[ARFC] Deploy Aave V4 on Base

This post has been updated on 2026-09-21 to reflect the strategic direction discussed and agreed with the DAO Service Providers, also to include the parameters provided by @LlamaRisk in this thread.

Summary

This ARFC seeks community feedback on deploying Aave V4 on Base and activating an initial tokenized equities market.

Base is an Ethereum Layer 2 incubated by Coinbase, with low transaction costs, native USDC liquidity, and access to a broad retail user base. Aave has operated on Base since 2023, and the network was among Aave’s first multi-chain deployments to surpass $1 billion in deposits.

The deployment would bring the Hub and Spoke architecture to one of Aave’s largest existing deployments. It would establish a dedicated Equities Hub, separating liquidity and risk for the tokenized equities market from other Aave markets on Base, while providing a foundation for specialized markets as the network develops.

Motivation

Base combines an established Aave deployment, growing stablecoin activity, and an expanding set of consumer applications, payments products, and trading venues. These conditions support demand for onchain credit infrastructure that can serve distinct use cases while maintaining clearly defined liquidity and risk boundaries.

Deploying Aave V4 on Base would:

  • Upgrade one of Aave’s largest existing deployments to the current protocol architecture.
  • Establish a dedicated market for tokenized equities, with USDC as the borrowable asset and tokenized equities used exclusively as collateral at launch.
  • Separate the liquidity and risk of this market from other Aave markets on Base through a dedicated Equities Hub.
  • Extend Aave’s presence within the Base and Coinbase ecosystem as additional users, applications, and tokenized assets come onchain.

The initial market provides a focused implementation of the V4 architecture. A single pooled spoke will support the initial set of tokenized equities as collateral, allowing users to borrow USDC against diversified positions while maintaining asset-specific collateral factors. The equities themselves will not be borrowable at launch.

Coinbase’s distribution, infrastructure, and institutional relationships may support a broader range of tokenized assets on Base over time. A dedicated V4 market provides a measured starting point for this activity, with a market structure designed to accommodate distinct collateral types and risk profiles without extending their exposure to other Aave markets.

Market Structure

The tokenized equities market is deployed through a dedicated Equities Hub on Base. USDC suppliers to the hub explicitly opt into lending against tokenized equity collateral, and this exposure is isolated from other Aave markets.

The Hub contains a single USDC reserve and two spokes. The Mag-7 Spoke supports USDC borrowing against the seven specified tokenized equities. The Tokenized USDC Spoke is supply only, providing vaults and aggregators with a composable USDC position without collateral exposure.

Within the Mag-7 Spoke, users may supply any combination of the seven tokenized equities as collateral and borrow USDC. Each asset retains its own collateral factor, so borrowing capacity remains determined by the composition of each position. Pooling the assets within one spoke allows diversified collateral positions to be managed within a single market.

The tokenized equities are collateral only at launch. Borrowing of equities, and positions that borrow one equity against another, are not enabled.

Specification

Dynamic Liquidation Bonus Configuration

Chain Hub Spoke Liquidation Bonus Factor Target Health Factor Health Factor For Max Bonus
Base Equities Hub Mag-7 Spoke 90.00% 1.24 0.90

Spoke Parameters

The liquidation fee is 10%, consistent with the rest of Aave V4. The risk premium threshold is 0 for every reserve, as risk premiums are not in use, and liquidators can receive collateral as shares on every reserve.

Chain Hub Spoke Reserve Collateral Factor Max Liquidation Bonus Borrowable Collateral Risk Liquidation Fee Risk Premium Threshold Receive Shares
Base Equities Hub Mag-7 Spoke AAPLc 78.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke AMZNc 73.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke GOOGLc 76.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke METAc 65.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke MSFTc 79.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke NVDAc 70.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke TSLAc 65.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke USDC 0.00% - TRUE - - 0 TRUE

Add and Draw Caps

Chain Hub Spoke Reserve Add Cap Draw Cap
Base Equities Hub Mag-7 Spoke AAPLc 15,000 0
Base Equities Hub Mag-7 Spoke AMZNc 10,500 0
Base Equities Hub Mag-7 Spoke GOOGLc 15,000 0
Base Equities Hub Mag-7 Spoke METAc 5,800 0
Base Equities Hub Mag-7 Spoke MSFTc 5,200 0
Base Equities Hub Mag-7 Spoke NVDAc 24,000 0
Base Equities Hub Mag-7 Spoke TSLAc 14,000 0
Base Equities Hub Mag-7 Spoke USDC 32,000,000 21,000,000
Base Equities Hub Tokenized USDC Spoke USDC 1,000,000 0

