StakeWise osETH Boost — product breakdown and four questions

Scope: Ethereum mainnet · StakeWise Boost (boost.stakewise.io) Date: 2026-08-05 · Block stamp: all on-chain reads at block 25,688,659 unless stated Companion docs: wstETH yield options on Ethereum (the wstETH loop this is compared against) · oseth-looping-strategy-research.md (research repo, not checked in here — the manual version of this exact trade, studied 2026-07-06)

Data provenance note: boost.stakewise.io is a dashboard/simulator front-end (page title: “StakeWise Boost Mock”) that loads a daily-generated dataset (generatedAt 2026-08-05T12:15Z, subgraph block 25,688,663) from the StakeWise subgraph, the Aave GraphQL API and validatorqueue.com. Its data is genuine — every load-bearing value it carries was cross-checked on-chain at block 25,688,659 and matched (§6).

Answer

Verdict: not usable at our size, and currently not worth using at any size on the vaults that have capacity. Re-verified at block 25,688,768.

For NOCA the question is academic at size: remaining Boost capacity is 830 osETH ($1.7M) of new 14x deposits, because the Aave osETH supply cap (123,000) is 90.9% full and every deposited osETH consumes ~13.6 osETH of cap. §1.4

Why, in four points:

1. Leverage is high — and fixed. ~14x on the flagship vaults (Genesis, Chorus One; measured 13.6x on the largest live position), ~6x on default vaults. You don’t choose it: it falls out of Aave’s 93% eMode LTV × the vault’s osETH mint LTV (99.99% / 90%, verified on-chain). Consequence: every 1bp move in the Aave WETH borrow rate moves Boost APY by ~13bp.

2. Deposit capacity is 830 osETH ($1.7M). The Aave osETH supply cap (123,000) is 90.9% full (111,752 supplied, headroom 11,248), and each deposited osETH consumes ~13.6 osETH of cap because the strategy supplies your osETH plus the flash-minted collateral. Via a 6x vault the ceiling is 1,845 osETH ($3.7M). The headroom is shared and first-come — existing positions compound into it — and only grows if Aave governance raises the cap.

3. Outcome APY: a 13x-leveraged bet on the vault-minus-borrow spread — window-dependent, spot-negative. Boost APY is all-in (it replaces osETH APY, it does not stack on it):

   • Spot: Genesis 2.03% = −7bp vs holding osETH (2.11%), −16bp vs holding wstETH (2.20%); past-week realized 1.60% (−51bp). But rolling 30d = 3.93% and 90d = 4.31% beat holding by ~180–220bp, while 180d = −0.47% (the Aave rate spike through 13x leverage). The window dispersion is the product: ±30bp of average spread = ±400bp of APY — see §2.1.
   • Breakeven average spread vs just holding osETH: ~+11bp (breakeven borrow ≈2.13% against Genesis’s 2.24% vault APY; the market is at 2.14%) — a coin-flip around zero edge today, same root cause as our wstETH loop rejection (staking yield ≈ borrow cost).
   • NodeSet 4.62% (+251bp vs osETH / +242bp vs wstETH) is real but sits in a 24k-ETH vault: capacity-capped at ~$1.7M anyway, and a large deposit would multiply itself 14x into the vault, spend ~43 days in the activation queue, and dilute the very APY it came for.
   • The Boost page’s Future simulation showing +45 ETH/yr on 1,000 osETH (4.2% year one) is arithmetically correct but close to a best case: it assumes network-average vault performance, the borrow rate frozen at 2.14% for 12 months, and a negotiated 40% fee rebate. Re-run with Genesis’s trailing APY, the realized 365-day average borrow rate and no rebate, year one is ≈ −2.4% — see the scenario table in §2.1.

4. Withdrawal concerns and the main risk chain. Exit is queue-bound (validator exit queue: ~0 days today, weeks in stress) and during the queue you keep paying borrow interest on 13.6x debt while the exiting ETH earns nothing (~28bp of deposit per queued week at current rates). The forced unboost at 94.5% LTV fires precisely in the bad regime — sustained borrow spike → LTV drifts up → auto-exit into a long queue at peak rates. And you receive osETH, not ETH: NAV redemption is disabled on-chain and DEX depth is only ~$20M, so converting size is a second, harder exit. What is not a concern: liquidation from a market depeg (Aave prices osETH at redemption rate), DEX slippage in the unwind (there is none), and Aave WETH liquidity (exit repays debt, doesn’t withdraw).

