; the quantities are identical and were verified against the report dashboards.

Backdrop

The relevant background:

  1. The 2026 IPS (6.46) passed on June 29 with hard limits: a ~$49.34M liquidity floor (three years of runway), a 60/40 ETH/stablecoin target, a single-protocol cap raised to 30%, and ≥90% of the book in low-risk strategies. So for the first time there are hard policy limits to compare stress results against.

  2. An options-mandate expansion is under discussion (covered calls and cash-secured puts, with a worked example of a $1M put at a $1,150 strike). One open question in that thread is what assignment would do to the portfolio in a drawdown; section 7 runs those numbers.

  3. Bear market context. ETH entered July ~68% below its August-2025 high after three consecutive red quarters, and the April Aave/rsETH incident showed what a single staking-derivative event can do to money-market infrastructure ($195M bad-debt exposure, $10B of TVL flight). This book holds staked-ETH derivatives across three venues and money-market positions on Aave, Morpho and Fluid.

Six scenarios were run against the portfolio, one per section below, followed by a summary. Headline figures are model outputs under stated parameter choices. Where a different defensible choice would move a number materially, that sensitivity is noted inline.

1. Composition

At July-14 prices: ETH-based assets $47.2M (63.7%): 1. StakeWise vault ETH $18.7M, 2. Stader ETHx $18.1M, 3. ether.fi weETH $10.4M. Stablecoins $26.9M (36.3%): 1. USDC $12.4M (Morpho kpk vault + Fluid), 2. DAI $11.1M (Aave), 3. USDS $3.4M (Sky SSR).

The ETH / stables split is the endowment’s own published design and will surprise nobody. Two observations on top of it:

  • The book sits on its target. At the report’s own end-June prices the split was 60.7 / 39.3 against the 60/40 target adopted June 29 (May: 58.5 / 41.5). The drift to 64/36 since then is just ETH’s July bounce; the allocation itself hasn’t changed. Without rebalancing, the ratio will keep moving with ETH.

  • The wrappers are the exposure. ETH, ETHx and weETH moved at pairwise correlations of 0.89–0.99 over the measurement window (section 4). Spreading the ETH across three staking venues limits the damage any one protocol can do; however, it does nothing to reduce the ETH exposure itself.

2. Protocol concentration

With each position attributed to the respective protocol, the largest single-protocol exposure is StakeWise at 25.3%, then Stader 24.5%, Aave 15.0%, ether.fi 14.0%, Morpho 9.5%, Fluid 7.2%, Sky 4.6%. Concentration measured as a Herfindahl index is 0.182, which is equivalent to spreading the portfolio evenly across about 5.5 protocols (1/HHI, the standard “effective number” reading of that index). No single point of failure is flagged and nothing breaches the IPS’s 30% cap, though the top two staking venues together hold about half the book.

Stress applied: a single-protocol incident (temporary freeze with 80% eventual recovery, plus a 20% token haircut during the event), run separately against each of the top three:

Protocol Exposed Modeled loss % of book
StakeWise (vault ETH) $18.7M $6.74M 9.1%
Stader (ETHx) $18.1M $6.53M 8.8%
ether.fi (weETH) $10.4M $3.73M 5.0%

The two large staking venues are nearly interchangeable in size, so a bad month at either one carries the same ~9% price tag. To be clear about what the table measures: it multiplies position sizes by an assumed incident severity, with no view on how likely such an incident is at any of these venues or on the quality of their engineering. The April rsETH incident is the relevant precedent for the weETH line: a staking-derivative event transmitted stress to lending markets without any code failure at the derivative itself. The severity assumptions are deliberately moderate, and the loss scales roughly proportionally with them.

3. Exit liquidity in two regimes

Modeled as disciplined liquidation (fixed participation in daily volume, order books only, staking-tag exit delays):

Normal (10% participation, 14-day horizon) Crisis (5% participation, 7-day horizon)
Exitable within horizon $23.5M (31.7%) $12.4M (16.7%)
Full-liquidation slippage cost $10.9M (14.7%) $13.4M (18.1%)
Outside the model (no venue data) $18.1M $18.1M

Position level (crisis): USDC exits in under a day. DAI needs ~10.6 days and misses the week. The StakeWise ETH has market depth for ~3-day exits, but the staking delay pushes full exits to ~24 days. weETH takes ~127 days at disciplined participation. USDS takes ~1.4 years; its order-book depth is thin relative to even a $3.4M position.

