Maple Finance

Rejected venues wait the longest for re-review; a rejection has to earn another look before the scheduled date.

REJECTED, and re-argued in full at this review because the product changed and the rejection’s grounds had to change with it. Maple today is not the Maple that failed in 2022. Its docs say every loan behind syrupUSDC, syrupUSDT, and syrupUSDG is overcollateralized with liquid digital assets; third parties report ratios above 150% (OAK Research). Underwriting moved in-house to Maple Direct, with 24-hour margin calls and contractual liquidation levels; the third-party pool-delegate model that produced the 2022 losses is gone from the core products. The scale is real: Maple reports $4.6B AUM at H1 2026, up 81% year over year, and put syrupUSDC supply near $2.8B in June 2026, while DefiLlama shows $2.39B on-chain TVL with $1.92B borrowed at the 2026-08-14 review. Credit the redesign honestly: it fixed underwriting. What it did instead is concentrate the risk in the operator, and nothing about the operator can be verified from outside. Collateral does not sit in protocol contracts; it sits with off-chain custodians (BitGo, Copper, Hex Trust, Kraken Financial, per Maple and third-party reporting). Margining runs on Maple’s private three-feed alert system and liquidation is a manual process through OTC desks, all invisible on-chain. The High Yield product, in Maple’s own docs, redeploys collateral into staking and secured lending, so that collateral must be recalled before it can be liquidated. A Security Admin can pause every function, and the GovernorTimelock behind upgrades, only added in September 2025, is managed by a multisig whose threshold and signers are unpublished. Withdrawals queue behind a stated 30-day ceiling, processed by Maple Direct. And the claim of no defaults since 2023, across $20B+ of originations, is Maple’s own figure relayed by third parties; no independent loan-tape audit exists, and an impairment workout would not show as a default. The 2022 record stays because it prices what trusting an operator’s representations can cost. On 2022-12-05 Orthogonal Trading defaulted on $36M across eight loans, about 30% of all active loans: $31M in the M11 USDC pool, where reporting at the time put remaining lenders’ losses near 80%, plus $5M (3,900 wETH) in the M11 wETH pool. Maple expected to recover about $2.5M of the $36M. Orthogonal had told lenders in November its FTX exposure was about $2.5M; on December 3 it disclosed far more. Babel Finance had already defaulted that July, a $7.9M loss, a 3.8% haircut on its pool. One firm now underwrites, prices, margins, liquidates, and pays out. A client cannot hold this and also hold the belief that their downside is observable. The rejection extends to syrupUSDC and syrupUSDT wherever they are accepted as collateral, or it leaks back in through the side door.

Kill criteria: the events that revoke this verdict
• The Governor multisig’s signers and threshold are published and every privileged action runs through the timelock
• Collateral custody and margining become verifiable: positions on-chain, or independent attestations of custodian balances and margin events
• Rehypothecation of borrower collateral ends, or is ring-fenced entirely outside the products under consideration
• Maple publishes a full independently audited default, impairment, and loss record covering 2023 onward

The research file

The 2022 record

Orthogonal Trading received a notice of default on 2022-12-05 for eight loans totalling $36M, about 30% of all active loans on Maple at the time, verified against contemporaneous reporting by The Block, CoinDesk, and Bloomberg. The pool split is verified: $31M sat in the M11 Credit USDC pool across four loans, and a further $5M, 3,900 wETH, in the M11 wETH pool was past due. M11 Credit, a sister company of Maven 11, was the pool delegate for both.

The deception is verified. Through November 2022 Orthogonal repeatedly told M11 Credit its FTX exposure was about $2.5M. On December 3 it disclosed it had substantially more stuck on FTX and could not repay. Maple severed ties the same week, publicly accusing Orthogonal of ”misrepresenting its financial position.”

The roughly 80% lender loss in the M11 USDC pool is reported by third parties; press summaries state investors in that pool faced a roughly 80% loss on remaining capital. CoinDesk, 2022-12-12, shows the mechanism: by December 9 that pool had 80% of all its loans to one borrower, Orthogonal, up from 14% on August 31, and pool cover was mostly depleted, with less than $2M across the three troubled pools’ covers. The exact realized percentage is a reconstruction from reporting, not a Maple-published number. On recovery, third parties report Maple conceded it expected to recover only about $2.5M of the $36M, from pool cover and Orthogonal’s accrued fees, a roughly 7% recovery. Orthogonal entered provisional liquidation in the British Virgin Islands in late December 2022; no final liquidation recovery figure was found. The scale of the wreck is verified: CoinDesk totalled $36M defaulted plus $18M distressed, $54M, which was 66% of everything outstanding across Maple’s four active pools.