Also worth knowing (beyond the four points): the osToken 5% fee applies across the whole ~14x notional (~155bp/yr drag — the real cost of the packaging); the position adds StakeWise (operator + osToken controller + strategy contracts) to the counterparty surface on top of Aave; and the re-entry trigger is the same as for our wstETH Path A — a staking-to-borrow spread ≥75bp, plus (here) fresh Aave osETH cap headroom. Monitor allocatorMaxBoostApy vs osETH APY on the StakeWise subgraph: Boost is only interesting when that difference is meaningfully positive on a vault that can take the size.

1. Product breakdown

1.1 What Boost is

StakeWise Boost (live since February 2025) is a one-click leveraged-staking strategy built into StakeWise V3 vaults. Per deposited osETH, in a single transaction the strategy:

1. Takes the user’s osETH and flash-mints additional osETH (OsTokenFlashLoans, no fee);
2. Supplies the combined osETH to Aave v3 core as eMode 1 collateral;
3. Borrows WETH at up to 93% LTV;
4. Stakes the WETH in the user’s chosen StakeWise vault;
5. Mints osETH against the new stake (up to the vault’s osETH mint LTV) and repays the flash mint.

No iterative looping and no DEX swap anywhere — the “loop” is collapsed into one flash-mint transaction. Each user gets a dedicated StrategyProxy (EIP-1167 minimal proxy of 0x2CbE7Ba7f14ac24F3AA6AE2e1A8159670C9C7b75) that holds the Aave position.

Leverage is not chosen by the user. It is fixed by two parameters: Aave eMode LTV (93%) and the vault’s osETH mint LTV:

Verified empirically: the largest live position deposited 2,007 osETH and holds 27,291 osETH supplied / 27,350 WETH borrowed — 13.6x, at 93.1% LTV.

1.2 Fees

• No Boost-specific fee. StakeWise charges nothing for the strategy itself.
• osToken fee: 5% of staking rewards on all outstanding osETH (OsTokenVaultController.feePercent() = 500). Because the strategy mints osETH against ~93–100% of the entire levered notional, this fee applies ~14x over per deposited osETH — a ~155bp/yr drag at current rates. This is the material cost of the packaging.
• Vault operator fee (e.g. Genesis 5%) is already inside the quoted vault APY. Some vaults rebate part of their fee to boosters (a “vault fee rebate” — visible as a NodeSet/CCM boost premium in §1.3).
• Aave WETH borrow interest: 2.139% APY at the stamp block (2.115% APR, utilisation 82.8%).

1.3 Current size and rates

Per-vault Boost APY (StakeWise subgraph allocatorMaxBoostApy, i.e. instantaneous at max leverage):

The two vaults with real size are both below plain osETH today. The attractive numbers sit in small vaults whose extra APY comes from higher operator performance/fee rebates — and whose TVL a large deposit would immediately dilute (a Boost deposit multiplies itself ~14x into the vault — even a fraction of our size would swamp NodeSet’s 24k-ETH TVL and sit in the 43-day activation queue, §5).

Realized Genesis Boost APY, recomputed from the site’s own daily dataset using its published formula (14x):

The −0.47% 180-day window is the Aave WETH rate spike our wstETH study documented (utilisation hit 98.48% in that quarter) passing through 13x leverage. StakeWise’s own case study reports 4.415% annualized since inception (Feb 2025, published Apr 2025 — when staking yield was ~3%+ and the spread real).

1.4 Capacity — the binding constraint

Aave osETH supply cap 123,000, supplied 111,752 → headroom ~11,248 osETH. Since each deposited osETH consumes ~13.6 osETH of cap (own + flash-minted collateral), new 14x Boost deposits max out at ~830 osETH ≈ $1.7M (~1,850 osETH ≈ $3.7M if routed via a 6x vault). Existing positions keep compounding into the cap as well. This is not an edge case — it is the product’s advertised “Remaining Boost capacity” stat and its own first listed risk.

2. Q1 — What “Boost APY” means

Boost APY is the total APY of the boosted position, denominated on the original osETH deposit. It is not additive to osETH APY — it replaces it. Once deposited, your osETH is Aave collateral inside the strategy; you no longer hold an appreciating osETH balance. Your return is the Boost APY.

Three independent confirmations:

1. The site’s own formula (front-end source, function calculateBoostComponents): BoostAPY = (base vault return on deposit + vault return on borrowed stake − osToken fees on all outstanding osETH − Aave borrow interest) / original deposit. The base return on the deposit itself is inside the numerator — it’s an all-in figure.
2. StakeWise’s case study: “1–3 percentage points above the usual staking rate, for a total APY of approximately 4–6%.”
3. The numbers: Genesis Boost APY 2.03% < osETH APY 2.11%. An “APY on top” would make no sense below the base.

