Skip to content
DeFi Reference

Modular lending

1. TL;DR

Section titled “1. TL;DR”

Modular lending splits what used to be one governance-run pool into a minimal, permissionless market primitive — pick a collateral asset, a loan asset, an oracle, a liquidation LTV, and an interest-rate model, and the market exists, immutable, isolated from every other market — plus a separate layer of curators or vaults who decide which of those markets are worth depositing into. Morpho Blue and Euler v2 (rebuilt after a 2023 hack) both take this approach on Ethereum; Aave V4 instead keeps one shared liquidity pool per network but modularizes the borrower-facing “Spokes” that draw from it. Kamino’s curator and market-creation tooling brings a similar isolated-market-plus-vault pattern to Solana.

2. Explain it simply

Section titled “2. Explain it simply”

Analogy

Section titled “Analogy”

A pre-modular lending pool is a single department store where head office decides every product on every shelf, and a defect on one shelf can force the whole store to close. Modular lending is a shopping mall instead: anyone can rent an empty unit and open a tiny, single-purpose shop with their own rules, a defect in one shop doesn’t shut the mall, and most shoppers don’t wander the mall themselves — they hire a personal shopper (a curator) who already knows which shops are trustworthy and puts their money there for them.

ELI5

Section titled “ELI5”

Imagine instead of one huge shared piggy bank with one set of rules for everyone, anyone can set up their own tiny piggy bank with its own rule: “this bank only takes toy cars as security and only lends out coins.” Because each tiny bank is separate, if one bank’s rule turns out to be a bad idea, only the people who used that specific bank are affected — everyone else’s coins in other tiny banks are safe. Most people don’t want to pick a tiny bank themselves, so they let a trusted helper spread their coins across several tiny banks the helper has already checked out.

Step-by-step walkthrough

Section titled “Step-by-step walkthrough”

Scenario: creating a Morpho Blue market and having a curated vault allocate into it (illustrative, modeled on Morpho Blue’s published market parameters and Kamino’s near-identical curator flow).

  1. Before any market exists. Nobody has deployed a WETH-collateral / USDC-loan market with an 86% liquidation LTV yet; the primitive contract exists, but no state for this combination does.
  2. Anyone creates the market. A market creator specifies five immutable parameters: collateral (WETH), loan asset (USDC), oracle, liquidation LTV (86%), interest-rate model. The market now exists permanently — nobody, including the creator, can ever change these parameters again.
  3. A curator’s vault allocates into it. A vault (an ERC-4626 token depositors already trust) reviews the new market and routes, say, 40% of its assets there and the rest into other vetted markets — depositors never had to evaluate the WETH/USDC market themselves.
  4. A lender deposits into the vault, not the market. Alice deposits 10,000 USDC into the curator’s vault, gaining diversified exposure to whichever mix of isolated markets the curator has chosen.
  5. A borrower interacts with the market directly. Bob deposits WETH and borrows USDC — his position’s risk is contained entirely within that one market, isolated from every other market the vault has deployed to.
  6. If the WETH/USDC market’s oracle or LTV turns out to be poorly chosen, only that market’s lenders (including the fraction of the vault allocated there) are exposed — not the whole protocol.

Common misconceptions

Section titled “Common misconceptions”
  • Myth: “Modular” always means many small isolated markets, the way Morpho and Euler do it. Reality: Aave V4 is also a modular redesign, but keeps one shared Liquidity Hub per network and modularizes the “Spokes” instead — a genuinely different axis of modularity (Aave, “Understanding Aave V4’s Architecture,” 2025-06-12).
  • Myth: A permissionless, isolated market is automatically safer than a governed pool. Reality: Isolation only contains a bad market’s blast radius to whoever interacts with it; it does nothing to stop a curator from allocating a vault’s funds into a bad market on depositors’ behalf.
  • Myth: Morpho Blue itself decides which assets are safe to lend against. Reality: “The protocol itself makes no choices about which assets are credit-worthy or how capital should be allocated. Those decisions are deliberately left to lenders” (Morpho Midnight whitepaper, 2026-05, §1).
  • Myth: Euler’s 2023 hack means isolated-market designs are inherently unsafe. Reality: The hack hit Euler’s original, pre-modular architecture; its post-hack relaunch doubled down on isolation, rebuilding around the Euler Vault Kit so that “Euler markets can compose without forcing every asset into one shared risk pool” (Euler Finance blog, fetched 2026-08-29).
  • Myth: Cross-vault composability is the same thing as the old aggregated-pool model. Reality: Euler’s Vault Connector lets selected vaults recognize each other, with builders retaining “control over which vaults connect,” unlike a pool where every listed asset is automatically cross-collateral for every user (Euler Finance blog, fetched 2026-08-29).