Interest Rate Curve

Chain Hub Asset Base Drawn Rate Rate Growth Before Optimal Rate Growth After Optimal Optimal Usage Ratio Liquidity Fee
Base Equities Hub USDC 0.00% 4.00% 20.00% 90.00% 10.00%

Price Feeds

Each reserve is priced by the Chainlink total return feed for the token, which carries the issuer multiplier and follows the 24/5 US equities market hours. These are standard feeds. Smart Value Recapture variants are not used at rollout. USDC is priced by the Chainlink USDC/USD feed on Base through the stable price cap adapter, with the cap set at 1.04.

Next Steps

  • Gather community feedback and service provider recommendations on this ARFC during a five day forum discussion period.
  • Aave Labs and LlamaRisk will complete their assessments and append the proposed initial assets and the Hub and Spoke configuration to this proposal.
  • Escalate the proposal to an ARFC Snapshot for a three day off-chain vote.
  • Following a successful Snapshot, submit an AIP for an onchain vote to deploy Aave V4 on Base.

Useful Links

Disclosures

Aave Labs is the author of this proposal and has received no compensation from third parties for its creation.

Copyright

Copyright and related rights waived via CC0.

8 Likes

Avalanche, Arc and Tempo each ran a Temp Check before their ARFC, and Base going straight to ARFC only reads as consistent if this is treated as an upgrade of an existing deployment rather than a new market. If it is an upgrade, then the migration approach for existing V3 positions is not a detail to finalize during the discussion period, it is the proposal. The Adoption Paths framework already set the expectation that v4 runs in parallel with v3 for 24 to 36 months and that early TVL is dominated by migration flow, with incentives and rate differentials inflating borrows without creating net new credit demand, and Base is where that distortion will be largest because the existing V3 book there is one of the biggest Aave has.

Three shapes for the V3 posture, and the ARFC should pick one before Snapshot rather than after. Run V4 alongside V3 with no scheduled change and let the market migrate on its own, which forces no exits but leaves two venues quoting the same assets on one chain for years while rates diverge and depth thins unpredictably on one side. Deploy with a dated V3 posture change written into the same proposal, for example freezing new supply on V3 Base once V4 Hub caps clear a stated utilization threshold, which produces a clean end state but commits to a trigger before the Hub has any Base specific operating history. Deploy with a deliberately narrow initial scope covering the collateral types V3 serves worst, leaving the majors on V3 until the Hub has been through a real liquidation event, which slows headline TVL but keeps the migration flow small enough to actually read.

The third, paired with one reporting commitment: report supply migrated from V3 Base separately from net new supply from the first day the Hub is live. Without that split the Base deployment will produce the largest and least informative TVL number in the V4 rollout, and every later cap decision on Base will be argued from it.

It’s been two weeks, hasn’t there been a snapshot vote yet?

Hi Aave Labs,

I have been reviewing the proposed V4 deployment on Base. Since the Hub configuration, oracle setup, deployment contracts and V3 migration approach are still being finalized, has the security review plan for the Base specific deployment boundary also been defined?

A focused review could cover the Base Hub and Spoke configuration, oracle and asset onboarding controls, deployment parameters and the V3 migration path before the AIP is submitted.

I run Arctek Audits and can send a concise proposed scope mapped to the final contracts and commit if an independent reviewer has not yet been selected.

1 Like

Tokenized Equities on Aave V4 Base: Initial Market Parameters

Update, 24 September 2026. The full technical review of the Coinbase B20 issuance stack, the token standard, and the price feed has been posted in the Coinbase B20 Equities on Base Assessments thread. Section 4.6 below adds the risk steward bounds for the Equities Hub.

Summary

This post presents LlamaRisk’s recommended initial parameters for the tokenized equities market on Aave V4 Base. The market is deployed as a dedicated Equities Hub with a single pooled spoke, in which the Magnificent Seven US technology stocks in their Coinbase B20 form, B20 being Base’s native token standard (AAPLc, AMZNc, GOOGLc, METAc, MSFTc, NVDAc, and TSLAc) serve as collateral and USDC is the only borrowable asset. The equities themselves are not borrowable at launch. An in-depth review of the Coinbase B20 issuance stack is available in the assessments thread.