Algebraically, per 1 osETH deposited at leverage L: BoostAPY = L·v − (L−1)·b − L·ltv·(v_os/0.95)·5% where v = vault APY, b = Aave WETH borrow APY, v_os = osETH APY. At Genesis today: 14(2.24%) − 13(2.14%) − 1.55% ≈ 2.0%. Breakeven borrow rate vs just holding osETH: ≈2.13% — the market is at 2.14%. Boost is a coin-flip around zero edge right now.

Is osETH APY = wstETH APY? To first order yes — both are net ETH staking yield with a ~10% total fee take (Lido: 10% of gross; osETH: vault fee then 5% osToken fee, ≈9.75% of gross for Genesis). Measured same-day: osETH 2.107% vs wstETH 2.197% — osETH runs ~5–10bp lower on the big vaults; smaller high-MEV vaults (NodeSet 2.43%, CCM 2.62%) beat Lido but carry single-operator concentration.

2.1 Verifying the Boost page’s own numbers — and what actually drives the outcome

“Boost APY (Past week)” (About tab). The page computes it via calculateHeadlineBoostApy over the trailing 7 days of vault, osETH and borrow APYs. Reproduced exactly from its dataset (2026-08-05): 1.60%, vs 2.11% for holding osETH the same week. Per 1,000 osETH (1,076 ETH): +24.1 ETH/yr base vault return, +313.6 on the 13,994 borrowed ETH, −303.9 borrow interest, −16.7 osToken fees = +17.2 ETH/yr vs +22.7 just holding. The levered leg earned only a +7bp spread (+9.7 ETH/yr) — less than the osToken fee on the ~14,000 osETH the strategy keeps outstanding. That is the whole mechanism by which “boosted” can pay less than holding.

Future simulation (Projections tab) — 1,000 osETH into Genesis with a 40% vault-fee rebate (the “+45.00 ETH / 4.55% headline” screen). Every waterfall bar reproduces to the cent, so the arithmetic is right; the content is in the embedded assumptions:

What drives the final outcome, ranked by weight:

1. Average vault-minus-borrow spread × 13 — the whole trade. Breakeven vs holding osETH ≈ +11bp of average spread; ±30bp of spread = ±400bp of APY. A +50bp move in the average borrow rate over the holding period = −650bp.
2. Vault performance vs network average — every 10bp of vault APY = ~±130bp at 14x. The simulator credits Genesis with network-average ETH.STORE going forward; it has trailed that by ~12bp net.
3. osToken fee drag: −165 to −174bp/yr, structural — 5% of gross staking rewards on the full ~14x notional. This is why a small positive spread still loses to holding.
4. Vault fee rebate, if negotiated: +70bp at 40% — the ingredient behind NodeSet/CCM screens showing 4.4–4.6%.
5. Queues — entry (43.1 days today): the extra ETH earns nothing while the debt accrues (~35bp year-one drag in the simulator’s own blend, worse for deposits large relative to vault TVL); exit: the same bleed in reverse, at whatever the queue is when you leave.

Scenario comparison — year-one outcome, 1,000 osETH (1,076 ETH) into Genesis at 14x:

The 40% rebate is worth +66bp at trailing rates — it softens the realistic case from −2.4% to −1.7% but cannot flip the sign; only the spread can. The “more realistic” column is not a constructed downside: it simply reuses what the market actually paid over the last 365 days (borrow 2.445% avg) and what Genesis actually earned (2.241%). The best case needs all three optimistic inputs at once to reach ~+200bp over holding; the downside needs only one borrow-rate episode — which happened within the last two quarters (utilisation 98.48%, ≈7.2% APR under the current curve).

3. Q2 — “Borrowing is disabled” vs Boost’s borrowing

The Aave UI statement and Boost’s operation concern opposite directions of the same market:

• What Aave disables: borrowing osETH from the pool. On-chain at block 25,688,659: borrowingEnabled = false, borrow cap = 1 (belt-and-braces), total osETH debt ≈ 0.0094 osETH (dust). Same deliberate governance choice we documented for wstETH (wstETH yield options study §2): nobody rationally borrows a yield-bearing LST — you’d pay interest while being short an appreciating asset — so a borrow market would be a liquidation surface with no offsetting revenue. The 0% supply APY on the Aave page is the same fact: no borrowers → nothing to pay suppliers.
• What Boost does: supplies osETH as collateral and borrows WETH. That is governed by collateral parameters, not the borrow switch:
  • Base (non-eMode) LTV = 0 — osETH gives no borrowing power outside eMode; liquidation threshold 75%.
  • eMode 1 “ETH correlated”: collateral bitmap includes reserve id 29 = osETH (bits {0, 1, 7, 9, 28, 29, 31}); LTV 93% / LT 95% / bonus 1%; borrowable bitmap = 1 → WETH only.