If you only remember one thing

Section titled “If you only remember one thing”

Modular lending separates “does this specific market’s parameters make sense” (a permissionless, isolated, immutable primitive) from “which markets should my money actually go into” (a curated vault layer) — two questions that a single aggregated pool used to answer with one governance vote.

3. How it works

Section titled “3. How it works”

The minimal isolated market primitive

Section titled “The minimal isolated market primitive”

Morpho Blue’s core insight was to strip a lending market to the smallest set of parameters that still make it a market: “Morpho Blue proposed a different architecture for onchain credit based on isolated and immutable markets with permissionless creation” (Morpho Midnight whitepaper, 2026-05, §1). Each market is defined once, at creation, by a fixed collateral asset, loan asset, oracle, liquidation LTV, and interest-rate model (/lending/interest-rate-models/, /lending/collateral-ltv-health/) — none of it can change afterward. Because the market never had governance-adjustable parameters, there is no admin key, oracle-swap vote, or LTV change to attack; the entire risk surface is fixed at deployment.

Curators and vaults: where the actual risk decisions live

Section titled “Curators and vaults: where the actual risk decisions live”

Because a bare isolated market is too fragmented for most depositors to evaluate one at a time, the practical lending experience happens one layer up, through curated vaults: “much of the supply comes from vaults built on top of the protocol. The market layer therefore stays minimal, while curation and allocation become a competitive layer above it” (Morpho Midnight whitepaper, 2026-05, §1). A vault is an ERC-4626 token: depositors hold one fungible share, and a curator (a team, a DAO, or an algorithm) splits the vault’s assets across many isolated markets, re-creating diversification without a shared risk pool underneath. Kamino runs a close Solana analogue: curators who “deploy vault,” “set allocation strategy,” “route assets,” and separately “create market” as “an isolated lending market with your own risk parameters” (Kamino docs, 2026-08 fetch) — market creation and curation are explicitly two separate steps, the same split Morpho formalizes.

Composable isolation: the Euler Vault Kit and Vault Connector

Section titled “Composable isolation: the Euler Vault Kit and Vault Connector”

Euler v2’s relaunch is built on two primitives. The Euler Vault Kit (EVK) lets “market builders… launch ERC-4626 lending vaults with configurable collateral rules, borrow and supply caps, oracle inputs, interest rate models, liquidation settings, governance paths, and operating roles” (Euler Finance blog, “Introducing Euler V2,” fetched 2026-08-29) — each vault is its own isolated market, packaged as a standard ERC-4626 vault. The Ethereum Vault Connector (EVC) adds controlled composability: it lets vaults “recognize one another as collateral, coordinate account checks, delegate scoped permissions, and batch multi-step actions” (Euler Finance blog, fetched 2026-08-29). This composability is opt-in per vault pair, not automatic the way collateral recognition is inside an Aave-style pool — Euler frames it as solving a real tension: “a single shared pool cannot fit every asset. Fully isolated markets can limit composability and fragment liquidity” (Euler Finance blog, fetched 2026-08-29).

The other axis of modularity: Aave V4’s Hub-and-Spoke

Section titled “The other axis of modularity: Aave V4’s Hub-and-Spoke”

Aave took a structurally different path to the same complaint — that a single Aave V3 market “can’t be used to meet borrowing demand in another” — without fragmenting into fully isolated markets. In V4, “all assets are stored in a unified Liquidity Hub” per network, which “tracks which Spokes are authorized to access which assets and enforces limits on how much liquidity each Spoke can draw,” while users interact only through Spokes, each with “its own rules and risk settings” — an E-Mode Spoke, an Isolation Spoke, an RWA Spoke for tokenized treasuries, or a Vault Spoke letting a user “borrow against assets held in a Safe” without ever supplying them to the Hub (Aave, “Understanding Aave V4’s Architecture,” 2025-06-12). The Hub still pools liquidity — Spokes draw from one shared source — so risk containment comes from per-Spoke draw limits, not market-level isolation.

