Moonwell and MAMO: 9.13 million in debt, 4.8 years of revenue

Moonwell closed August 27, 2026, with $9,131,342 in residual debt in its MAMO market on Base —representing 4.78 years of the protocol's annual revenue— according to the post-mortem published on August 28 by Anthias Labs, its risk curator. Residual debt is that which remains active after liquidations, and revenue refers to the income retained by the protocol, not the interest earned by lenders: $1,911,054 over twelve months according to DefiLlama. This same debt is equivalent to 35.9 times Moonwell's total reserves ($254,352 as of August 24). MAMO is the token for Mamo, the automated savings application on Base, listed as collateral in October 2025. The exploit opened the hole in 24 minutes by combining two independent flaws: a supply cap that only counted formal deposits and a price feed —the oracle that determines collateral value— lacking a deviation band. This article breaks down the four figures of the incident, reconstructs the mechanism contract by contract —including the oracle provider not named in the post-mortem— and measures the debt against the money Moonwell generates and holds.

What happened in Moonwell's MAMO market on August 27, 2026?

At 09:06:11 UTC on August 27, the MAMO/USD feed was at $0.010597. By 09:28:43, it reached $0.43127363. In between, an account accumulated 94,305,863 MAMO purchased on Base DEXs, formally supplied 15,089,595 to the Moonwell market, and transferred another 53,393,290 directly to the mMAMO contract (the receipt token issued by the market to each depositor) in two transfers at 09:19:59 and 09:21:09. The final loan was issued at 09:30:13. The first liquidation occurred 32 seconds later.

Four different amounts are circulating regarding the incident, and all four are correct: they measure different things. The Anthias Labs post-mortem publishes them labeled as follows:

• $11,028,762: The gross value borrowed across 18 operations involving cbBTC, WETH, USDC, and wstETH, valued using the Moonwell oracle at the block of each loan. This does not account for liquidations or repayments.
• $9,131,342: The debt remaining active after 595 liquidations, measured at oracle prices at the time of evidence collection. This is the amount the protocol carries.
• 8,729,454 USDC: The amount that left Base via Circle's CCTP bridge in two twin burns at 09:41:43 and 09:42:43. This is the outflow, not the loss.
• $6,784,655: The traced increase in stablecoin balance over the initial capital. The post-mortem marks this as the estimated profit.

The starting external capital was 1,947,391 USDC, obtained by selling 799 ETH arriving from Tornado Cash on August 21 and 23: the $6.78 million traced represents a 3.48x multiplier of that principal (own calculation). Moonwell reacted at 10:53:43, one hour and 23 minutes after the last loan, by lowering borrow caps to 1 wei, and at 11:09:43 with the MAMO supply cap.

How much do the $9.13 million weigh on Moonwell's accounting?

DefiLlama publishes two annual figures for Moonwell, and the gap between them defines the problem. What the aggregator labels as Fees is $8,650,625: the total interest paid by borrowers over twelve months, which goes almost entirely to lenders. What it labels as Revenue is $1,911,054: the portion the protocol retains via the reserve factor. The first measures the size of the business; the second, the cash reserves with which that business can cover its errors.

The $9,131,342 in residual debt from the MAMO market is equivalent to 4.78 years of Moonwell's annual revenue —$1,911,054 according to DefiLlama as of August 28, 2026— and 1.06 years of its fees. The same loss and the same aggregator yield 1.1 or 4.8 years depending on whose money is being counted.

The other metric is worse. The reserve report from August 24, three days before the incident, puts Moonwell's reserves across its three active chains at $254,352, with an accumulation of $27,786.19 over thirty days. The residual debt of the MAMO market is equivalent to 35.9 times all the reserves Moonwell had accumulated as of August 24, 2026. The MAMO market specifically had $495.72 in reserves that day, 49,571.509 MAMO.

Own calculation based on figures frozen on August 28, 2026, at 09:29 UTC. The TVL provides the external scale: Moonwell's total dropped from $71,761,050 on August 27 to $44,702,974 on August 28 at 08:47 UTC (−37.7%), and Base TVL fell from $68,437,976 to $41,698,650 (−39.1%). The residual debt weighs 13.3% of the total supplied on Base the previous day.

