Entropy is pleased to present the fourth Arbitrum Treasury Management Quarterly Report. The attached report provides an overview of the DAO’s financial performance and treasury management portfolio, offering detailed insights into income, expenses, runway, RWA holdings, ETH and ETH-correlated positions, stablecoin deployments, and more.
If there are specific charts, data points, or other adjustments you’d like to see included, please reach out to @BricksIntern on Telegram.
Disclosure first: I’m part of the team behind Sentralis, a cryptocurrency portfolio risk and scenario analysis solution, which produced this analysis. Nobody at Arbitrum, the Foundation, Entropy Advisors, or the ATMC asked or paid for it, and none of it is advice. It takes no position on any governance question, and it is not a critique of the managers: the TM portfolio’s operational reporting is among the best published by any DAO, and this analysis exists because that reporting made it possible.
Some notes about the data background: the position feed behind Entropy’s own arbdata.com dashboards (August 11 snapshot), the May ATMC report, the June and July monthly updates, and the Q4-2025 Investment Policy Statement. The July update and the executed Foundation drawdown changed the book materially (about $19M of outflows, two RWA positions closed, the GMX vault divested), so the composition analyzed here is several weeks fresher than the last published position tables. If anything in the data basis is wrong, please let me know and it will be corrected.
Scope: this covers the $89.7M Treasury Management Portfolio only, not the DAO’s ARB holdings (the May report’s “Tokens in Treasury” line, $274M).
TL;DR:
Snapshot date: 2026-08-11 (positions from the arbdata feed, same-day; prices at August 11 daily closes; ETH $1,870)
Covered value: $89,709,966 of the feed’s $89,719,673 total (99.99%; excluded: ~$10K of dust across seven residual lines). The feed carries no ARB row; the May report’s ~1% ARB line is either disposed or outside the tracked segments, and at that size it does not move any number below.
Reproducibility: Monte Carlo runs are seeded (25,000 paths, seed 42, one-year horizon, 365-day covariance, both a zero-drift GBM and an empirical bootstrap).
Two delegates asked pointed risk questions in the 6,000-ETH transfer thread in April: @OliverBuilds asked what level of risk is being assumed, including exposure limits, liquidity constraints, and drawdown tolerance (#19), and the L2BEAT team asked for continued scrutiny of risk frameworks and exposure limits (#17, #24). The monthly reports publish composition and yield-against-benchmark, and they do it well, but no quantitative tail, liquidity-stress, or scenario figure has appeared in the public record. The July update also executed the Continued Funding drawdown (1,740 ETH plus $15.8M of RWAs to the Foundation, with ~$300K of USTBL still to move), which reshaped the book enough that any earlier intuition about its composition is stale. This is an outsider’s attempt to answer the questions that were actually asked, with the DAO’s own policy document as the measuring stick.
From the manager’s own feed, August 11:
| Segment | Position | Value | Share |
|---|---|---|---|
| ETH 43.6% | eETH under covered calls (6,750) | $12.54M | 14.0% |
| eETH, liquid staking (6,127) | $11.38M | 12.7% | |
| weETH (3,228) | $6.66M | 7.4% | |
| wstETH (3,668) | $8.54M | 9.5% | |
| Stablecoins 30.2% | Spark sUSDC | $11.92M | 13.3% |
| USDai | $9.13M | 10.2% | |
| Maple syrupUSDC | $5.88M | 6.6% | |
| Morpho Gauntlet Core (residual) | $0.12M | 0.1% | |
| RWA 26.2% | Spiko USTBL | $10.63M | 11.9% |
| Franklin BENJI | $6.96M | 7.8% | |
| WisdomTree WTGXX | $5.95M | 6.6% |
The effective number of independent positions is about 5.5 at protocol level. Three structural observations first:
The July drawdown was funded by closing exactly the two RWA instruments with discretionary redemption paths (BUIDL, a BVI private fund with transfer-agent settlement cycles, and USDY, redeemable by wire through an SPV). What remains is the cleanest wrapper set the tMMF segment has had: two SEC-registered government money-market funds (BENJI, WTGXX, statutorily outside the SEC’s mandatory liquidity-fee regime) and one EU VNAV fund (USTBL, which sits outside MMFR Article 34’s fee-and-gate regime entirely, and whose NAV accumulates rather than distributes). Whether that was a deliberate liquidity-quality choice or a convenience of settlement, the effect on redemption risk was positive.