4. Worked numeric example

Section titled “4. Worked numeric example”

A curator’s vault allocates 10,000,000 USDC of depositor capital across three isolated Morpho-Blue-style markets (illustrative parameters):

MarketCollateralAllocationSupply APY
AWETH, LLTV 86%4,000,000 USDC (40%)5.0%
BwstETH, LLTV 94.5%4,000,000 USDC (40%)3.5%
Ca newly listed, higher-risk token, LLTV 62%2,000,000 USDC (20%)9.0%

The vault’s blended supply APY, before the curator’s own performance fee, is

Rˉ=0.40×5.0%+0.40×3.5%+0.20×9.0%=2.0%+1.4%+1.8%=5.2%\bar{R} = 0.40\times5.0\% + 0.40\times3.5\% + 0.20\times9.0\% = 2.0\% + 1.4\% + 1.8\% = 5.2\%

If Market C’s collateral collapses to zero, generating bad debt equal to its entire 2,000,000 USDC allocation, only vault depositors absorb that loss — proportionally, 20% of the vault’s assets — while lenders who supplied Market A or B directly (outside this vault) are entirely unaffected, since the markets share no collateral or liquidity. That is the isolation guarantee in numbers: the loss stays contained to whoever chose exposure to that specific market, not spread across the whole protocol the way an Aave V2/V3-style aggregated pool would spread it.

5. Where it’s used

Section titled “5. Where it’s used”

Ethereum

Section titled “Ethereum”
  • Morpho Blue — the reference minimal isolated-market primitive described throughout this page; vaults built on top (often called MetaMorpho vaults) are the primary way most users actually supply. docs.morpho.org
  • Euler v2 — relaunched with the EVK and EVC after the March 2023 hack (below), live since 2024 (secondary: Blockworks, CryptoBriefing, 2024, accessed 2026-08-29). euler.finance/blog
  • Aave V4 — in development as of the June 2025 architecture post; Hub-and-Spoke rather than isolated markets, described in §3. aave.com/blog

Solana

Section titled “Solana”
  • Kamino Lend curators — permissionless-style “Create Market” plus a separate vault/curator allocation flow, closely mirroring Morpho’s market-plus-vault split (Kamino docs, 2026-08 fetch).
  • marginfi (rebranded Project 0 / P0) — a related but distinct approach: rather than a vault allocating across isolated markets, P0 lets a single margin account hold “cross-venue collateral” from “multiple DeFi venues” (including Kamino) “in a single unified margin account” (marginfi/P0 docs, 2026-08 fetch) — composability at the account level, not the vault-allocation level.

6. Risks, attacks, and incidents

Section titled “6. Risks, attacks, and incidents”
  • Euler Finance hack, 13 March 2023 (Ethereum). Euler’s original (pre-modular) protocol had a donateToReserves function, added via a governance-approved update, letting a caller donate collateral to the protocol’s reserves — but, unlike every other fund-moving path, it carried no health/solvency check. Euler Labs’ own post-mortem is direct: “the function lacked a health check to make sure that an existing user wouldn’t be able to put themselves in an unhealthy state just by donating to the reserves” (Euler Labs blog, “War & Peace,” accessed 2026-08-29). The attacker donated collateral to break their own health factor, then self-liquidated for a bonus exceeding what they’d donated, repeating the pattern for a reported ~$197M loss (as of 2023-03; widely corroborated: Bloomberg, Elliptic, Chainalysis). After roughly three weeks of on-chain negotiation — “the simplest way to move forward today is to return 90% of both the DAI and ETH under your control to the EulerDAO treasury address… without needing to go the legal route” (Euler Labs blog, accessed 2026-08-29) — the attacker returned the funds by 3 April 2023, no prosecution pursued. Euler rebuilt from the ground up as the modular EVK + EVC architecture in §3, relaunching as Euler v2 in 2024 — this page’s central case study in how a single missing check becomes existential, and why the industry’s response was more isolation.
  • Curator risk reappears one layer above the market. Because most depositors interact with a vault rather than a bare market, the curator becomes a new trust dependency: allocating too much into a poorly parameterized market re-creates concentrated risk even though the underlying markets remain individually isolated (worked numerically in §4).
  • Connector-mediated contagion. Euler’s EVC lets vaults recognize each other as collateral — a feature, but also a reintroduced contagion path if connected carelessly; Euler frames this as a builder responsibility (“builders maintain control over which vaults connect”), not a protocol-enforced guarantee (Euler Finance blog, fetched 2026-08-29).
  • Aave V4’s Hub concentrates a different kind of risk. Because every Spoke draws from one shared Liquidity Hub, a Hub-level bug or under-capped malicious Spoke threatens every Spoke simultaneously — the same aggregated-pool risk isolated-market designs avoid, just moved to a different layer (Aave, 2025-06-12).
  • Immutability cuts both ways. An isolated market’s parameters cannot be patched if a flaw is later discovered — only abandoned — unlike Aave’s governance-adjustable pool parameters (see /lending/interest-rate-models/, §6).

