KETJU Research

← The Register

Historical record. Archived 2026-08-16: Superseded by the 2026-08-16 research pass, which reached and published a new active verdict.. This preserves the last issued verdict and does not count toward current coverage.
stable-lending

Euler v2

Under review
Max sleeve
Reviewed
2026-07-31 · v1
Next review
2026-10-31
Research basis
Individual research
Chains
Ethereum · sovereign
Symbols
USDC USDT WETH

The scheduled date is the outside bound. Kill criteria are checked every day, and a trigger reopens the memo that week.

The original Euler lost about $197M in March 2023 and recovered 100% of the stolen assets; their value was about $240M when returned. Euler V2 launched in September 2024 after a ground-up rebuild. Those facts deserve weight, but they do not support protocol-wide approval. This review corrects a material error: Euler did not report $137M of protocol bad debt from Stream Finance. Its 2026 retrospective says Ethereum DAO-managed markets had zero direct Stream exposure, while the incentivized Plasma deployment lent heavily to Stream and was effectively wiped out. That is still a serious allocation and governance failure, not a second core-contract exploit. V2 is permissionless infrastructure: each EVK vault chooses collateral, LTVs, interest-rate model, oracle, hooks, governor and upgrade posture; EulerEarn can allocate across as many as 30 ERC-4626 strategies, including external ones. “Known” vault metadata is explicitly not an endorsement and curators can later change parameters. The formal under-review verdict is preserved, but the decision recommendation is clear: reject Euler as a protocol-level exposure and reopen only a named vault after an independent, address-specific dossier. Brand, aggregate TVL and audit count are not substitutes.

The research file

The mechanism

Euler V2 separates the lending primitive from the market product. The Euler Vault Kit deploys isolated ERC-4626 vaults; a vault holds one asset and its configuration sets collateral relationships, loan-to-value ratios, an interest-rate model, price oracles, caps and optional hooks. The Ethereum Vault Connector authenticates and batches operations across vaults, supports subaccounts and operators, and permits one vault’s shares to serve as collateral elsewhere. Isolation can contain a bad market, but composition can transmit the economic risk selected by its creator. EulerEarn adds a managed meta-vault: one deposit asset can be allocated across up to 30 Euler or external ERC-4626 strategies under curator-set caps and ordered supply and withdrawal queues.

Who controls it

Control is vault-specific. A creator chooses whether an EVK vault is upgradeable or immutable and governed or finalized. For upgradeable vaults, the factory upgrade administrator controlled by Euler DAO can replace implementation code; a governor can change permitted parameters; finalized immutable vaults remove both patch paths. Governed vaults may use curators, timelocks, guardians and emergency roles. EulerEarn owners assign roles and fees, curators add or remove strategies and caps, and allocators reorder flows. The docs warn that a “known” listing only means initial configuration was reviewed, not ongoing safety; curators retain control and can misconfigure it. An institutional memo therefore must name every contract, governor, signer threshold, timelock, oracle and strategy—not “Euler.”

Failure and operating record

Euler V1’s March 2023 exploit removed about $197M. Euler’s recovery account says all stolen assets were returned after negotiations and were worth about $240M at recovery. V2 is different code and launched in September 2024. Euler now reports more than 60 security reviews by more than 16 firms and a Cantina bounty up to $7.5M; these are issuer counts and do not certify each third-party vault, hook, oracle or external strategy.

The late-2025 Stream episode is governance evidence. Euler says DAO-managed markets had zero direct toxic-asset exposure and isolation contained them, but also says Stream borrowed much of the incentive-driven liquidity on Plasma and that deployment was effectively wiped out. Its own retrospective attributes the growth to weak curation, fragile deposits and unclear responsibility, and records a 2026 leadership transition. That does not justify the old $137M bad-debt claim; it does justify requiring named accountability and loss history for every vault.

Exit and liquidity

An EVK lender can call the ERC-4626 withdrawal path only while the vault has unborrowed underlying liquidity. A high-utilization or impaired borrow market can make an economically solvent share temporarily or permanently unredeemable. In EulerEarn, the curator’s withdrawal queue determines which underlying strategies are tapped first; a reserve strategy can improve responsiveness at the cost of yield. Strategy losses, exhausted queues and simultaneous withdrawals are thus part of the client exit, not implementation details. Any proposed position needs a same-block measurement of cash, utilization, borrower concentration, queue, caps and proposed-size withdrawal—not protocol TVL or share-token DEX volume.

Comparison and decision recommendation

Against Aave, Euler offers much wider market design freedom and finer isolation, but transfers far more asset, oracle and parameter diligence to each vault owner. Against Morpho, the underwriting problem is similar: the code may be common while the investable product is a curator and a finite allocation set. EulerEarn further permits external ERC-4626 strategies, so a familiar deposit symbol can conceal non-Euler risks. Against a single immutable pair, a governed market can react to a failing asset, while an immutable one removes administrator risk but cannot be patched or re-parameterized.

Recommendation: do not approve Euler V2 protocol-wide. Maintain under review only as a venue capable of producing an approvable named vault. A future approval must be address-specific, curator-specific and strategy-specific, with no inference of safety from the Euler label.

Open questions and observable reopen tests

For a candidate vault, record the factory and implementation hash; upgrade flag; governor, guardian, owner and allocator addresses; signer thresholds and timelocks; underlying asset; every collateral and strategy; oracle route and fallbacks; LTV, caps, utilization and borrower concentrations; prior bad debt; incentives; and withdraw queue. Reconcile the live configuration to the exact audit scope and simulate collateral impairment, oracle staleness, full utilization and a proposed-size run.

Reopen a named vault only after six months with no realized bad debt, every asset already approved, a verified accountable curator, risk-increasing changes subject to at least a one-day exit window, and the proposed position withdrawable inside the written slippage and timing limit under stress. Any unapproved strategy, unannounced governor change, audit-scope mismatch or loss closes that vault file. No aggregate protocol metric can satisfy these tests.

Sources

The claims above trace to these. Where a number could not be independently verified, the thesis says so.

Inherited controls

The verdict above grades the protocol layer. Every position also inherits the asset it holds and the chain it settles on. The least safe layer sets the position’s grade, and the position table names which one that is.

ChainVerdictGradeControl constraint
EthereumApproved sovereign No sequencer, no upgrade key, no operator who can be compelled — rule changes require social consensus.
AssetGradeWho can freeze it
USDC freezable Issued by Circle, backed by bank deposits and T-bills. Circle can and does freeze addresses on request from law enforcement.
USDT freezable Issued by Tether. Has frozen addresses on request. Reserve composition is less transparently attested than USDC.
WETH sovereign Wrapped ETH. Immutable contract, no admin key, no blocklist.
The memo is public. The watching is the product: the terminal reads your clients’ wallets against this Register and flags the events above when they fire. $49 per advisor per month, first 14 days free. Start the trial.