● DeFi & On-chain

DeFi Lending in 2026: Modularization and Curator Risk

On-chain lending is the biggest business in DeFi, holding tens of billions in deposits. Here is how it works, and why the Stream and Resolv blowups exposed its fragility.

By Yuki Tanaka · Aug 7, 2026 · 14h ago ~22 min read

On-chain lending is the biggest business in decentralized finance, and in 2026 it is also the most contested. Around $54 billion in deposits sit across more than 380 protocols tracked by DefiLlama, which makes credit the largest single category in DeFi by a wide margin. Yet the same year that on-chain lending pulled in Wall Street capital and institutional collateral also produced two of the ugliest blowups the sector has seen: the Stream Finance implosion in November 2025 and the Resolv exploit in March 2026, each of which turned one hidden point of failure into tens of millions of dollars of bad debt scattered across markets that were supposed to be isolated from one another. This guide explains how on-chain credit works, how it is being rebuilt around an idea called modularization, and why the design choices that make it powerful are the same ones that make it fragile.

What On-Chain Lending Actually Is

At its core, a lending market is a set of smart contracts that lets one group of users deposit assets to earn interest and another group borrow those assets against collateral they lock up. There is no loan officer, no credit score, and no paperwork. Everything runs on public code: you connect a wallet, supply an asset such as USDC or ETH, and immediately start earning the interest that borrowers pay. If you want to borrow, you post collateral worth more than the loan and draw funds against it. Because there is no way to chase a defaulter through the courts, almost all of this lending is overcollateralized, meaning the borrower always locks up more value than they take out.This model was pioneered by Compound and Aave between 2019 and 2020, when they turned the idea of a “money market” into a live protocol. The appeal is straightforward. A supplier earns a yield set by real borrower demand rather than a bank’s discretion, and can withdraw at any time as long as there is free liquidity. A borrower gets instant, permissionless leverage or liquidity without selling a position, which is useful for traders, for people who want cash without triggering a taxable sale, and increasingly for institutions that want to keep their Bitcoin and borrow dollars against it. The catch, which the rest of this article keeps returning to, is that the whole structure rests on two assumptions: that the collateral is worth what the system thinks it is worth, and that a struggling loan can be closed before it goes underwater.

The Engine Room: Utilization, the Kink, and Interest Rates

Interest rates in a lending pool are not set by a committee; they are set by a formula that responds to how much of the pool is being borrowed. That measure is called utilization: if suppliers have deposited $100 million and borrowers have drawn $80 million, utilization is 80 percent. As utilization climbs, both the borrow rate and the supply rate climb with it, because scarcer liquidity should cost more.Most protocols use a piecewise interest-rate curve with a deliberate bend in it, known as the kink or the optimal utilization point. Below the kink, rates rise gently as borrowing increases. Above it, rates rise steeply, sometimes almost vertically, to punish very high utilization and pull the pool back toward a safe level. The logic is defensive: suppliers need to be able to withdraw, so the system makes the last slice of available liquidity expensive to borrow, which nudges borrowers to repay and tempts new suppliers to deposit. Suppliers receive the borrow interest minus a reserve factor that the protocol keeps as revenue or as a safety buffer. All of this is variable and recalculated block by block, so a rate that looks attractive at deposit time can move sharply if a large borrower enters or exits.When you supply, most protocols hand you an interest-bearing receipt token that tracks your share of the pool and quietly grows as interest accrues, so you rarely see a manual interest payment; your balance simply increases. Redeeming that receipt returns your principal plus everything it earned. This is also why withdrawals can occasionally stall: if utilization is very high because borrowers have taken out most of the pool, there may not be enough idle liquidity to redeem every supplier at once until some loans are repaid or the elevated rates lure fresh deposits. It is a design that works smoothly in calm markets and shows its edges in panicked ones.

Health Factors, Liquidations, and the Keepers Who Enforce Them

Every borrower position carries a health factor, a single number that tells you how close the loan is to being forcibly closed. It is derived from the value of the collateral, the size of the debt, and a risk parameter called the liquidation loan-to-value, or LLTV, sometimes called the liquidation threshold. If you post $10,000 of ETH at an 80 percent LLTV, you can borrow up to $8,000; the moment your debt plus accrued interest crosses that line, whether because you borrowed more or because ETH fell, your health factor drops below one and the position becomes eligible for liquidation.Liquidation is handled by independent bots, often called keepers or liquidators, that watch every position and pounce the instant one goes underwater. A liquidator repays part of the bad loan and seizes the borrower’s collateral at a discount, the liquidation bonus, which is the profit that makes the whole enforcement mechanism run without any central operator. When markets fall fast, liquidations can cascade: forced selling of seized collateral pushes prices lower, which pushes more positions below their threshold, which triggers still more liquidations. A healthy lending market is one where this enforcement happens cleanly and on time. As the Stream and Resolv episodes later in this article show, the danger is usually not that liquidations happen; it is that they fail to happen when they should.