7. Open problems

Section titled “7. Open problems”
  • Does market-level isolation actually reduce systemic risk, or just relocate it to curators? None of this page’s sources resolve whether capital flowing through a few dominant curated vaults ends up any less concentrated than the aggregated pools modular lending replaced.
  • Which modularity axis wins: isolated markets (Morpho, Euler) or a shared Hub with capped Spokes (Aave V4)? Aave’s justification — “liquidity in one market can’t be used to meet borrowing demand in another” under full isolation — is a real capital-efficiency argument neither Morpho’s nor Euler’s materials directly rebut in this page’s sources (Aave, 2025-06-12).
  • How much cross-vault/cross-market connectivity is safe? Euler’s EVC and Aave V4’s Spoke-to-Hub draw limits both bet on bounding contagion through a connector; whether either bound is calibrated correctly hasn’t been tested by a comparable-scale incident yet.
  • Governance-free security. Isolated markets have no admin keys to attack, but push all safety-critical judgment onto whoever creates and allocates to the market — a shift Morpho treats as a feature (“the protocol itself makes no choices”) that leaves open whether creators and curators are, in aggregate, better risk-assessors than governance was.

8. Ethereum vs Solana

Section titled “8. Ethereum vs Solana”
AspectEthereumSolana
Isolated-market primitiveMorpho Blue (5 immutable parameters); Euler v2’s EVK vaultsKamino “Create Market” — isolated market with custom risk parameters
Curation layerMetaMorpho-style vaults (ERC-4626) allocating across marketsKamino curator vaults (deploy, allocate, route, earn)
Cross-market/vault composabilityEuler’s EVC — opt-in mutual collateral recognition between vaultsmarginfi/P0’s cross-venue collateral — one account spanning multiple protocols’ positions
Alternative to full isolationAave V4 Hub-and-Spoke — one shared pool, capped modular SpokesNo Solana source in this cluster describes an equivalent Hub-and-Spoke redesign
Best-documented incident driving the designEuler Finance hack, 2023-03-13 — prompted a full modular rebuildNone of this page’s sources document a comparable Solana modular-lending incident

The two ecosystems reach for similar tools — an isolated market primitive plus a curation layer — but Ethereum’s version grew out of a specific trauma (Euler’s 2023 hack) and treats immutability as the safety feature, while Solana’s Kamino curator system reads as a natural product extension of an already-pooled design rather than a reaction to a comparable incident.

9. Reference doc

Section titled “9. Reference doc”

The reference

Section titled “The reference”

Understanding Aave V4’s Architecture — Aave, 12 June 2025. aave.com/blog

Summary of the reference

Section titled “Summary of the reference”

The post opens by reviewing Aave V3’s limitation as motivation: each market (like Ethereum “Core” and “Prime”) “has its own isolated liquidity pool,” so “liquidity in one market can’t be used to meet borrowing demand in another,” creating “friction when launching new markets” and fragmenting liquidity across chains and market types.

