Aave moves to retire 96 markets and exit six chains, with $98.1 million of deposits in scope

Aave founder Stani Kulechov announced a governance proposal on Thursday, July 30, that would strip dozens of asset markets out of the lending protocol and shut down its deployments on six blockchains entirely. The Block, which reported the announcement, put the scope at roughly $98.1 million in supplied assets and $15.6 million in outstanding debt.

The package has two halves. The first offboards 50 low-adoption reserves and 21 matured Pendle Principal Tokens spread across 11 Aave V3 deployments — 71 markets, accounting for $85.3 million in supplied assets and $11.5 million in debt, per the proposal figures cited by The Block. The second retires six deployments outright: Sonic, Scroll, zkSync, Metis, Soneium and Aptos, which between them carry another 25 reserves, $12.8 million in supplied assets and $4.1 million in debt.

The proposal was prepared by LlamaRisk, a risk service provider to the Aave DAO, and applies the protocol's newly proposed Aave Risk Framework across every deployment rather than reacting to any single asset or incident. Its stated targets are markets whose activity no longer justifies the standing cost of price oracles, liquidation infrastructure and risk monitoring; bridged assets that duplicate a native listing already on the protocol; expired Pendle PTs that no longer generate yield; and deployments where revenue has fallen below maintenance cost.

The economics

CoinDesk went through the revenue case. Each of the six chains slated for closure now generates under $5,000 a quarter, and Metis, Soneium and Aptos bring in less than $1,000 each. For scale, CoinDesk cites Ethereum mainnet at more than $142 million a year and Base at about $4.7 million, against roughly $3,000 for Metis.

Deposits on all six have fallen sharply over six months, per the figures CoinDesk reports from the proposal: Soneium down 95%, Aptos available liquidity down 94%, zkSync down 88% to about $844,000, Scroll down 86% to roughly $2 million, Metis down 79%, and Sonic — the largest of the group — down 74% to just under $8 million. DefiLlama data cited by CoinDesk puts the six chains' combined deposits at about $13 million against roughly $14 billion protocol-wide across 23 chains, under 1% of assets. That $13 million sits slightly above the $12.8 million in the proposal itself; the two are almost certainly different snapshots, but neither source reconciles them.

Aave's own economics compress that further. CoinDesk, citing DefiLlama, says borrowers paid about $888 million in interest over the past year, of which the protocol retained roughly $117 million — about 13 cents on the dollar, with the rest flowing to suppliers. On that take rate, Aave's share of a $5,000 quarter is a few hundred dollars.

Mechanics: no forced exits

Nothing gets liquidated by decree. For individual reserves, Aave would freeze new activity, cut supply and borrow caps to a single unit, and raise reserve factors on borrowable assets, according to The Block. For the six full wind-downs, both sources describe a harsher setting: reserve factors raised to 99%, so nearly all borrower interest is routed to the treasury rather than to suppliers, plus — per CoinDesk — a 5% base borrowing rate. Markets are frozen to new deposits, new borrowing and collateral use. The design intent is that holding a position becomes expensive enough that remaining users close out voluntarily.

Where this came from

The risk framework behind the review was introduced in June, after the roughly $292 million KelpDAO bridge exploit left Aave exposed to potential bad debt when stolen rsETH was deposited as collateral. Kulechov said at the time the framework would govern listings, ongoing reviews and deprecations across Aave V3, V4 and Aave Horizon, with assets failing the new bar removed. A companion LlamaRisk proposal flags a further set of long-tail reserves for elevated Chainlink price-feed risk and would swap their live feeds for fixed-price adapters.

The economic argument is older than the framework. CoinDesk notes that in December, the Aave Chan Initiative proposed rolling back zkSync, Metis and Soneium as having "proven to lack product market fit," and pushed a rule requiring any future deployment to commit to at least $2 million in annual revenue.

Key facts

• $98.1 million supplied, $15.6 million debt in scope; split $85.3M/$11.5M for individual reserves and $12.8M/$4.1M for the six deployments — The Block, citing the proposal
• 50 low-adoption reserves + 21 matured Pendle PTs across 11 V3 deployments; 25 further reserves on the six chains — The Block
• Six chains generate under $5,000/quarter each; Metis, Soneium, Aptos under $1,000 each — CoinDesk, citing the proposal
• Ethereum mainnet: >$142M/year; Base: ~$4.7M; Metis: ~$3,000 — CoinDesk
• Six-month deposit declines: Soneium −95%, Aptos liquidity −94%, zkSync −88% ($844k), Scroll −86% ($2M), Metis −79%, Sonic −74% (just under $8M) — CoinDesk
• Aave gross revenue: $198M in Q1 2026 → $156M in Q2, down about a fifth; liquidation fees $27M in Q2 → under $200,000 one month into Q3 — CoinDesk
• Borrower interest paid over the past year ~$888M, of which Aave retained ~$117M — DefiLlama via CoinDesk
• AAVE traded at $95.52 on July 30, down 2.75% on the day — The Block's price page