Why didn't the 20 million MAMO supply cap stop the attack on Moonwell?

MIP-B48 (Moonwell Improvement Proposal, the protocol's governance proposal format) listed MAMO on October 15, 2025, with a 50% collateral factor, a 20 million MAMO supply cap, a 3 million borrow cap, and a 30% reserve factor. On paper, that cap limits how much MAMO collateral can exist in the market. In practice, it only monitors one of the two doors through which value enters.

A Compound-style lending market values a position by multiplying mMAMO receipts by an exchange rate that divides the MAMO backing the market by the receipts issued. The supply cap is checked in the deposit function, which issues receipts. A standard ERC-20 transfer to the contract increases the backing without going through that function and without issuing a single receipt.

This is exactly what the post-mortem recorded: 53,393,290 MAMO entered via direct transfer; neither of the two transactions emitted a Mint event, the mMAMO supply remained unchanged in those blocks, and the exchange rate rose from 0.020513 to 0.075460 — a factor of 3.6788. The 20 million MAMO supply cap approved by Moonwell in October 2025 only counted formal deposits —15,089,595 MAMO, below the limit— and failed to see the 53,393,290 that entered via direct transfer on August 27, 2026. The account controlled approximately three-quarters of the mMAMO in circulation, according to the post-mortem, so it captured nearly all of the exchange rate increase.

The two multipliers compounded: the oracle price increased by 40.7x and the exchange rate by 3.68x — the dollar value of each mMAMO receipt rose approximately 150 times (own calculation). Without that second lever, the same price movement on 15 million MAMO would not have provided the borrowing capacity for eleven million.

Which Chainlink oracle powered Moonwell's MAMO market on Base and what did it verify?

The Anthias Labs post-mortem refers to a "source feed" without naming the provider. The contract chain on Base reveals it, and it can be traced with three read calls (verified on August 28, 2026, at 09:29 UTC):

The source feed is a Chainlink MAMO/USD oracle deployed on Base, with eight decimals, which Moonwell does not read directly: it reads it through a proprietary wrapper designed to capture OEV (oracle extractable value, the value liquidators extract by being the first to use a new price). By design, this wrapper returns the previous round when the latest one is less than ten seconds old and no one has paid to fast-track it. The post-mortem does not explain why the maximum accepted by Moonwell ($0.40248571 at 09:29:31) remained below the feed's peak ($0.43127363 at 09:28:43); the published sequence of rounds is not exhaustive.

What neither layer checks is magnitude. There is no deviation cap between rounds nor a plausible price band in the wrapper, and upstream, the Chainlink aggregator for MAMO/USD has minAnswer at 1 ($0.00000001) and maxAnswer at 2¹⁷⁶−1, with no operational ceiling. A jump from $0.0106 to $0.4313 in 22 minutes and 32 seconds passes all checks because none monitor velocity. The ChainlinkOEVWrapper (0xdbd3…afe6) used by Moonwell to read the Chainlink MAMO/USD feed on Base validates positive price, completed rounds, and non-stale responses, but none of the three checks measure deviation between rounds (contract reading, August 28, 2026).

The contrast with other markets lies in where illiquid collateral resides. Aave does not list such assets in its core market; Morpho Blue only has isolated markets with specific parameters, such as the cbXRP market we analyzed on August 26. MAMO was in Moonwell's core market on Base with a 50% collateral factor.

How many oracle incidents has Moonwell accumulated since October 2025?

Moonwell recorded four incidents linked to its price layer between October 10, 2025, and August 27, 2026 —321 days— totaling $16.31 million in added bad debt. The count reaches four because it includes October 10, 2025, which Anthias Labs documented with its own post-mortem on the Moonwell forum.

The cbETH incident on February 15 was not caused by an attacker: it was caused by the execution of MIP-X43, the proposal through which Moonwell governance activated Chainlink OEV wrappers in the core markets of Base and Optimism. A configuration error left cbETH valued at $1.12, liquidation bots took 1,096.317 cbETH by repaying minimal amounts, and correcting the oracle required five days of voting and timelock. The wrapper reading the MAMO/USD feed belongs to that same family of contracts and was deployed on May 17, 2026, three months later.

