dYdX Treasury SubDAO — Independent Risk Teardown

Everything below is built from public data: the July 2026 report’s position table, on-chain balances, delegations and staking parameters read on August 17, and the governance record (data basis and limitations at the end). If anything in the data basis is wrong, say so and it will be corrected.

TL;DR:

1. The July report’s $16.04M book comes to about $14.85M at August 17 prices. Same positions, same units; the whole difference is DYDX repricing from the report’s $0.1137 mark to ~$0.1006.
2. The book changed shape in July, and mostly in the direction these numbers favor. The $3.5M USDY deployment took the own-token share from 77% of the June book to 64% of the July one. Both sides of that trade are priced in §1 and §3.
3. Market depth under the DYDX position collapsed while that was happening. The venue-limited daily volume we track fell from ~23M to ~5M DYDX over five weeks, and all-venue aggregates show a similar decline. The staked 90.0M-token position now stands near 18 times daily venue volume (about 4 times in mid-July). After the chain’s 21-day unbonding, a modeled full market exit runs ~260 days under normal assumptions and multiples of that in crisis; §3 has the range and what shrinks on deeper volume figures (the day-counts) and what does not (the 21 days).
4. The modeled 1-in-20 year is −52% to −58% (−$7.7M to −$8.7M) depending on model choice, narrower than the −59% to −67% band computed for the June book: the USDY share does that work. The trailing-year bootstrap still puts the median year at −49%, and a >20% drawdown occurs in essentially every simulated path under both models.
5. Cash convertible within a week fell to ~$2.3M, from ~$4.7M in June: $2.4M of USDC became USDY, which is redeemable at NAV through Ondo for eligible holders but is not a 7-day order-book exit at this size on our numbers. The two quantities suggested for tracking in July moved in opposite directions.
6. Two results cut the other way: replaying the Oct-2025→Jul-2026 bear on today’s book, the USDY position gains 5.7% while DYDX loses 76%, which is the diversification doing exactly what it was bought for; and the FTX-collapse replay still marks this book up ~9%.

Snapshot: 2026-08-17 prices; position units from the July 2026 report (July-31 basis); module balances, delegations and staking parameters read via chain REST on August 17. Covered value: >99.9% of the report’s stated total ($52 of ETH/OSMO dust excluded).

Which treasury this is about

The analysis runs the reported book (Book A) as primary and the combined SubDAO-controlled view (Book A′ = A + Buyback Account) where the buyback question makes it relevant.

1. Repricing and the July restructuring

Report units, repriced at August 17:

Two things happened to this book in July. It got smaller, mostly from the DYDX decline (−23.7% on the month, another −11% since the report’s mark). And it got materially less concentrated: the $3.5M USDY deployment under the #372 mandate moved the composition from 77% DYDX / 23% stablecoins at the June report’s marks to 64% / 22% / 14% at the July report’s, which is 61% DYDX at today’s prices. The DYDX token count itself rose, from 88.10M to 90.04M, on buyback inflows (§4).

2. Simulations

Two models, 25,000 one-year paths each, same seed. Zero-drift GBM answers “what does this volatility structure do with no view on direction”; the empirical bootstrap resamples the trailing 365 days and answers “what if next year statistically resembles the last one,” which for DYDX makes it the more negative model by design.

The July numbers for the June book were −59% to −67%. The band moved to −52% to −58% for the same 1-in-20 measure, and the difference is composition, not calmer markets: the DYDX that drives every tail is a smaller share of the book. On the combined SubDAO-controlled view (Book A′, $17.06M, 64.8% DYDX) the band widens back to −55% to −62%.

Two replays bracket the simulations. The realized Oct-2025→Jul-2026 bear applied to today’s composition comes to −45% (−$6.7M), with DYDX at −76% and the USDY position up 5.7% over the same window on real price data: that is the decorrelation the July deployment was buying, visible in a replayed drawdown. The FTX-collapse window still marks the book up about 9% (DYDX rallied as trading moved to decentralized venues; USDY did not exist in 2022 and is held flat, so treat it as order-of-magnitude).

3. Liquidity

The staked DYDX sits behind the chain’s 21-day unbonding (unbonding_time read on-chain; some help pages still say 30 days). What changed since July is the market under it. The venue-limited order-book volume we track averaged 22.9M DYDX/day ($2.8M) in mid-July and 5.0M/day ($0.5M) now; all-venue aggregates run several times deeper but show the same ~5× decline over the period. Against the narrow basis, the position went from roughly 4 days of all trading to roughly 18.

Modeled as fixed participation in daily volume, on the narrow basis:

On deeper volume figures the market-exit days shrink proportionally; the 21 days do not, and neither does the direction of the volume trend. The same arithmetic on the community treasury module (77.25M DYDX, unstaked): ~205 days normal, ~1,230 crisis. As in June, order books are not the only path: OTC blocks exist at treasury scale, the USDY is redeemable at NAV through Ondo for eligible holders rather than sold on-market (our 52-day/310-day order-book figures for it overstate the practical constraint for exactly that reason), and nothing in the SubDAO’s obligations forces an exit on any clock. The day-counts price what a full conversion would cost if governance ever wanted one.