Babel Finance came first. Verified via Orthogonal Credit’s own Medium posts and CoinDesk: Babel defaulted on a $10M USDC loan from the Orthogonal-delegated USDC pool, formally liquidated around 2022-07-10. Creditors booked a $7.9M loss, about a 3.8% haircut on the pool because the pool was large. Note the delegate irony: Orthogonal Credit was the pool delegate that underwrote Babel, five months before Orthogonal Trading itself defaulted.

Auros Global rounds out the season. Third parties report Auros missed payments on $17.7M of Maple loans starting late November 2022, first a $3M principal payment on 2022-11-30, citing funds frozen on FTX, and disclosed BVI provisional liquidation on 2022-12-20. M11 chose restructuring over default: 55% repaid, 40% reissued at 8.64% annualized over three 90-day cycles, the rest renewed at zero interest. M11 said lenders should expect full recovery; the Blockworks headline about M11 pool investors expecting full debt repayment refers to the Auros debt, not Orthogonal. Whether the full recovery completed was not independently confirmed.

The product now

Maple v2 shipped December 2022, rebuilt around withdrawal managers, impairments, and Governor-timelock upgrades; docs list three audits for that release, Trail of Bits in August 2022, Spearbit in October 2022, and Three Sigma in October 2022. The pivot that matters came after: docs say current lending is overcollateralized, that all syrupUSDC, syrupUSDT, and syrupUSDG loans are overcollateralized, and that yield comes from fixed-rate, overcollateralized loans to institutional borrowers, supplemented by DeFi liquidity provision and futures basis trading. Third parties report collateralization ratios kept above 150% (OAK Research). Exact per-loan LTVs are set in private term sheets, not published; docs describe LTV, margin call, and liquidation levels per loan visible to the borrower, not to the public.

The product structure now runs on several lines. The Syrup pools, syrupUSDC, syrupUSDT, and syrupUSDG, are permissionless ERC-4626 vaults for open deposit, launched 2024; syrupUSDC grew from about $100M at launch to about $2.8B of supply by June 2026, per Maple, and syrupUSDG launched on Robinhood Chain and reached $200M in eight days, per Maple’s July 2026 memo. The permissioned institutional pools are Blue Chip Secured, BTC and ETH collateral only, held in qualified custody, with third parties reporting about 7.5% yield, and High Yield Secured, which accepts riskier collateral such as XRP, SOL, and POL, with third parties reporting 11% or more; docs say High Yield earns the extra yield by ”reinvesting the collateral in staking and/or secured lending opportunities,” which is rehypothecation in Maple’s own words. A BTC Yield product runs with Core Foundation, collateral held at BitGo, Copper, and Hex Trust per third-party reporting. A Kraken warehouse lending facility has Maple providing senior financing through a bankruptcy-remote SPV with Kraken Financial holding the collateral, per Maple’s July 2026 memo. A Cash Management pool holds T-bills via a single borrower, Room40, with J.P. Morgan as custodian and prime broker, separate from the credit pools.

The size: Maple reports $4.6B AUM at H1 2026, up 81% year over year, with more than $22B of total originations since 2022. DefiLlama, verified via API on 2026-08-14, shows the protocol at $2.39B TVL on Ethereum with $1.92B borrowed; Solana TVL rounds to zero on their accounting. The AUM figure and DefiLlama’s TVL differ by methodology, AUM counting the loan book and off-chain facilities, TVL counting on-chain deposits. syrupUSDC and syrupUSDT deployed beyond Ethereum to Solana, Arbitrum, Base, and Plasma in 2025, and syrupUSDG to Robinhood Chain, per Maple; this entry lists Ethereum only.

The risk shape

The borrowers, docs say, are creditworthy crypto-native institutions that post liquid digital assets as collateral: trading firms, market makers, exchanges, with Kraken named. Individual borrower books are still not public; the improvement is the posted collateral, not borrower transparency.

The collateral is BTC and ETH in Blue Chip and XRP, SOL, POL and similar in High Yield, per third parties. Docs confirm collateral can even be the same asset as the loan, and that High Yield redeploys collateral into staking and secured lending. Redeployed collateral is not sitting in a vault waiting to be liquidated; it must be recalled first.