The Liquidity Hub is V4’s answer: “in V4, liquidity is no longer siloed by market. Instead, all assets are stored in a unified Liquidity Hub per network.” The Hub is the sole source of truth for accounting — it “tracks which Spokes are authorized to access which assets and enforces limits on how much liquidity each Spoke can draw,” and enforces the core invariant that total borrowed can never exceed total supplied. Users never touch the Hub directly; it “works behind the scenes” using “a share-based system” for efficient interest accounting.

Spokes are the user-facing layer: “the component of Aave V4 that users interact with directly,” each with “its own rules and risk settings,” optimized for a particular asset class or use case, and each managing its own oracle integration, collateral tracking, and pause controls. The post walks through four illustrative Spoke types: an E-Mode Spoke for correlated-asset high-LTV borrowing; an Isolation Spoke for newer/riskier collateral behind a Hub-enforced draw cap; an RWA Spoke for tokenized real-world assets like treasury bills, with “stricter access, custody, or redemption rules”; and a Vault Spoke letting a user borrow against assets held in something like a Safe without ever supplying them to the Hub, since “risk management and collateralization are handled entirely within the Spoke.”

The post closes on stated benefits — capital is “no longer fragmented across markets on the same chain,” governance becomes “lighter” since Spokes can be added or upgraded without disrupting the rest of the system, and “anyone can build a Spoke,” turning the Hub’s liquidity into a shared credit line for permissionless innovation, much as Morpho Blue’s minimal market primitive does for isolated markets, but without giving up pooled liquidity.

Key quotes

Section titled “Key quotes”

“Liquidity in one market can’t be used to meet borrowing demand in another.” (§Aave V3 Basics)

“In V4, liquidity is no longer siloed by market. Instead, all assets are stored in a unified Liquidity Hub per network.” (§Liquidity Hub Overview)

“Each Spoke has its own rules and risk settings. They can be optimized for stablecoins, staked ETH derivatives, or higher-risk assets.” (§Spokes Overview)

“A Liquidity Hub would determine how much liquidity is made available to the Isolation Spoke by setting a cap on the amount it can draw.” (§Specialized Spokes in V4)

“Anyone can build a Spoke. If it adds value, it can tap into the Liquidity Hub as a credit line.” (§Benefits of the V4 Architecture)

How to read the original

Section titled “How to read the original”

Background needed: Aave V3’s Isolation Mode and E-Mode from /lending/collateral-ltv-health/, since V4’s Spokes are largely a re-packaging of those same ideas into a new architectural layer. Nothing needs to be skipped — the post is short and written for a general audience. The hardest thing to hold in your head, not because any one paragraph is hard but because it’s easy to conflate with Morpho/Euler’s model: V4 is not an isolated-market design. Every Spoke still draws from one shared Hub; the isolation is only in each Spoke’s rules, not in the underlying liquidity, which is the opposite trade-off from Morpho Blue or Euler v2.

What changed since

Section titled “What changed since”
  • Aave V4 was still explicitly “in development” as of this June 2025 post’s own disclaimer; whether the final shipped design matches this architecture description, and when it goes live, is not resolved by this page’s sources.
  • Euler’s relaunch as Euler v2 (2024) predates this Aave V4 post and represents the competing, fully-isolated-market answer to the same fragmentation problem Aave V4 describes — the two designs shipped in the opposite order from what a reader might assume (Euler’s modular rebuild came first, as a reaction to being hacked; Aave’s is a proactive redesign of an already-dominant, never-hacked-at-this-scale protocol).

Secondary references

Section titled “Secondary references”
  1. Morpho Midnight Whitepaper, §1 (2026-05) — read for the clearest short primary description of Morpho Blue’s isolated-market-plus-vault-curation split, even though the whitepaper’s main subject is the fixed-rate Midnight protocol built on top of it.
  2. Euler Finance blog, “Introducing Euler V2: The Credit Layer for Programmable Finance” (fetched 2026-08-29) — read for the EVK/EVC architecture in the vault-builder’s own words, including the “single shared pool cannot fit every asset… fully isolated markets can limit composability” framing this page quotes.
  3. Kamino Lending docs (curators and markets pages, 2026-08 fetch) — read if you want to see the same market-creation-plus-curation split expressed as a concrete Solana product flow rather than an abstract architecture.

10. Sources

Section titled “10. Sources”