Oracles: The Price Feed That Holds It All Together

A lending contract has no innate idea what ETH or any other collateral is worth. It learns the price from an oracle, an external service that pushes market prices on-chain. Chainlink is the dominant provider, but many markets use bespoke feeds, and this single dependency is arguably the most sensitive part of the entire system. If the oracle reports a price that is too high, the protocol will happily let someone borrow against collateral worth far less than the loan, and the lenders eat the difference.Historically, the classic attack was to manipulate the price an oracle reads, often with a flash loan that distorts a thin market for a single block. The industry hardened against that over years of painful lessons, a saga covered in detail in our report on oracle manipulation in 2026. But a subtler failure has become the defining risk of the modular era: the hardcoded oracle. To save gas or to simplify a market for a yield-bearing token that should trade near a fixed value, some vaults simply assume a price of, say, exactly one dollar, or a fixed exchange rate, rather than reading a live feed. That shortcut works right up until the token depegs, at which point the market keeps pricing discounted or worthless collateral at full value, liquidations never fire, and borrowers walk away with the lenders’ money. Two of 2026’s biggest disasters were exactly this.

The 2026 Market Map

Lending is concentrated: the top handful of protocols hold the large majority of all deposits, and the composition has shifted over the past year as Aave shipped a new version and Morpho scaled aggressively. The table below sketches the field in mid-2026. Treat every figure as a snapshot; total value locked moves constantly and different trackers define deposits and TVL differently, so cross-check DefiLlama before acting on any number.

Protocol	Architecture	Main chains	Approx. size	Notes
Aave	Pooled plus V4 hub-and-spoke	Ethereum, L2s, Avalanche	~$19B+	Category leader; native GHO stablecoin
Morpho	Immutable core plus curator vaults	Ethereum, Base	~$11B deposits	Powers Coinbase and Binance loans
Spark	Aave v3 fork, stablecoin-focused	Ethereum	~$6.8B	Sky (ex-MakerDAO) ecosystem
Compound	Pooled (V3 Comet)	Ethereum, L2s	~$2.7B	The original, now mature
JustLend	Pooled	Tron	~$2.4B	Large but chain-specific
Fluid	Smart collateral / smart debt	Ethereum	~$1.6B	Instadapp team
Kamino	Pooled	Solana	~$1.1B+	Solana lending leader
Euler	Modular vaults (EVK/EVC)	Ethereum, multi-chain	~$1B+	Rebuilt after 2023 hack

One caveat on that table: Morpho reports more than $11 billion in deposits, a figure that counts gross supplied capital, while its net value locked on trackers reads lower because the two metrics measure different things. It is a useful illustration of why headline numbers in DeFi should always be checked against a definition.

The Great Modularization

The most important structural story in on-chain credit is a shift the research firm Tiger Research calls modularization. For years, a lending protocol was a monolith: one team wrote the code, chose which assets to list, set every risk parameter, and pooled all deposits together. That is simple and it concentrates liquidity, but it also means one bad listing can threaten the whole pool, and it forces every user to trust a single risk committee.Modularization splits that monolith into two layers: a neutral piece of infrastructure that just moves money safely, and a separate layer of risk managers who decide what to lend against and on what terms. Tiger Research draws an analogy to how traditional finance restructured after 2008, separating execution from risk oversight and spreading exposure across many specialized firms. On-chain, roughly $7.4 billion sits in curation vaults that embody this design. The report sorts the three leading approaches into three recognizable Wall Street shapes, summarized below.

Protocol	TradFi analogy	Who manages risk	How it works
Morpho	Prime brokerage	External, independent curators	Immutable core executes; curators pick markets and allocate capital
Aave V4	Universal bank	The protocol, internally	Central hubs pool liquidity; spokes get capped credit lines
Euler V2	Multi-strategy hedge fund	Per-vault operators	EVK vaults and the EVC connector allow flexible, cross-collateral strategies

Each design answers the same question, how do you isolate a bad bet so it does not sink the ship, in a different way. And each, as the events of the past year proved, has a distinct failure mode.