Custody is off-chain. Docs say qualified custody; third parties name BitGo, Copper, and Hex Trust for BTC Yield, and Kraken Financial holds the Kraken facility’s collateral. Collateral is held with custodians, not in protocol contracts, so a lender is trusting Maple’s operational pipeline, not a liquidation bot.

Margining, docs say, runs on a proprietary alert system with three price-feed sources and 24/7 monitoring. At the Margin Call Level the borrower has 24 hours to restore the Initial Collateral Level; at the Liquidation Level Maple may liquidate immediately, margin call in process or not. Liquidation runs primarily through OTC desks, with CEX and DEX as fallback. This is a manual, discretionary process run by Maple Direct, Maple’s lending arm, not an on-chain mechanism. Docs claim margin calls over the past year were met within hours, often topped up before the call.

Withdrawals, docs say, are queue-based and first-in-first-out, processed as liquidity comes free. syrupUSDC normally has instant liquidity via Uniswap and Balancer buffer pools; queue processing is typically under 24 hours with a stated maximum of 30 days, and the Maple Direct team processes the queue. In a real credit stress the impairment mechanism cuts pool value first, so leavers exit at a penalty and forfeit any later recovery to those who stayed.

The oracles do not add independence. Pyth and Chainlink publish syrupUSDC prices computed from the ERC-4626 exchange rate, that is, from Maple’s own pool accounting. Loan margining prices come from Maple’s internal three-feed system, not from a public oracle.

Maple Direct underwrites, structures, and manages every loan in the pools Syrup feeds. Since the 2023 restructuring there are no third-party pool delegates in the Syrup path; the delegate role and its M11-style outsourced underwriting is gone from the core products. One firm is now underwriter, margin agent, liquidator, and queue operator.

Who controls it

SYRUP replaced MPL at 1 MPL to 100 SYRUP; SYRUP launched November 2024, and the conversion window closed 2025-04-30, with a 48-hour extension over 2025-05-19 to 21. Governance votes run on Snapshot by stSYRUP stakers, which is off-chain signalling, not binding on-chain execution.

On-chain authority, docs say, sits with the Governor, a GovernorTimelock contract managed by a multisig; privileged actions, upgrades, globals parameters, and fee rates, follow a schedule, a delay, then a bounded execution window. The timelock itself was only added in a September 2025 upgrade, audited by Sherlock and 0xMacro in September 2025. A Security Admin can pause every function in the protocol in an emergency, and an Operational Admin holds a subset of routine powers. Multisig thresholds and signer identities are not in the docs and could not be verified.

At pool level the Pool Delegate role still exists in the contracts, funding loans and setting caps and fees; in current products it is Maple Direct. Upgrades to pool contracts by pool-level admins require a scheduled call recorded in Globals, with the Security Admin holding direct execution rights on some contracts.

The record since 2022

No post-2022 default or lender loss surfaced in any source reviewed. Third parties report Maple claims zero losses since 2023, with margin calls executed without any loss to lenders even during periods of sharp market volatility, per OAK Research with June 2026 data. That is Maple’s own claim relayed by a third party, not an independent audit of the loan tape, and the impairment mechanism means a quiet workout would not necessarily look like a default.

2023 was the recovery year: pools relaunched under Maple Direct underwriting, the cash-management T-bill product launched, and TVL rebuilt from the post-FTX trough. Third parties report year-end 2024 figures around $445M before the 2025 jump past $4B.

Docs say Maple v2 has had seven or more audits: Trail of Bits in August 2022, Spearbit in October 2022, and Three Sigma in October 2022 for the December 2022 release, two more for June 2023, and Sherlock and 0xMacro for the September 2025 Governor-timelock release. Contracts are open source, and the audit PDFs sit in the maple-labs GitHub. A bug bounty is live on Immunefi, linked from the docs, but the bounty cap could not be verified. DeFiSafety scored Maple 92%, per a report the docs cite.

What changed and what did not

The old rejection’s core claim, that Maple is undercollateralised institutional credit lending to trading firms on reputation rather than posted collateral, was true of the product that failed in 2022 and is not an accurate description of the product sold in 2026. In 2021-2022, Maple v1 had third-party pool delegates, Orthogonal Credit and M11 Credit, underwriting unsecured or lightly secured loans on reputation and financial disclosures; the contracts even permit uncollateralized loans, and the 2022 book largely was. In 2026, docs say every loan behind the Syrup tokens is overcollateralized with liquid digital assets, margin-called at 24 hours, and liquidated at a contractual level, with third parties reporting ratios above 150%, and underwriting in-house at Maple Direct.