Second, ether.fi is now the portfolio’s dominant protocol dependency at 34.1%, up from under 7% at the May report (when only the weETH position existed), because the June deployment moved ~13K formerly idle ETH into eETH and the July drawdown then shrank everything else. IPS §7.1.1 caps any single protocol at 35%.
Third, the covered-call overlay (6,750 eETH of notional, re-written in late July) is not priceable from public data: strikes and tenor are unpublished. In downside scenarios the calls expire worthless and the numbers below are unchanged; in rallies they cap the upside above the strike and the +94% arithmetic in §3 would arrive sooner in practice; the premium income is real yield the models below ignore.
The IPS requires 30% of the portfolio withdrawable within 24 hours, 60% within 7 days, 80% within 30 days, and its §7.2.1 sets the counting rules (tMMFs at T+0 to T+2 count as 24-hour liquidity; Lido counts as 24-hour up to DEX sizing; other LSTs and LRTs “require individual liquidity assessment”). No public computation of this ladder against the actual book exists. Using those counting rules plus each instrument’s documented redemption mechanics:
| Rung | Requirement | Base counting | Conservative counting |
|---|---|---|---|
| 24 hours | ≥30% | 56.4% | 39.7% |
| 7 days | ≥60% | 65.9% | 49.2% |
| 30 days | ≥80% | 86.0% | 86.0% |
The 24-hour rung holds on any counting, and it is carried by the tMMFs plus Spark’s sUSDC (atomic redemption per its docs). The segment doing that work is also the one the Foundation drawdown drains first.
The 7-day rung is where the assumptions decide the answer. The base column counts USDai (documented near-instant redemption, though current docs are ambiguous about whether direct redemption is now restricted to approved market makers) and syrupUSDC (normally instant per Maple’s docs) inside the week. The conservative column takes both at their documented worst cases: syrupUSDC’s stated maximum is 30 days, and USDai’s fallback if direct redemption is gated is a thin secondary market. On those assumptions the rung fails at 49.2%. The distance between 65.9% and 49.2% is exactly the $15.0M in those two OTC-lending strategies, so the rung’s pass or fail is a question about two specific redemption mechanisms, not about the market.
The 30-day rung holds at 86% provided the ether.fi withdrawal queue behaves as documented (about 10 days typical), and it holds without needing the covered-call tranche at all. The documented tail matters, though: ether.fi withdrawals ride Ethereum’s validator exit queue when the protocol’s buffer is exhausted, and that path has stretched to weeks-to-months in past congestion. In that state, 34% of the book sits beyond every rung, and the on-market alternative is thin for a position this size (tracked spot volume in weETH is under $2M a day against a $30.6M position).
Even when the rungs pass, they pass on redemption mechanics, not order books. On tracked spot volume alone, at crisis participation assumptions, more than half this book cannot reach cash inside a week. Redemption paths are the plan; the plan’s dependencies are the queue lengths above.
The IPS gives ETH a 30% long-term target and a 60% maximum, with quantified triggers (§6.5.1: more than 5 points over the cap at quarter end forces a rebalance to 10 points under the cap within 14 days; more than 10 points intra-month forces it as soon as possible). Computed on today’s book, with non-ETH values held flat:
The asymmetry is presumably intentional (the policy text itself argues for wide bands and upside flexibility), but it has a quantitative consequence: rebalancing pressure exists only in rallies, and the portfolio’s risk in crashes is governed by nothing except the allocation it happens to have.