So osETH on Aave is configured to permit exactly one trade — the correlated leveraged-staking loop — and nothing else. Verified live: 397 Boost proxies hold 95,720 osETH of collateral against 95,967 WETH of debt.

One more piece makes the 93% LTV safe to run: the Aave oracle prices osETH at exactly the StakeWise redemption rate — oracle ratio 2012.69/1869.77 = 1.076438 vs convertToAssets(1e18) = 1.076439. A market depeg of osETH cannot move the LTV, identical to the wstETH exchange-rate-pricing finding.

4. Q3 — The looping, vs our wstETH loop (Path A)

Boost is the same trade as Path A in the wstETH yield options study §4 — long staking yield, short Aave WETH borrow, through the same eMode 1 of the same pool — differing in packaging:

The verdict transfers unchanged. Our Path A rejection was not about mechanics — it was that the staking-to-borrow spread is ~nothing, so leverage multiplies ~nothing. Boost is mechanically better packaging (no swap costs, no depeg-liquidation risk, atomic entry) but adds a 5%-of-gross fee across the whole levered notional, which at today’s spread eats more than the leverage earns on its flagship vaults. StakeWise’s own docs concede it: “avoid using Boost when its APY is consistently negative.” The July conclusion of oseth-looping-strategy-research.md (“leveraged ETH staking that earns a modest, rate-sensitive premium — not a high-yield play”) still stands, on a thinner spread (Genesis vault APY 2.53% → 2.24%).

What would change the conclusion

What would make Boost interesting: a high-MEV vault with fee rebate (NodeSet at 4.62%) and room under the Aave cap and a spread ≥75bp — the same trigger condition as Path A (§10 of the wstETH study). None hold today at our size.

5. Q4 — Withdrawal issues

Unboosting reverses the flash-mint: the strategy exits staked ETH from the vault, repays the WETH debt, and releases collateral. The issues, in order of how much they’d matter to us:

1. Exit is queue-bound, not instant. Repaying the debt requires real ETH from validator exits: vault exit queue → Ethereum exit queue → sweep delay. At the stamp date the network exit queue is ~empty (96 ETH, ~0 days), so an unboost today clears in days. But the queue is regime-dependent — the entry side currently sits at 43.1 days / 2.48M ETH, and exit queues extend exactly in the scenarios (rate spikes, depeg panics, slashing news) in which one wants out. This is the same correlation problem we flagged for Maple in the wstETH study: the exit lengthens precisely when the position bleeds.
2. Negative carry during the exit window. From unboost to claim, Aave interest keeps accruing on the full debt while the exiting ETH earns nothing after the validator leaves. At 13.6x, each week of queue at a 2.14% borrow rate costs ~28bp of the deposit. A long queue converts a mediocre exit into a materially negative one; StakeWise’s own timeline warns “Boost APY may be reduced during the sweep delay.”
3. Forced unboost at 94.5% LTV — at the worst moment. Anyone can (and the team does) trigger exits for positions at 94.5%. LTV only drifts up through sustained negative spread, so the mechanism structurally sells at the bottom: it exits you into the exit queue while borrow rates are spiked. Not a safety flaw — it is the anti-liquidation design working — but it means the tail scenario is “locked in the queue, paying 7% APR on 13.6x debt for weeks.”
4. You receive osETH, not ETH. Unboost returns your (appreciated) osETH collateral; getting to ETH is a second, separate problem: (a) redemption at NAV via OsTokenRedeemer only works when unbonded ETH is available and was found disabled on-chain in our July research; (b) burning osETH requires holding a matching vault stake position and again rides the unstake queue; (c) selling — osETH DEX liquidity is an order of magnitude thinner than wstETH’s (Fluid ETH-osETH $10.7M, Curve osETH/rETH $10.4M, Balancer ~$1.3M, vs wstETH’s ~$54M single-clip routability). Fine for hundreds of osETH, not thousands.
5. What is not a problem: Aave-side liquidity (exit repays WETH rather than withdrawing it; collateral withdrawal has no cap), DEX slippage inside the unboost itself (none — no market sell anywhere in the flow), and partial exits (supported, exitingPercent is per-position; live positions show partial exits in progress).

6. Sources and contracts

Contracts

APIs / documents (all accessed 2026-08-05)

Verification log (cast, block 25,688,659)

Unit convention: bp (basis points) for rate spreads and for gaps between two percentages · % for levels. (Companion docs write percentage-point gaps as pp; here they are expressed in bp: 1pp = 100bp.)