The initial stage operates on Chainlink’s 24/5 tokenized equity feeds, which publish from Sunday evening to Friday evening ET and hold the last published value over weekends and market holidays. While the underlying market is closed, the collateral value of a position does not change, its health factor can fall only through interest accrual, and the price published at the reopen reflects, in a single step, any information that arrived during the closure. The parameters are designed around this property. The liquidation bonus is sized on the cost of each pathway a liquidator can take to turn seized tokens back into USDC, since a liquidator may hold the collateral for hours or days while it redeems, sells, or hedges. Continuous monitoring of market and external data will be in place from launch.

This setup is an intermediate step. Chainlink is expected to bring 24/7 feeds for the same tokens to Base. Once those feeds are in production, the parameters may be revisited so that liquidations can execute through the weekend, with session-dependent parameters, automated handling of corporate actions, and feed safeguards. The full assessment of the issuer structure, the token standard, and the price feed is in the technical review linked above, and the parameter methodology and the rollout plan will be presented in a separate post.

1. Market Structure

Source: LlamaRisk, 17 September 2026

The market is deployed as a dedicated Equities Hub on Base. Stablecoin suppliers who join the hub opt into equity-collateralized lending explicitly, and the exposure does not extend to other Aave markets. The hub holds one USDC reserve, one lending spoke, and a supply-only tokenized spoke for USDC that gives vaults and aggregators a composable position without collateralization.

The lending spoke pools the seven names. A user can post any combination of the seven tokens as collateral and borrow USDC against it, with each token counted at its own collateral factor. Pooling reduces the volatility of position health for a diversified basket, since a single name can gap 20% to 30% on an earnings surprise while a basket of the seven rarely moves by double digits together. Collateral factors remain per token, so pooling does not by itself increase borrowing power. The equities are collateral only. Equity borrowing and equity-against-equity positions are not enabled.

2. Pricing Setup for the Initial Stage

US equities trade in sessions, while the lending market accepts deposits, borrows, and liquidations at any hour. The price feed connects the two, and the way it behaves while the stock market is closed determines which risks the parameters have to cover.

Each token is priced by a Chainlink tokenized equity feed that multiplies the price of the underlying share by the issuer’s token multiplier and publishes the product as a single total return price. The multiplier starts at 1.0 and moves only on corporate actions. A reinvested dividend raises it by the dividend net of the distribution fee and withholding tax, and a stock split multiplies it by the split ratio at the same time as the share price divides by that ratio. The market consumes the published value directly.

The underlying share price is assembled from the four sessions in which US equities trade, which gives the feed an operating window from Sunday 20:00 ET to Friday 20:00 ET.

Session Hours, ET Days Sourcing
Pre-market 04:00 to 09:30 Monday to Friday Extended-hours venues, several providers
Regular 09:30 to 16:00 Monday to Friday Exchange data, multi-sourced
Post-market 16:00 to 20:00 Monday to Friday Extended-hours venues, several providers
Overnight 20:00 to 04:00 Sunday evening to Friday morning Blue Ocean ATS
Closed 20:00 Friday to 20:00 Sunday, and US market holidays No updates, last value held

Onchain, the feeds are configured with a 0.5% deviation threshold and a 24-hour heartbeat, so during open sessions a new value is written whenever the offchain price moves 0.5% from the last published value. While the market is closed the feed publishes nothing, so the last value published before the Friday close stands, with a stale timestamp, until the Sunday evening reopen. The same applies on US market holidays.

Two consequences follow for the market. First, during a closure the collateral price does not move and a health factor can fall only through interest accrual. The market itself does not pause, and a position that crosses the threshold this way can be liquidated at any time, including over the weekend. Second, when the feed resumes, the price published at the reopen reflects whatever information arrived while the market was closed, and a position that this repricing pushes under water becomes liquidatable only at that point. In practice, the first opportunity to liquidate a position that deteriorated on a Friday afternoon is the Sunday evening reopen, and the first opportunity to hedge the seized collateral in a deep book is the Monday open.

During a corporate action, the issuer pauses its onchain registry flag, the feed holds its last value, the new multiplier is applied, and the feed resumes once the new underlying price multiplied by the new multiplier has been confirmed to match the held value. Minting and redemption pause for the same window while token transfers continue. The affected reserve is expected to be paused for that window as well, since a token that keeps changing hands against a held price is the state in which a mispriced liquidation could occur.