Three caveats, each of which makes real exits look better than the model: (a) protocol redemption paths are not modeled. USDS redeems near-instantly through Sky, and the staked positions have withdrawal queues that convert to ETH at par in days. Read these figures as time-to-cash through open markets, which is the measure that matters when everyone is drawing on those same redemption queues at once. (b) The Stader ETHx position (~$18M, 24% of the book) has no usable venue-volume data in our source and sits outside the model entirely. (c) These are current volumes; crisis volumes differ, in both directions.

Against the floor: the IPS floor figure is $49.34M. The stablecoin side is $26.9M, of which $12.4M can be sold within a crisis week in this model. Whether the floor is defined over stablecoins alone or over all low-risk strategies is a question of the IPS text; these figures are meant to give that discussion a sense of scale.

4. Correlation under a crisis regime

Baseline (365-day correlations): annualized portfolio volatility 40.9%, 1-day 95% VaR $2.61M (3.5%), diversification ratio 1.04. Applying crisis-regime correlations moves VaR by +0.7%.

The near-zero crisis effect is expected for an ETH-plus-stables book. ETH, ETHx and weETH already sit at pairwise correlations of 0.89–0.99 in the baseline, so a crisis has almost no additional correlation to destroy; the diversification ratio of 1.04 says the book is already close to a single-asset risk profile. What diversification exists lives entirely in the stablecoin side, which is why its size (section 1) and its exit speed (section 3) are the quantities worth tracking.

5. One-year Monte Carlo, two models

Two deliberately different simulation models, same book, same seed (25,000 paths each, 365 days):

  • Model-based (GBM): lognormal paths on the full covariance matrix with zero drift, so it captures the book’s volatility structure without taking a view on direction.

  • Empirical bootstrap: resamples the book’s actual daily returns from the trailing 365 days, inheriting both the real fat tails and the realized drift of a deep bear year.

1-year outcome GBM Bootstrap
5th percentile (bad year) $43.5M (−41.2%) $36.3M (−51.0%)
Median +1.4% −27.4%
1-year 95% VaR $30.5M (41.2%) $37.8M (51.0%)
Expected shortfall (95%) 46.0% 53.7%
Median intra-year max drawdown 35.7% 42.4%
Paths breaching a 20% drawdown 98.0% 99.3%
Paths ending below the $49.34M floor figure 12.0% 38.6%

The GBM median (+1.4%) simply falls out of the zero-drift lognormal setup. The bootstrap median (−27.4%) replays a trailing year that was deeply negative for two-thirds of this book. The real distribution sits somewhere between the columns, and which column a treasury plans against is a risk-appetite choice. Two results deserve attention: a 20%+ intra-year drawdown is the dominant outcome under either model (98–99% of paths), and the floor-breach probability moves by a factor of three (12% → 39%) on model choice alone. Anything sized off that floor has to live with that spread.

6. Historical replay: the current bear’s core leg

Replaying the October 2025 → February 2026 drawdown (the tariff-war bear market) on the current book, with full price history for every position: −$25.4M (−34.3%). The ETH positions contribute effectively all of it (ETH −54%, ETHx −54%, weETH −54%); the stablecoins net out to ±$18k. Two reasons for picking this event: it is the deepest drawdown in our replay catalog for which every position in this portfolio has full price history (ETHx, weETH and USDS launched too recently for the 2020–2023 events), and it is the bear market the DAO is actually in, so a repeat from today’s levels is the most concrete worst case to look at. It remains a hypothetical: it assumes the same path repeats from levels that already sit far below the October starting point.

One negative result to report: an SVB/USDC-depeg replay (March 2023) was attempted and is not usable for this book. Most positions (ETHx, weETH, USDS) have no 2023 price history, and the USDC leg itself predates our USDC series, so the replay sees the one event that specifically attacks stablecoin holdings as a −2% ETH-only event. Rather than publish a misleading number, we note the coverage gap; a depeg scenario for this book needs a parametric treatment (available on request).