What did not change is where the trust sits. It is still in an operator, not a mechanism. Collateral sits with off-chain custodians, margining runs on Maple’s private price feeds, liquidation is a human process through OTC desks, and the withdrawal queue is processed by Maple Direct with a 30-day outside limit; none of this is verifiable or enforceable on-chain by a lender. The concentration of function is total: one firm underwrites, monitors, margins, liquidates, and pays out. In 2022 the failure was a delegate lying to lenders; the delegate layer was removed, not disintermediated. High Yield explicitly rehypothecates collateral into staking and secured lending and accepts altcoin collateral, XRP, SOL, POL, whose liquidity in a crash disappears exactly when it is needed. And borrower books remain uninspectable on-chain, which is the test the off-chain-credit class rule actually applies.

So the rejection’s argument moves from reputation lending, which is stale, to operator-dependent, off-chain-margined secured lending whose collateral custody, pricing, and liquidation cannot be verified on-chain, run by the platform whose previous model produced an 80% pool loss. The 2022 record stays as the demonstrated failure mode of trusting the operator’s representations.

Comparison and rejection decision

Compared with Aave v3, Maple’s current loans may be overcollateralized, but the collateral balances, borrower-level LTVs, margin events and OTC liquidations are not independently observable on-chain; Aave exposes positions and liquidation rules in contracts. Compared with a Steakhouse-curated Morpho vault, Maple concentrates underwriting, price monitoring, liquidation and withdrawal-queue operation in Maple Direct, while Morpho allocations and market state are public. Compared with a tokenized Treasury fund, Syrup is a private-credit and trading counterparty claim rather than a short-government portfolio with an identified custodian and administrator. The newer collateralized structure improves on Maple v1, but it does not cure the institutional mandate’s observability and control defects; rejection and a zero cap remain appropriate.

Open questions

The exact roughly 80% M11 USDC pool lender loss is widely reported and the mechanism is confirmed, concentration plus depleted cover, but no Maple- or M11-published final loss percentage was found; it stays attributed to reporting. A 17% M11 wETH pool loss figure that circulated could not be sourced this pass; the verified figure is the $5M, 3,900 wETH, exposure. The final recovery from the Orthogonal BVI liquidation after the initial $2.5M estimate has no published outcome. Whether Auros lenders were in fact made whole after the restructuring was projected by M11 but not confirmed complete.

Actual current LTV, margin call, and liquidation levels are per term sheet and not published; above 150% collateralization is a third-party summary. The zero-losses-since-2023 claim is Maple’s, relayed by third parties, with no independent loan-tape audit, and impairments or workouts would not be visible as defaults. The Governor multisig threshold, signer identities, and the timelock delay length are unpublished. The Immunefi bounty maximum could not be verified. The split of the $4.6B AUM between the Syrup pools and the permissioned institutional pools is not solid enough to print, since the $2.8B syrupUSDC figure is a supply number, not an AUM number. Whether syrupUSDC is currently accepted as Aave collateral, asserted in the prior memo, was not re-verified this pass.

Sources

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

• Maple documentation export, fetched 2026-08-14 · primary · accessed 2026-08-14
  Supports: current loan structure, margin and liquidation process, withdrawal queue, governance controls
• Maple memo, July 2026 · primary · accessed 2026-08-15
  Supports: H1 2026 AUM, Syrup product growth, Kraken facility
• DefiLlama Maple protocol API · secondary · accessed 2026-08-14
  Supports: protocol TVL, borrowed amount, chain breakdown
• OAK Research Maple Finance overview · secondary · accessed 2026-08-15
  Supports: reported collateralization, reported post-2023 loss record, product yields
• The Block: Orthogonal Trading default, December 2022 · secondary · accessed 2026-08-15
  Supports: $36M Orthogonal default, M11 pool exposures
• CoinDesk: Maple’s $54M of sour debt, 2022-12-12 · secondary · accessed 2026-08-15
  Supports: $54M defaulted and distressed loans, borrower concentration, pool-cover depletion
• CoinDesk: Maple severs ties with Orthogonal, 2022-12-05 · secondary · accessed 2026-08-15
  Supports: Orthogonal misrepresentation, Maple relationship termination, default timeline

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.