3. Parameters

3.1 Collateral Factor

In Aave V4 the collateral factor is both the borrowing limit and the liquidation threshold. On a 24/5 feed, the risk it is meant to cover is the fall the collateral may suffer between the moment a position becomes liquidatable and the moment a liquidator is able to close it. Outside regular cash hours the length of that interval is uncertain. A position that crosses the threshold at 17:00 could be closed minutes later on an extended-hours venue, during the overnight session, or only at the next regular open when a desk can hedge in a deep book. A position that crosses the threshold on a Friday afternoon is unlikely to be closed before the Sunday evening reopen, and may not be closed before the Monday open.

The collateral factor is therefore sized without assuming anything about that timing, other than that the liquidation is completed within five minutes of the next regular open. For every off-hours minute in the last eight years, a position is placed exactly at the liquidation threshold and the collateral factor is required to leave no bad debt if the liquidation executes at any live minute between that moment and five minutes past the next regular open. Weekend and holiday closures count toward the delay. Because the feed republishes only on a 0.5% move, the published price may sit up to 0.5% above the market, and that allowance is charged on every starting minute. The debt accrues interest at 24%, the top of the USDC rate curve, over the longest closure in the history, which is 92 hours and 35 minutes from a half-day close into a long weekend. Writing D for the fall from the published price to the lowest executable price and i for the interest accrued over that span, a full seizure at the 5.5% maximum bonus leaves no bad debt when

\mathrm{CF}\,(1 + \mathrm{LB})\,(1 + i) \le 1 - D.

The price history consists of one-minute bars for the seven names on Nasdaq from May 2018 to August 2026 across the pre-market, regular, and post-market sessions, supplemented by the overnight venue from August 2025, with seven stock splits corrected. Minutes in which the 24/5 feed would be frozen are excluded as execution minutes.

Source: LlamaRisk, Nasdaq and Blue Ocean one-minute bars, 20 August 2026

The largest fall in the record gives a first reading of the collateral factor for each name. For AMZN, GOOGL, META, and NVDA it is a post-market earnings reaction. For AAPL and MSFT it is the evening before the March 2020 weekend, and for TSLA the evening before the Labor Day weekend of 2020. Each of these falls is a single observation, and the record carries no information about a fall rarer than it contains. To address this, an extreme-value tail is also fitted to the bad nights of each name and read at the fall expected once in ten years, once on each name alone and once jointly across the seven names with a shared tail shape. The two fits can disagree for a name whose own tail departs from the class, and the record cannot resolve which is right, so the lowest of the three readings is adopted.

Source: LlamaRisk, 16 September 2026

Name Collateral factor
AAPLc 0.78
AMZNc 0.73
GOOGLc 0.76
METAc 0.65
MSFTc 0.79
NVDAc 0.70
TSLAc 0.65

The collateral factors will be refitted as the record grows, and they are the first parameters expected to be revisited in the second stage.

3.2 Liquidation Bonus

A liquidator on this market cannot always dispose of the collateral as soon as it is seized. Redemption with the issuer is open only to vested holders and settles during regular hours, the secondary market on Base is thin, and outside regular hours the tokens can only be hedged until the next open. The bonus is therefore sized to compensate a liquidator for the full pathway from repaying the debt to holding USDC again, whatever form that pathway takes on the day. The net bonus a route requires is

b_n = \frac{1 + \delta}{(1 - d)\,(1 - k)} - 1,

where \delta is the allowance between the published price and the market at commitment, set at 0.5% from the feed’s deviation threshold, d is the move of the stock between commitment and the moment the hedge is in place, taken as a stressed 2.02%, and k is the sum of the cash costs of the route as a fraction of the hedge notional. The gross bonus paid by the borrower is b_n / (1 - \rho), with \rho the 10% liquidation fee. Three routes are available, and they differ mainly in k.

Redemption through the issuer. During regular market hours, a liquidator with redemption access repays the debt, receives the tokens, shorts the same number of underlying shares, submits an in-kind redemption, and delivers the redeemed shares against the short. The cash costs are the 5 bps redemption fee, brokerage, clearing, stock borrow, and the opportunity cost of the USDC over a 72-hour hold, partly offset by interest on the short-sale proceeds. This is the least expensive route and the floor for any bonus.

Sale on the secondary market. A liquidator that has not been whitelisted with the issuer cannot redeem and sells the tokens on Base instead. In this case k is the price impact of the sale. Between USD 0.3M and USD 1.1M of each token can be sold into USDC within a 2% price impact, so a liquidation in the low hundreds of thousands of dollars costs about 1% and one approaching the depth limit about 2%. Larger liquidations would have to be split across blocks and days or routed through a party that can redeem.