The other recurring detail is that remediation arrives late. The report from August 24, 2026, proposes repaying $22,554.36 of bad debt with reserves: 0.49% of the $4,583,019 pending at that time. On August 23, a cbETH supplier who was never liquidated opened a thread asking how to recover their position six months after the February incident.

With what funds can Moonwell cover the MAMO market debt?

Protocol Reserves. Adding the MAMO residual debt to what Moonwell already carried ($4,583,019 as of August 24, 99.87% on Base), the total reaches $13,714,361, covered by reserves at 1.85%. At the July-August accumulation rate of $27,786 per month, covering it solely with reserves would take 494 months: about 41 years. Furthermore, reserves are distributed by market, as the repayment mechanism only allows using a market's reserves against that specific market's debt.

Foundation Treasury. Borrowers affected by the cbETH incident requested in June 2026 that remediation be funded from there, citing a public statement from the Moonwell Foundation regarding a "substantial" treasury; without a verifiable primary source from the Foundation, that figure remains governance context. The precedent is a data point: cbETH remediation had distributed 12.29% of the obligation by June 4, 2026, nearly four months after the incident, and the affected parties themselves estimated about five years at the protocol's pace.

External Recovery. This is the path the post-mortem closes without explicitly saying so. The 8,728,319 USDC that reached Ethereum (1,135 less than those burned on Base) were converted to canonical DAI at 09:44:47, two minutes and four seconds after the second burn, and moved to another account a minute later. USDC is freezable by its issuer; DAI does not have this function. The complete conversion in just over three minutes from the first burn eliminates the issuer-freezing lever that did operate in other 2026 cases (Tether and IoTeX, in the Q1 DeFi security report).

What signals will indicate if Moonwell absorbs the MAMO market loss?

As of August 28, 2026, at 09:29 UTC, the most recent topic on the Moonwell governance forum is the post-mortem itself. There is no remediation MIP, and any presented requires five days of voting and timelock before execution. Four specific signals will resolve the issue in the coming weeks:

• A MAMO market remediation MIP, with an explicit source of funds. If it comes from reserves, the timeline is measured in years; if it comes from the Foundation treasury, it breaks the cbETH precedent.
• The late September 2026 reserve report from Anthias Labs, the first to incorporate the MAMO market. The figure to watch is the total net bad debt: on August 24, it was $4,583,019.
• Phase 2 oracle circuit breakers proposed on June 8, 2026, in the risk management evolution thread. They had been discussed for 80 days with only four messages and no MIP when the MAMO incident occurred.
• Treatment of the supply cap. A cap checked only on the deposit path leaves the direct transfer door open in any other market using the same code.

The thesis of this analysis could be overturned by a specific data point: if the late September reserve report or an approved MIP shows that Moonwell can cover at least half of the $9,131,342 with reserves or its own income —not from the Foundation treasury or external recoveries— in less than twelve months; or if DefiLlama corrects the protocol's annual revenue to above $9 million due to an adapter error. Based on the snapshot from August 28, 2026, neither condition is met.

Anyone evaluating a lending market on Base has two calculable ratios before depositing: what the protocol retains annually versus the bad debt it already carries, and how long its governance takes to change an oracle parameter. At Moonwell, as of August 28, 2026, these are $1.91 million against $13.71 million and a five-day timelock. The underlying mechanics can be found in our guide on oracles, and the ecosystem where this occurred in DeFi on Base.

Sources and links: MAMO Post-mortem, Anthias Labs (Aug-28-2026) · Reserve Recommendations 8/24/26 · MIP-X43 cbETH Oracle Incident Summary · wrsETH Oracle Malfunction (Nov-4-2025) · Remediation of Bad Debt From 10/10 Using Reserves (Nov-10-2025) · Report on the Events of October 10th, 2025 · Evolving Moonwell's Risk Management (Jun-8-2026) · DefiLlama: Moonwell revenue · DefiLlama: Moonwell TVL by chain · Moonwell ChainlinkOEVWrapper source code · OEV Wrapper for MAMO/USD feed on Base · Chainlink MAMO/USD Aggregator