Steakhouse Prime USDC — "USDC Prime Instant" (Base)

TL;DR

• What it is: ERC-4626 MetaMorpho vault on Base (0xBEEFE94c…, steakUSDC), curated by Steakhouse Financial, supplying USDC into Morpho Blue markets — in practice the supply side of Coinbase’s BTC/ETH-backed retail loan product. ~$229M TVL.
• ⚠ Exit trigger fired: the curator switched on the vault fee on 2026-06-10 — instant, no timelock, the exact action Risk #4 said to exit on. Yield-only, no principal risk, and net yield still rose as utilization climbed. Needs a hold-or-exit call (Risk #4).
• Yield source: borrower interest on overcollateralized cbBTC loans; no emissions in NAV.
• Biggest risk: the vault now sits almost entirely in a single cbBTC market whose oracle prices cbBTC via Chainlink BTC/USD — a cbBTC-specific depeg is invisible on-chain and would socialize as bad debt to suppliers (Risk #1).
• For detail: Staking Rewards profile on sibling Ethereum vault (link); Morpho/Coinbase integration (Morpho post).

How it works

USDC deposits mint steakUSDC (ERC-4626, non-rebasing). The vault (0xBEEFE94c8aD530842bfE7d8B397938fFc1cb83b2) has 6 enabled Morpho Blue markets but is now effectively a single-market vault: cbBTC/USDC 99.5% ($227.6M), then wstETH $896K, cbETH $144K, and the WETH sleeve run down from $19.4M to $89.9K (2026-07-21) — the four non-cbBTC sleeves total ~$1.13M. Morpho Blue is an immutable lending singleton; MetaMorpho is the curator layer on top, with role changes (caps, markets, guardian) gated by a 7-day timelock (timelock() = 604800) — but not setFee, which is what moved this quarter (Risk #4).

The borrowers are Coinbase retail users taking BTC-backed (BTC→cbBTC 1:1→USDC) and ETH-backed loans through Coinbase’s product. A large share of depositors are also Coinbase users routed in via the “Lend your USDC” feature — per-user smart-contract wallets pointed at this exact vault. The 86% LLTV on the dominant cbBTC market (lltv = 8.6e17) matches Coinbase’s stated liquidation threshold. Coinbase captures its margin off-chain; the on-chain MetaMorpho fee() is now 5% — set in the vault’s only-ever SetFee event, block 47,141,158, 2026-06-10 06:07 UTC, called by the owner Safe, recipient 0x255c7705…085a. Yield accrues via Morpho’s AdaptiveCurve IRM into convertToAssets; cbBTC-market spot supply APY ~4.53% gross / 4.30% net of the fee (borrowRateView at 90.15% utilization, 2026-07-21), up from ~3.87% pre-fee as utilization climbed — so the fee’s bite was more than offset by rate, and a holder’s realised yield went up. MORPHO rewards Coinbase advertises are not in NAV: totalAssets() reconciles exactly to the sum of market positions plus lostAssets() (= 0), and the vault holds zero MORPHO.

Exit is ERC-4626 withdraw/redeem — same-block at NAV when underlying market liquidity exists, queued in practice when it doesn’t. This is the Base V1 vault (MetaMorpho V1.1); a separate 0xbeef0880… “Instant V2” on Ethereum is a different rating.

Capital quality. The vault scaled from $8.5M (2025-10) to a peak of ≥$574M (2026-04-21) and held roughly flat through the Apr-2026 bridge attack itself ($527M → ~$521M across 18–19 Apr) — durable on the architecture-led read (isolated, immutable Morpho markets are contagion-resistant) and the utility-driven read (capital doing Coinbase’s lending job). But “no drawdown” is true only of NAV, not TVL: four days after the attack, on 2026-04-23, TVL fell ~$100M (−17.6%) inside 12 hours to ~$467M, where it then sat for weeks. Share price never flinched through it. But the marginal supplier is mobile, not sticky, and this quarter measured exactly how mobile. On-chain TVL: $336.9M (06-07, fee 0) → $265.4M (06-15) → $238.4M (06-22) → ~$229M flat since — −$98.5M over the two weeks after the fee, −32% in all. The fee switch and the outflow are one event: about a third of the vault left, then the bleed stopped dead — what remains is capital that tolerates the fee. Integrity held throughout: convertToAssets(1e18) strictly monotonic (1.0128 in 2025-10 → 1.049921 now), lostAssets() = 0, every exit at NAV. Legibility caveat: depositors route through per-user Coinbase smart wallets, so the holder base is partly opaque — we read retention off exit behavior, not a roster.

Risks

Verified on-chain (2026-07-21, Base block 48,933,322): TVL, share price, per-market allocation and utilization, the cbBTC oracle feed, fee() + its single SetFee event, timelock(), owner/curator thresholds, guardian revoke authority, instant capacity via maxWithdraw, and the absence of any steakUSDC AMM pool. Taken on trust: Coinbase’s stated 86% liquidation threshold and its retail-loan product mechanics (both pages 403 to automated fetch), Steakhouse’s zero-bad-debt record, cbBTC’s 1:1 Coinbase backing. Not derivable on-chain: the guardian (0x894d59…) is an ERC-1967 proxy, not a Safe — its revoke authority is confirmed by call, but its signer set is not readable.