Perpetual hedge outside regular hours. Between the post-market close and the next open, a liquidator shorts a linear perpetual with matched share exposure, holds it until the redeemed shares can be sold at the open, and then closes both legs. In addition to the redemption costs, this route pays the basis between the perpetual and the stock when the legs are closed, funding while the short is open, and taker fees. The stress case takes a 0.5% basis at exit, 0.85% of funding and 0.15% of fees.

Route Oracle allowance Hedge latency stress Route cost k Net bonus required Gross bonus at a 10% fee
Redemption, regular hours 0.50% 2.02% 0.06% 2.64% 2.93%
Secondary sale on Base 0.50% 2.02% 2.00% 4.67% 5.18%
Perpetual hedge, stressed 0.50% 2.02% 1.50% 4.13% 4.59%

The maximum liquidation bonus is set at 5.5% for every name, so that each of the three pathways is paid for. It covers the redemption route with a wide margin and the stressed perpetual route with a comfortable one, and it covers the secondary-market route up to the depth Base offers today. It is in line with the 5.55% maximum bonus on the BTC and WETH reserves of the Main Spoke on Ethereum. The spoke uses the same bonus curve as the Main Spoke. With a 90% liquidation bonus factor the bonus is 4.95% at a health factor of 1.0, which already covers the redemption and perpetual routes, and reaches the 5.5% maximum at a health factor of 0.90, above the point at which any of the seven collateral factors would produce bad debt. A liquidation restores the position to a health factor of 1.24. Liquidations on this market do not run through Smart Value Recapture at rollout, so the liquidator keeps the whole bonus after the 10% fee.

3.3 Caps

The collateral factor and the bonus protect the market against price moves, and the caps protect it against size. A liquidator that seizes these tokens may hold them for hours or days while it redeems, sells, or hedges, so the caps keep the amount a liquidator may have to absorb within the depth of the venues it can use.

  • Redemption through the issuer is available to allowlisted authorised participants and carries no onchain rate limit. Primary throughput is observable on the minting side: the issuer’s supply manager contract on Base (0xd1ca…664c) limits the amount of each token its minter can create per 24 hours, which corresponds to about USD 5M per token per day at current prices.
  • The secondary market on Base is thin. Between USD 0.3M and USD 1.1M of each token can be sold into USDC within a 2% price impact, and the outstanding supply of every token is between USD 1M and USD 4M.
  • Perpetual futures are the venue in which an off-hours liquidation is hedged. Across Hyperliquid, Binance, OKX, and Lighter, the seven names carry between USD 44M and USD 302M of open interest each, most of it on Hyperliquid and Binance.

Each add cap is set so that a liquidation of the entire cap could be processed through the issuer within one day of its mint allowance and hedged on the perpetual venues within a small fraction of the name’s open interest. For AAPL, GOOGL, NVDA, and TSLA one day of the mint allowance is the lower of the two bounds and sets the cap, for META 5% of open interest sets it, and for AMZN and MSFT the caps sit at about half of one day’s mint allowance and 6% of open interest. Together the caps amount to USD 29.3M of collateral. A loss-budget view of the same book would allow more at these collateral factors, so liquidity is the binding constraint at launch.

Reserve Perp open interest, four venues 5% of open interest Mint allowance per day DEX depth at 2% impact Supply outstanding Add cap Add cap, USD
AAPLc USD 125.4M USD 6.27M 15,500 (USD 5.14M) USD 0.72M 6,077 (USD 2.02M) 15,000 USD 4.98M
AMZNc USD 45.2M USD 2.26M 19,100 (USD 4.69M) USD 0.32M 5,642 (USD 1.39M) 10,500 USD 2.58M
GOOGLc USD 178.1M USD 8.90M 15,700 (USD 5.36M) USD 0.69M 5,914 (USD 2.02M) 15,000 USD 5.13M
METAc USD 79.1M USD 3.96M 8,200 (USD 5.55M) USD 0.64M 2,365 (USD 1.60M) 5,800 USD 3.93M
MSFTc USD 44.4M USD 2.22M 10,300 (USD 5.06M) USD 0.32M 2,652 (USD 1.30M) 5,200 USD 2.56M
NVDAc USD 301.5M USD 15.08M 24,000 (USD 5.11M) USD 1.08M 19,428 (USD 4.14M) 24,000 USD 5.11M
TSLAc USD 103.0M USD 5.15M 14,300 (USD 5.10M) USD 0.27M 2,988 (USD 1.07M) 14,000 USD 4.99M