Two single-point scenarios for the positions that concentrate infrastructure: a staking freeze on the book’s staked DYDX (temporary freeze, 80% recovery, 20% haircut on the affected portion; parameter choices, not probabilities) produces an estimated loss of $3.3M (22%); the same parameterization on the Ondo position produces $1.3M (8.5%). The USDY deployment traded market concentration for a new single-issuer exposure; both sides of that trade are now on the table. In fairness, the staking spread stays clean: ~28 validators, the largest near 6.9%.

4. Buybacks

Under #313’s split (75% of net protocol fees to DYDX purchases, 15% to stakers, 5% to this treasury’s stablecoins, 5% to MegaVault), July bought 895,295 DYDX for $111,626 at an average $0.1247, open-market only. Cumulative since April 2025: ~22.3M DYDX for ~$6.7M, a blended ~$0.30 per token against today’s ~$0.10.

Where those tokens sit takes one reconciliation, because the monthly report and the chain answer it differently. The H1 update puts buyback-sourced staked DYDX at 22.2M by end-June. On-chain, the segregated Buyback Account holds 19.97M staked (unchanged between our July 16 and August 17 reads), while the reported book’s own DYDX count rose 88.10M → 90.04M during July against only 0.90M of July purchases. The reading that fits both: bought-back DYDX ends up staked partly in the segregated account and partly inside the reported book, with transfers landing periodically rather than monthly. That split is our inference from the report plus chain data, not a kpk statement; if kpk states the mechanics differently, this section updates.

The direction is not in question, only the address: every buyback dollar becomes staked DYDX under the same 21-day gate, ~10.7M tokens a year at July’s pace. There is a case for the program and it deserves equal weight: the DAO chose it at 59% approval, and accumulation at $0.125 is cheap against the $0.30 blended history if the protocol’s fee base recovers. What the July numbers add is that the stablecoin side of the split (5%) received ~$7.4k while the buyback side converted $112k of fee revenue into the book’s least liquid asset, over the same weeks its market depth was collapsing.

5. Funding

July’s buyback spend implies net protocol fees of about $149k for the month, ~$1.8M annualized, down from ~$256k/month implied in June. No fee holiday was active in July (the series ended with #389’s rejection), so this decline is volume, not fee policy; August’s #393 then wound down fifteen low-volume markets. The treasury’s own staking income was $10.1k for July (1.15% APY on the staked position).

The structure carries over from the June analysis: the treasury is denominated mostly in the token of the protocol whose fees fund it, and the mechanism recycling those fees buys more of the same token. The USDY deployment is the first material weight on the other side of that structure. None of this is a solvency claim.

6. Consolidated view

1. The simulated bad year is still the big number: −$7.7M to −$8.7M on the reported book, −$9.4M to −$10.6M on the combined view. The replayed 2025-26 bear (−$6.7M) sits just below the band.
2. Liquidity, not composition, deteriorated this period: week-scale cash fell from ~$4.7M to ~$2.3M by conversion into USDY, and the DYDX exit runway roughly quadrupled as market volume thinned. Whether USDY’s NAV-redemption path substitutes for the order book at need is now a material operational question, and one kpk can answer better than an outside model.
3. The counterweights are real and one of them is new: the bear replay shows the USDY position appreciating through DYDX’s −76%, the validator spread stays clean, an FTX-style CEX crisis was historically good for this book, and no obligation forces any exit on any clock.

None of this says the current structure is wrong. Token holders chose the buyback bet deliberately, and the July restructuring spent $3.5M making the book harder to break. The two quantities worth tracking on a standing basis are unchanged from the July forum reply on the June book, with updated readings: the own-token share of SubDAO-controlled assets (64.8% today, falling on the USDY move, rising ~10.7M tokens/year under the current split) and the share of the book convertible within a week ($2.3M today, down on the same move).

Method & limitations

• Every headline figure is a model output under stated parameters (participation rates, regime haircuts, freeze/recovery assumptions, covariance windows). Different defensible choices move results materially; treat them as orders of magnitude.
• Position units are the July report’s; prices are August-17 daily data; on-chain reads (balances, delegations, unbonding_time) are August 17. Coverage >99.9% of the report’s stated total. Report totals are compared as stated; we do not re-derive its internal marks.
• Simulations: 25,000 paths, one-year horizon, seeded and reproducible; zero-drift GBM plus an empirical bootstrap of the trailing 365 days, which inherits that year’s direction by design. USDY enters with its real price history (near-zero volatility) and zero expected return, so its ~4% yield is not credited in the simulations.
• Liquidity model: fixed participation in daily order-book volume, venue-limited basis (~5.0M DYDX/day 30-day average at the read date), with the proportional adjustment for deeper bases stated in §3. It ignores OTC and ignores Ondo’s NAV redemption path for USDY; both would shorten realistic exits. At ~18× daily volume, slippage estimates are unreliable extrapolations, so we publish day-counts and omit slippage dollars.
• The staking-freeze and Ondo-freeze scenarios are parameter choices; no probability estimate and no claim about either system’s actual security is implied.
• The FTX replay approximates 39% of the book (USDY pre-dates its own history, USDC held flat) and is order-of-magnitude. The bear replay uses full window data for all three assets.
• The buyback staking split (19.97M segregated + ~2.2M in-book) is our reconciliation of the H1 report against on-chain delegations, stated as inference in §4; corrections welcome and will be incorporated.
• Nothing here is investment advice, and we take no position on any proposal or vote.