Aave V4 and the Universal Bank

Aave remains the incumbent, with well over $19 billion in value locked, and in March 2026 it shipped Aave V4 on Ethereum mainnet after more than two years of development. V4 rebuilds the protocol around a hub-and-spoke architecture. Liquidity is concentrated in a small number of hubs, currently a low-risk Prime hub, a risk-adjusted Core hub, and a higher-risk, higher-return Plus hub, while individual spokes plug into those hubs to serve specific use cases with their own parameters. Crucially, each spoke receives a capped credit line from its hub, so a blowup in one spoke cannot drain the shared liquidity behind it. Early spoke operators include large staking and stablecoin players such as Lido, EtherFi, Kelp, Ethena, and Lombard.V4 also deepens the role of Aave’s native stablecoin GHO, which functions as a settlement asset, and its yield-bearing cousin sGHO, pitched as an on-chain savings product. Aave founder Stani Kulechov has been explicit that the next phase is about borrowing, not just supplying. “Now what we want to focus is on the borrow side, creating significant borrow demand,” he told The Block, describing a plan to channel on-chain liquidity toward real-world economic activity. That ambition is visible in Aave Horizon, the protocol’s institutional and real-world-asset initiative, which has courted tokenization players and asset managers. The through-line is that Aave is trying to become a full-service on-chain bank rather than a single shared pool.

Morpho and the Curator Economy

If Aave is the universal bank, Morpho is the clearest expression of the prime-brokerage model, and it has become the venue of choice for institutions that want DeFi rails without building their own. Its base layer, Morpho Blue, is a deliberately minimal and immutable core: anyone can spin up an isolated market defined by just five parameters, the collateral asset, the loan asset, the LLTV, the oracle, and the interest-rate model. The core never changes and never lists assets itself. All the judgment lives one layer up, with curators.A curator, a term Morpho popularized, runs a vault that accepts depositors’ funds and spreads them across Blue markets in search of yield, taking a fee and, in principle, owning the risk decisions. Firms such as Steakhouse Financial and Gauntlet manage billions this way, and the overwhelming majority of Morpho volume is stablecoin lending; after the shocks of late 2025, institutional capital has concentrated in a handful of the most established curators. The model has attracted serious volume: Morpho reports more than $11 billion in deposits, and it quietly powers consumer products at the largest exchanges. Coinbase’s Bitcoin-backed loans run on Morpho and have grown to more than $1 billion borrowed against Bitcoin, part of over $2 billion in loans originated through the integration; Binance has built on it too. In June 2026, Morpho raised $175 million at a $2 billion valuation in a round co-led by Paradigm, a16z crypto, and Ribbit Capital, with Apollo, VanEck, and Circle among the backers, to build what it calls an “open credit network.” The pitch is that Morpho is not a lender at all; it is the neutral pipe that everyone else lends through.

The Challengers: Euler, Spark, Fluid, and Compound

Below the two giants sits a diverse field. Euler returned from near-death: after a nine-figure exploit in 2023, from which the attacker eventually returned the funds, it rebuilt around Euler V2, a modular stack of vaults (the Euler Vault Kit) wired together by the Ethereum Vault Connector, giving it the flexible, cross-collateral character that earns it the hedge-fund label. Spark, part of the Sky ecosystem that grew out of MakerDAO, runs a stablecoin-optimized fork of Aave v3 and channels the Sky Savings Rate to holders of its dollar, making it one of the larger venues by deposits.Fluid, from the Instadapp team, has pushed a “smart collateral, smart debt” design that lets the same assets do double duty as collateral and liquidity, chasing capital efficiency. Compound, the protocol that arguably started the category, has settled into a mature, slower-growing role with its V3 Comet markets. And on Solana, Kamino has become the dominant lending venue, a reminder that on-chain credit is not an Ethereum-only phenomenon; Venus on BNB Chain and JustLend on Tron similarly anchor their own ecosystems. What unites the leaders is the move toward isolation and modularity; what separates them is whom they trust to set the risk.

When Curators Fail: The Stream Finance Contagion

On 4 November 2025, a protocol called Stream Finance disclosed that an external fund manager had lost roughly $93 million of the assets backing its yield-bearing stablecoin, xUSD, and froze deposits and withdrawals. xUSD fell from $1 to about $0.26, a 77 percent collapse in 24 hours. What turned a single fund’s blowup into what many called the darkest week in DeFi history was the plumbing around it.xUSD had been used as collateral in isolated lending vaults across Morpho, Euler, and Silo, where third-party curators let users borrow against it. Several of those vaults priced xUSD at a hardcoded $1 even as its market value cratered, so the liquidations that should have unwound the leveraged positions never triggered; borrowers simply drained the real stablecoins that lenders had supplied. Because xUSD had also been rehypothecated through recursive borrowing loops, its footprint was far larger than its backing, and research firm Yields And More estimated total exposure across the ecosystem at around $285 million. The damage jumped a further firewall when Elixir’s deUSD, which had lent a large share of its reserves against Stream, was itself wiped out, falling from $1 toward a cent. Roughly $1 billion flowed out of DeFi lending in the following days. The uncomfortable detail, widely reported afterward, was that curators chasing fees had funneled depositor money toward a strategy run by a single trader with minimal oversight. Curators including MEV Capital and Re7 Labs were among those with exposure, and on Lista DAO a governance vote moved to force-liquidate the affected vaults, with Re7 Labs later acknowledging a multimillion-dollar hit. Isolation had been designed into the contracts; it had not been designed into the incentives.