The real-world read

The reversal is the story, and it's being narrated backwards. Aave spent years expanding onto exactly these chains. It is now withdrawing from six of them, and the December ACI proposal already named three — zkSync, Metis, Soneium — as lacking product-market fit, along with a $2 million annual revenue floor for new deployments. That is a commercial verdict rendered roughly seven months before the risk framework existed. Presenting the wind-downs primarily as an output of a June risk process gives a tidier institutional gloss to a decision that was, on the sources' own timeline, already made on economics.

Risk framing and the triggering incident don't line up. The framework was born from a $292 million exploit involving rsETH collateral. Closing a Metis deployment that earns about $3,000 a quarter does not address that class of exposure. The genuine risk reduction here is operational surface area — fewer oracles, fewer liquidation paths, fewer things to monitor — which is real but not the same thing.

The savings are rounding error against the actual problem. Six chains at under $5,000 a quarter gross is under $20,000 combined; at Aave's ~13% take, the protocol's own share is low four figures. Meanwhile gross revenue fell $42 million between Q1 and Q2, and Q3 liquidation fees are running at under $200,000 versus $27 million the prior quarter. Cleaning up the long tail is defensible housekeeping. It is not a response to that decline, and nobody quoted claims otherwise — but the timing invites the inference.

Numbers to hold loosely. CoinDesk's framing leads with Aave exiting six chains "affecting about $98 million in deposits"; the proposal's own breakdown puts the six chains at $12.8 million, with the $98.1 million covering the entire package including 71 markets on chains Aave is keeping. The Aptos figure in the six-month decline list is available liquidity, not deposits, sitting unlabelled among five deposit figures. And CoinDesk's Q2 revenue split trails off mid-sentence without stating what Aave kept of the $156 million.

Whose numbers these are. Nearly every figure originates in the proposal itself, authored by LlamaRisk — a paid service provider to the DAO whose remit expands under the framework it is now applying — and promoted by Aave's founder. That does not make the numbers wrong; the revenue collapse is corroborated by DefiLlama data CoinDesk pulled independently. It does mean the cost-benefit judgment is being made by parties with a stake in the framework's adoption.

Conspicuously unsaid: no vote date or execution timeline appears in either account; neither says what happens to suppliers left in a frozen market if the last borrowers simply don't repay, or whether any backstop exists; and there is no mention of whether the six chains' foundations were notified, or whether any grant or incentive commitments are being unwound. Aave was an anchor DeFi tenant on several of these networks — what its exit does to them goes unaddressed.

Opinion, and whose

• Stani Kulechov (Aave founder): said in June that the risk framework would become the standard governing listings, reviews and deprecations across V3, V4 and Horizon, with non-compliant assets removed.
• LlamaRisk: judges that these markets' activity no longer justifies the operational overhead, and that revenue on the six deployments no longer covers maintenance. A cost-benefit call, not an observed fact.
• Aave Chan Initiative (December): zkSync, Metis and Soneium had "proven to lack product market fit"; proposed a $2 million annual revenue commitment for future deployments.
• CoinDesk: characterises Metis's economics as "worth roughly the price of a dinner" — the outlet's editorial gloss on the proposal's figures, not a claim by Aave.

None of the above is a forecast about AAVE's price, and none of the sources offers one.

Sources

• The Block, Brian Danga (July 30, 2026) — the proposal's structure and scope: the $98.1M/$15.6M totals and their split, the count of reserves and Pendle PTs, the 11 deployments, the freeze mechanics, the KelpDAO exploit background, the Chainlink fixed-price-adapter companion proposal, and the AAVE price quote. The Block discloses that Foresight Ventures is its majority investor and that Bitget is an anchor LP for Foresight; the article is editorial, not sponsored. Unrelated promotional units for LMAX Digital and Polymarket appeared alongside it on the page.
• CoinDesk (July 30, 2026) — the revenue economics, per-chain deposit declines, comparative figures for Ethereum mainnet and Base, quarterly revenue and liquidation-fee trend, the DefiLlama take-rate and TVL data, the 5% base rate detail, and the December ACI history. Editorial, not sponsored.
• Underlying primary material — the LlamaRisk-authored governance proposal and DefiLlama data, both accessed here via the two reports above rather than read directly.

Not financial advice.