The limit with materially less headroom is the single-protocol cap. At 34.1%, ether.fi crosses 35% at about a +5% ETH move, because the non-ETH 56% of the book stays flat while the whole ETH side, ether.fi included, appreciates. Two precision notes, because this audience includes the people who wrote the document: §7.1.1 places its “ETH Staking as a category” exception under the 45% single-strategy-type cap, not under the 35% single-protocol line, and a general carve-out for opportune strategic deals with OAT approval applies to everything. So whether a 36% ether.fi share is a breach or an exception is a committee reading. What the models can add is the size of the event the cap is guarding against: a temporary ether.fi freeze with 80% recovery and a 20% haircut on the affected portion comes to an estimated $11.0M, or 12.3% of the portfolio. The same event at Spark is $4.3M; at Spiko, $3.8M.
The dual mandate turns every one of these numbers into two. Revenue is ETH-denominated (§6.1.1’s hold-ETH baseline) while a large portion of expenses are dollar-denominated (§6.4.2). A −40% ETH crash costs the book −17.4% in dollars and gains +37.6% in ETH terms; a +40% rally is the reverse. Neither number is a manager outcome; both are consequences of an allocation the DAO chose. Any evaluation that quotes only one currency is implicitly taking a side of that mandate.
The seeded one-year simulation on the current book puts the 95th-percentile loss at −28.9% under zero-drift GBM and −36.3% under an empirical bootstrap of the last year’s daily moves ($25.9M to $32.6M). The spread between those two numbers is itself information: the bootstrap resamples a year in which ETH fell by more than half, so it inherits that drift, while the GBM is drift-free by construction. A 1-in-20 year for this book lands somewhere in that band, and anyone quoting a single VaR number for it has made the model choice silently. Median simulated max drawdown is 27% to 31%; the probability of at least a 20% drawdown inside the year is 83% to 93%.
Replaying the 2025-26 tariff bear on today’s book produces −27.9% (−$25.0M); the October 10 flash crash, −5.4%. Both land almost entirely on the ETH side, and the RWA and stablecoin legs mark flat through both.
Correlation stress produced this cycle’s non-result: crisis-regime correlations change the book’s modeled one-day VaR by $772 on $89.7M (from $2.7324M to $2.7332M), and the diversification ratio does not move at three decimal places. The reason is structural. The volatile half of this book is one asset wearing four wrappers (eETH, weETH, wstETH all track ETH at correlations near 1), and the other half is dollar instruments with almost no variance to re-correlate. There is no diversification for a crisis to destroy, so correlation stress has nothing to bite on. The book’s tail is composition, not correlation, and it is already fully present in the numbers above.
The IPS names a long-term target: 30% ETH. Scaling today’s book to that target (non-ETH strategies pro-rata, same total) and re-running both models gives a 1-in-20 year of −20.6% to −25.1%, against −28.9% to −36.3% for the current allocation, and cuts the probability of a 20%+ drawdown under GBM from 83% to 48%. The ceiling arithmetic moves the same way: at 30% ETH, reaching the 60% maximum takes about +250%.
That is not a recommendation to move; the glide path, its pace, and the ETH-maximalist revenue mandate are the DAO’s stated policy, and holding 43.6% is within it. It is the measured size of the choice: roughly a quarter to a third of the simulated tail is the premium currently paid for the extra 13.6 points of ETH exposure.
Analysis produced with Sentralis, a crypto portfolio risk and scenario analysis solution suited for beginner HODLers and semi-professional investors alike. If the ATMC or any delegate wants these scenarios re-run under their own assumptions (different haircuts, participation rates, allocation variants), we will do it at no cost: contact@sentralis.io
Entropy is pleased to present the 8th Arbitrum Treasury Management Monthly Report. The report combines the months of July and August, as Entropy worked through certain data updates during the period. The attached report provides an overview of the DAO’s Treasury Management portfolio and offers detailed insights into RWA holdings, ETH and ETH-correlated positions, stablecoin deployments, and more. In response to community feedback, we’ve added new risk metrics to the report structure.
If there are specific charts, data points, or other adjustments you’d like to see included, please reach out to @BricksIntern on Telegram.