Source: LlamaRisk, Hyperliquid, Binance, OKX, Lighter, KyberSwap, and Base onchain data, 17 September 2026

The USDC draw cap on the spoke corresponds to the debt the collateral caps can support at their collateral factors, USD 21.1M, rounded to USD 21M. The USDC add cap is set at USD 32M, so that supply can run ahead of the draw cap and keep utilization within the optimal range. Several caps are above a token’s outstanding supply today. The tokens are minted on demand by authorised participants, and supply is expected to follow the collateral demand of the market. The caps will be revisited through the risk steward process as supply and venue depth grow.

3.4 USDC Interest Rate

USDC is the only borrowable asset. Its rate follows the standard kinked curve, with a 0% base drawn rate, 4% rate growth before the optimal usage ratio and 20% after it, at a 90% optimal usage ratio and a 10% liquidity fee. The borrow rate therefore peaks at 24%, which is the rate charged on the debt over the longest closure in the collateral factor derivation. The optimal usage ratio is set at 90%, below the 92% used on deeper stablecoin reserves, because the hub starts with a single borrowable reserve and no credit line from a general market.

4. Specification

4.1 Dynamic Liquidation Bonus Configuration

Chain Hub Spoke Liquidation Bonus Factor Target Health Factor Health Factor For Max Bonus
Base Equities Hub Mag-7 Spoke 90.00% 1.24 0.90

4.2 Spoke Parameters

The liquidation fee is 10%, consistent with the rest of Aave V4. The risk premium threshold is 0 for every reserve, as risk premiums are not in use, and liquidators can receive collateral as shares on every reserve.

Chain Hub Spoke Reserve Collateral Factor Max Liquidation Bonus Borrowable Collateral Risk Liquidation Fee Risk Premium Threshold Receive Shares
Base Equities Hub Mag-7 Spoke AAPLc 78.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke AMZNc 73.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke GOOGLc 76.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke METAc 65.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke MSFTc 79.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke NVDAc 70.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke TSLAc 65.00% 5.50% FALSE 0 10.00% 0 TRUE
Base Equities Hub Mag-7 Spoke USDC 0.00% - TRUE - - 0 TRUE

4.3 Add and Draw Caps

Chain Hub Spoke Reserve Add Cap Draw Cap
Base Equities Hub Mag-7 Spoke AAPLc 15,000 0
Base Equities Hub Mag-7 Spoke AMZNc 10,500 0
Base Equities Hub Mag-7 Spoke GOOGLc 15,000 0
Base Equities Hub Mag-7 Spoke METAc 5,800 0
Base Equities Hub Mag-7 Spoke MSFTc 5,200 0
Base Equities Hub Mag-7 Spoke NVDAc 24,000 0
Base Equities Hub Mag-7 Spoke TSLAc 14,000 0
Base Equities Hub Mag-7 Spoke USDC 32,000,000 21,000,000
Base Equities Hub Tokenized USDC Spoke USDC 1,000,000 0

4.4 Interest Rate Curve

Chain Hub Asset Base Drawn Rate Rate Growth Before Optimal Rate Growth After Optimal Optimal Usage Ratio Liquidity Fee
Base Equities Hub USDC 0.00% 4.00% 20.00% 90.00% 10.00%

4.5 Price Feeds

Each reserve is priced by the Chainlink total return feed for the token, which carries the issuer multiplier and follows the 24/5 US equities market hours. These are standard feeds. Smart Value Recapture variants are not used at rollout. USDC is priced by the Chainlink USDC/USD feed on Base through the stable price cap adapter, with the cap set at 1.04.

4.6 Risk Steward Bounds

The Aave Risk Stewards operate on the Equities Hub within the bounds below. They follow the bounds activated on Aave V4 Ethereum and Avalanche in the ARFC to activate the Aave Risk Stewards on Aave V4, with a shorter cooldown on caps and on the spoke parameters. Caps move on a 12-hour cooldown because the market starts small, at about USD 29.3M in collateral caps, and we plan to keep the caps close to the supplied and drawn amounts and raise them as demand shows up. Two cap changes a day at fixed times keep signing predictable, and caps can come down overnight or over a weekend if AMM liquidity thins. The collateral factor, the maximum liquidation bonus, and the liquidation configuration share a 36-hour cooldown so that every spoke parameter moves on one cadence, with the maximum change per update unchanged from the other V4 instances.