Key-compromise threshold

Morpho Blue’s lending core is immutable and there is no admin NAV/oracle key at the vault — privilege concentrates in the MetaMorpho owner Safe, and the most dangerous actions route through the 7-day timelock (which the guardian can veto).

• Funds (theft / unbacked mint / principal loss): 5 of 9 + 7-day timelock — the owner Safe (0x0A0e…f8DD, 5/9, verified 2026-07-21) can enable a malicious or loose market and reallocate principal into it, but new-market enablement and cap raises are gated by the 7-day timelock() and the guardian (0x894d59…) can revoke a pending change inside the window — its revoke authority is confirmed by call, though the guardian is an ERC-1967 proxy rather than a Safe, so who controls it is not readable on-chain; treat that backstop as unverified at the key level. No instant principal-theft path exists: the Morpho Blue singleton is immutable (no upgrade-to-drain), and the only un-timelocked owner lever is setFee (≤50%), which siphons yield, not principal — and that lever has now been used (0 → 5%, Risk #4), so treat it as demonstrated rather than theoretical. The curator (0x827e…ECdB, 2/6) can only submit timelocked caps — but note all 6 curator signers are also owner-Safe signers (verified getOwners() on both, 2026-07-21), so the curator Safe is a strict subset of the owner Safe and the two roles are not independent parties: any 2 of those 6 act as curator, and 5 of the 9 act as owner.
• Liveness (freeze / NAV distortion / forced exit): 5 of 9 — the same owner Safe can reallocate the vault into fully-utilized markets to strand instant redemptions (a soft freeze), or setFee to 50% to force an economic exit. ERC-4626 redeem() is otherwise permissionless with no pause, and NAV is market-derived from the cbBTC market’s Chainlink feed — distortion requires compromising Chainlink, not a vault key.
• Single point of failure: no — both tiers require the 5/9 owner Safe; the curator (2/6) is timelock-bound and the guardian is veto-only. No single signer or standalone role breaches either tier. Caveat: the curator/owner signer sets overlap completely and the guardian’s own key structure is unreadable, so the layering is thinner than the three distinct addresses suggest.

Liquidity & exit

Primary path is ERC-4626 withdraw/redeem at NAV. convertToAssets(1e18) = 1,049,921 and totalAssets() = 228,691,750 USDC (2026-07-21, Base block 48,933,322). Instant capacity is $138.8M (60.7% of TVL) — measured directly as the vault’s own maxWithdraw under a balance override, which matches Σ min(vault position, market liquidity) to the cent; essentially all of it is the $138.4M of unborrowed cbBTC-market supply, at ~90% utilization. The vault’s $1.4M USDC balance is excluded: it is outside totalAssets(), outside maxWithdraw, frozen since April, and unrecoverable (skimRecipient() = 0x0). No secondary-sale path — getPool/getPair(steakUSDC, USDC) returns 0x0 on Aerodrome, Uniswap V3 (all fee tiers), and Uniswap V2 (re-checked 2026-07-21). The Public Allocator (0xa090…0467) is an allocator on this vault, but reallocateTo only moves this vault’s own assets between its own markets — with 99.5% already in cbBTC and ~$1.1M elsewhere, it offers no meaningful relief here; cross-vault help would have to come from third parties reallocating other vaults that share the cbBTC market. Cross-chain return uses Circle CCTP (minutes), not the OP Stack canonical bridge (~7d).

Defensible size at current liquidity (post-haircut): $30M — unchanged despite the smaller vault. Instant depth $138.8M; $30M = 21.6% of that, clears at NAV same-day. Apply a 3× stress haircut to available liquidity (→ $46.3M); $30M is 64.8% of that and still clears inside a 1-day window — tighter than at last review (44%) but intact. The haircut is corroborated twice over: the vault has met a ~$112M redemption in ~4 hours and a ~$98.5M fee-driven exit, both at NAV, without a queue. We size to instant liquidity, not TVL. Stress scenario (“NOCA exits 50% in 7d while peer redeemers also exit”): a $15M exit against post-haircut $46.3M is comfortable; the binding constraint remains a simultaneous spike to >98% utilization, which would queue the residual for borrower-repayment cycles. Note the sizing does not price the fee — that is a return question (Risk #4), not a liquidity one.

External reviews

Source links for the third-party takes in the Risks table. No standalone rating exists for this Base V1 vault — Staking Rewards rates the Ethereum sibling; no Bluechip, LlamaRisk, DeFiScan, or Exponential grade located. Per rubric.md, reviews older than 90d (⚠️) are stale — refresh before relying.

• Staking Rewards — rating BB- (rates the Ethereum V2 sibling, not this vault) · page updated 2026-04-09, accessed 2026-07-21 · ⚠️ stale (103d). The Base vault has its own Staking Rewards page with metrics but no rating letter.
• OAK Research — Resolv/USR curator-contagion investigation · 2026-03 · ⚠️ stale
• Summer.fi — Steakhouse curator profile · 2026-05-28
• Morpho — protocol security audits (12 audits, Certora-verified) · audit library