7. What the proposed options mandate would add

The mandate-expansion thread’s worked example is a $1M cash-secured put at a $1,150 strike, ~36% below the July-14 price. Here is what both sides of that trade look like against this portfolio:

  • The premium side: ~$70k per year per $1M of notional at the proposal’s ~7% figure.

  • The assignment side: at $1,150, the modeled portfolio is already down $16.8M (−22.7%). Assignment then converts $1M of stablecoins, the part of the portfolio doing all the defensive work in sections 3–5, into additional ETH exposure at that level. Per $1M of notional, that is $1M of added bottom-of-market exposure, and it scales linearly with program size.

Whether that trade is worth ~$70k a year per million is exactly the decision in front of the DAO; the numbers above are meant to make it concrete. The numbers can be re-run against the exact strikes, sizes and collateral rules once the mandate parameters are set.

The combined picture

  1. Two parts, on target. 61/39 at the June report against a 60/40 policy, and only the smaller part diversifies: the three staking positions move as one.

  2. Protocol risk is genuinely spread (exposure equivalent to ~5.5 equal-sized protocols, nothing over the 30% cap), but the two big staking venues are the same size, and a moderate single-protocol event comes to ~9% of the portfolio whichever one it hits.

  3. The volatility follows from the composition. A 20%+ drawdown year is near-certain within the models (98–99% of paths), the 1-in-20 year is −41% to −51%, and a literal repeat of the current bear’s core leg is −34%.

  4. The stablecoins are the whole defense. In crisis conditions, $12.4M of them converts to cash within a week through order books (more via the redemption paths not modeled here); the DAI position is roughly a week behind the USDC.

  5. The floor is model-sensitive. The probability of ending the year below the IPS’s $49.34M figure spans 12–39% across two defensible models, and anything sized off that floor inherits the spread.

An ETH-ecosystem endowment holding staked ETH is a coherent strategy: the book sits on its adopted target, and the new IPS explicitly accepts market risk in exchange for yield. What the numbers add is a measure of how much risk that acceptance currently amounts to. If the community finds these estimates useful, three quantities are worth tracking on a standing basis:

  • Floor-coverage probability (today: 61–88% across the two models), re-runnable monthly against the report book

  • Largest-protocol share vs the 30% cap (today: StakeWise, 25.3%)

  • Stablecoins sellable within a crisis week (today: $12.4M through order books, out of $26.9M in stablecoins)

Method & limitations

  • Every headline figure is a model output, conditional on (a) the published report book being complete and correctly attributed and (b) the stated parameter choices (participation rates, freeze/recovery/haircut assumptions, correlation windows, seeds). Different but defensible choices move the results materially, so read the figures to the nearest few percent rather than to the decimal.

  • Positions per the June 2026 report (period end 2026-06-30, open positions only; positions closed during June are excluded); prices as of 2026-07-14. July flows are not reflected. Coverage >99.9% of the open book; a ~$47k ETH balance and dust are excluded.

  • Liquidity model: fixed participation-of-ADV with venue-depth slippage, order books only. Protocol redemption paths (Sky, LST withdrawal queues) are NOT modeled and materially improve real-world exits. ETHx (~$18M) lacks usable venue-volume data and is excluded from the liquidity model only; DAI and weETH use aggregated market data rather than order-flow volumes.

  • Daily closes are aligned to a single day-boundary convention across price sources; a cross-source alignment defect found during internal verification on 2026-07-14 was fixed, and every scenario was re-run before this version.

  • Correlations from 365 days of daily returns; crisis regime from a 5-year window. Monte Carlo: two models, 25,000 paths each, seeded and reproducible. GBM (zero drift) and a block bootstrap of the trailing-365-day returns (which inherits that bear year’s drift). Both describe the spread of simulated outcomes; the gap between them is reported as model risk.

  • The historical replay uses actual per-asset price paths; the March-2023 depeg event is excluded for lack of asset history (noted in section 6) rather than approximated.

  • The protocol-incident severity (80% recovery, 20% haircut) is an illustrative assumption; how likely such an incident is at any given venue sits outside the model.

  • Snapshot-based; no protocol-code audit; nothing here is investment advice or bears on any custody or governance question before the DAO.

Analysis produced with Sentralis, a cryptocurrency portfolio risk- and scenario-analysis tool. For free teardowns for public treasuries contact riskanalysis@sentralis.io