Scope Parameter Cooldown Max change per update Mode Note
Hub Add cap 12 hours 100% Relative 36 hours on V4 Ethereum and Avalanche
Hub Draw cap 12 hours 100% Relative 36 hours on V4 Ethereum and Avalanche
Hub Optimal usage ratio 36 hours 3% Absolute
Hub Base drawn rate 36 hours 3% Absolute
Hub Rate growth before optimal 36 hours 3% Absolute
Hub Rate growth after optimal 36 hours 20% Absolute
Spoke Collateral risk 36 hours 300% Absolute
Spoke Collateral factor, update 36 hours 0.5% Absolute 72 hours on V4 Ethereum and Avalanche
Spoke Collateral factor, new reserve 36 hours 5% Absolute 72 hours on V4 Ethereum and Avalanche
Spoke Max liquidation bonus, update 36 hours 0.5% Absolute 72 hours on V4 Ethereum and Avalanche
Spoke Max liquidation bonus, new reserve 36 hours 0.5% Absolute 72 hours on V4 Ethereum and Avalanche
Spoke Target health factor 36 hours 5% Relative 72 hours on V4 Ethereum and Avalanche
Spoke Health factor for max bonus 36 hours 5% Relative 72 hours on V4 Ethereum and Avalanche
Spoke Liquidation bonus factor 36 hours 5% Absolute 72 hours on V4 Ethereum and Avalanche
Oracle Stable price cap 72 hours 0.5% Relative Applies to USDC, cap at 1.04
Oracle LST price cap 72 hours 5% Relative Not used
Oracle Pendle discount rate 48 hours 0.025 Absolute Not used

Disclaimer

This review was independently prepared by LlamaRisk, a DeFi risk service provider funded in part by the Aave DAO. LlamaRisk is not directly affiliated with the protocol(s) reviewed in this assessment and did not receive any compensation from the protocol(s) or their affiliated entities for this work.

The information provided should not be construed as legal, financial, tax, or professional advice.

2 Likes

The ARFC to Deploy Aave V4 on Base has been raised to snapshot. Voting will begin in less than 24 hours. You may vote here

1 Like

Aave Labs is finalizing preparations to activate Aave V4 on Base and wants to give the community formal advance notice ahead of execution. The first market is the Equities Hub, holding seven Coinbase issued tokenized equities (AAPLc, AMZNc, GOOGLc, METAc, MSFTc, NVDAc, TSLAc) alongside USDC, with a MAG7 Spoke where the equities are collateral only and not borrowable. USDC is the sole borrowable asset. The activation also enables the Equities USDC Tokenization Spoke, which acts as an ERC-4626 entry point for USDC deposits across all spokes of the USDC reserve on the Equities Hub. The market is deployed and fully configured with every spoke registration halted, so no supply, borrow, or liquidity movement is possible until activation.

The ARFC to deploy Aave V4 on Base has been raised to Snapshot, using the risk parameters LlamaRisk recommended in this thread. The Snapshot vote is taken as the binding vote for this execution, and activation proceeds only if the Snapshot result is positive.

This activation does not take the form of an AIP and carries no Aave Governance V3 vote. The Aave V4 Protocol Security Council 0x187AAE17d4931310B3fc75743e7F16Bdc9eD77e9 (5-of-8, the same signer set as on Ethereum) carries it out directly through its Executor 0xA9D9923A1ADC1200771aaaA38CFeD6A5b8483d70, which holds HUB_CONFIGURATOR_DOMAIN_ADMIN_ROLE on the Base V4 AccessManager. The deployed configuration was verified onchain against LlamaRisk’s recommendations.

These addresses are added to the aave-address-book and permissions-book. For transparency, the table below lists the price feed used for each Hub reserve, following LlamaRisk’s recommendations. The USDC reserve uses the Chainlink USDC/USD feed rather than an SVR feed:

Reserve Price source Kind Feeds
AAPLc 0x787f13dEa48Db0897CbCDD985de77809D837F988 Chainlink OCR2 proxy Coinbase AAPL/USD
AMZNc 0x06A8E4b3aBB3B7543d8396FB2B763d22820cB295 Chainlink OCR2 proxy Coinbase AMZN/USD
GOOGLc 0x5bF49E0ffA937CE2FfF033c739aD7C634c4D34F2 Chainlink OCR2 proxy Coinbase GOOGL/USD
METAc 0x6526aE6797A76123638b863AeE4dD27Ba4E4b27D Chainlink OCR2 proxy Coinbase META/USD
MSFTc 0xeB10A6c9aa7E537aEd766C08c35Dae35B321b18c Chainlink OCR2 proxy Coinbase MSFT/USD
NVDAc 0x04689a41629776563E6822F76f2e57D148d28513 Chainlink OCR2 proxy Coinbase NVDA/USD
TSLAc 0xFaf869185383a24F8cb00e27BdA6b63B9905DCb4 Chainlink OCR2 proxy Coinbase TSLA/USD
USDC 0xC7d0f8dCC1F860ca752054c59Ea82Ba2A5AaB50c PriceCapAdapterStable Chainlink USDC/USD, cap 1.04

Based on LlamaRisk’s recommendation in this thread, a different configuration compared to that of Ethereum and Avalanche (see Snapshot) is applied for the Risk Stewrads in the Equities Hub. They would operate within the same bounds as on those networks, but with shorter cooldowns on caps (12 hours instead of 36 hours) and on spoke parameters (36 hours instead of 72 hours), while the maximum change per update is unchanged. These configurations will be implemented through the AIP process in the upcoming days.

At execution, the payload clears the halted flag on every spoke registration across the Hub, making the MAG7 Spoke and the Equities USDC Tokenization Spoke fully operational on Base. Caps and oracle configuration are unchanged from deployment, and the action changes no roles, owners, or oracles.

References

2 Likes

Aave V4 on Base has been successfully activated. The Protocol Security Council lifted the temporary halt placed on the market at deployment, following the same execution described in the previous update, and the market is now fully operational.

The instance is live and accessible through pro.aave.com, with USDC, APLc, AMZNc, GOOGLc, METAc, MSFTc, NVDAc and TSLAc available as reserves from launch, per the configuration and price feeds shared previously in this thread.

3 Likes

On the line that the caps “will be revisited through the risk steward process as supply and venue depth grow”: I measured that depth at the venue level and published it, so that review has a starting number rather than a proxy. Pinned block 52,175,000; reads pinned by block hash with requireCanonical; both legs priced against the same state; a size ladder, a failure class per row, and the raw read logs in the same directory: https://github.com/dirkdiggler1026/executability-report/tree/main/measurements/base-depth

The finding that matters for sizing: the same underlying is not the same venue. At that block two NVDAc/USDC pools differ by roughly 600× on a USD 100 exit — 99.31% in one pool, and in the other no more than 0.168% (0.167746%, an on-chain holdings bound: that is all the pool holds; the walk value 0.166811% is shown beside it and is not the claim). A cap set from TVL or from a 2%-impact figure sees neither the difference nor its direction.

Two scope notes. In Panel A — the rows whose claim is the measured curve — the largest tier is USD 1,000; the largest row published at all is a USD 10,000 holdings bound. Tiers above that are measured but withheld pending one validation, with the reason stated in the disclosure rather than left implicit. And I have not measured this Hub’s assets on any other chain.

Would it be useful if I ran the same for the other assets in this Hub and sent the rows as a table — or is this already covered by the venue data behind your assessment?

A correction to my own wording above, which claimed more than the artifacts hold.

I wrote that tiers above USD 1,000 are “measured but withheld pending one validation”. For five of the 43 rows that is not what happened: their reads failed with HTTP 429 on tickBitmap, on the rows that issue the most reads, and they have no figure at all. The two runs lost different rows, so their absence is a property of my read budget in that run rather than of any pool. The README lists them as rows with no number; my sentence did not.

I also called reads-Y1.json / reads-Y2.json “the raw read logs”. They are a sample: each run issued about 2,450 calls and each file keeps the first 400, with the head of each response rather than the response. They show which rows failed and why, but they cannot recompute the figures. That is now stated in DISCLOSURE.md beside the files, along with what can still be rechecked - the chain at that block hash, while the roughly 30-day provable window still serves it - and what changes from the next run on: the complete list, each entry with its request parameters and the sha256 of its full response, and the full responses shipped beside it.

The finding is unchanged and so is the table: 99.31% and the 0.168% holdings bound are rows that read cleanly in both runs and were identical across them. Both corrections are about wording, not figures.