The Hardcoded-Oracle Curse: Resolv and the Repeat Offenders

If Stream was about human oversight, the Resolv exploit four months later was about the same technical flaw striking yet again. On 22 March 2026, an attacker who had compromised a signing key stored in Resolv’s Amazon Web Services key-management system minted 80 million units of the protocol’s USR stablecoin out of thin air, turning a few hundred thousand dollars into roughly $25 million in minutes. USR crashed toward $0.025 before partially recovering. The larger wound, again, was in the lending markets: because a related token, wstUSR, was priced by hardcoded oracles at a fixed value while it actually traded far lower, attackers deposited cheap collateral and borrowed at full face value. Fluid absorbed more than $10 million in bad debt and saw hundreds of millions of dollars of outflows in a single day, while Morpho markets faced a wave of liquidations.Security researchers were blunt that this was avoidable. Omer Goldberg, founder of the risk firm Chaos Labs, and others noted that hardcoded pricing for assets that can depeg is a known, repeatable trap rather than a novel exploit. By several counts it was the fourth hardcoded-oracle failure in about 14 months. The pattern is worth tabulating, because the through-line matters more than any single incident.

Incident	Date	The hardcoded assumption	Rough impact
Usual USD0++	Jan 2025	Floor price held while the token slid	Cascading liquidations
Moonwell	Autumn 2025	Stale or fixed feed on a paired asset	Bad debt on affected markets
Stream xUSD	Nov 2025	xUSD priced at $1 while depegging	~$285M exposure, ~$1B outflows
Resolv wstUSR	Mar 2026	wstUSR fixed above its market price	~$25M minted, $10M+ Fluid bad debt

Two lessons echo through that list. The first is a theme HOGE Wire has covered before: an audit is not a guarantee, as our piece on firms that were audited and hacked anyway laid out. The second is that the Resolv breach began not in a smart contract but in a leaked cloud key, a reminder that operational security and key management, the subject of our multisig best-practices playbook, remain the soft underbelly of on-chain finance.

Fixed Rates, Real-World Assets, and the Institutional Frontier

For all its drama, on-chain lending in 2026 is also maturing fast, and two frontiers stand out. The first is fixed-rate, fixed-term credit. Almost all DeFi lending has historically been variable-rate, which institutions dislike because it makes cash flows unpredictable. In July 2026, Morpho launched Midnight, a protocol that builds fixed-rate, fixed-term lending at the base layer rather than bolting it on top of variable markets, starting with a cbBTC/USDC market on Base across several maturities. Founder Paul Frambot frames it as the third wave of on-chain lending: first-generation protocols set the rate, risk, and term for users; Morpho Blue handed risk to curators; and Midnight now hands rate, risk, and term to an open market. He argues that without native fixed-rate credit, on-chain markets stay incomplete, unable to serve the borrowers who plan around predictable payments.The second frontier is real-world assets and the institutions that hold them. Tokenized US Treasuries and private-credit funds increasingly show up as collateral, and the plumbing to plug regulated players in is arriving: custody providers such as Anchorage Digital, Ledger Enterprise, and Taurus are wiring embedded access to on-chain yield for their clients, while asset managers including Apollo and Bitwise appear alongside Aave’s Horizon and in Morpho’s cap table. The same wave that produced a rush of crypto ETF approvals in 2026 is now pushing institutions one layer deeper, from simply holding tokens to borrowing and lending against them. The prize, if the risk controls hold, is a bridge between trillions of dollars of traditional credit and a settlement layer that never closes.That bridge comes with a philosophical tension. To satisfy compliance teams, several institutional markets are permissioned, gating who can supply or borrow through whitelists and know-your-customer checks, which sits awkwardly beside DeFi’s permissionless origins. The likely outcome is a split system: fully open, pseudonymous markets for crypto-native users, and walled, curated, compliant markets for regulated capital, both running on the same neutral base contracts. Morpho and Aave are each building for both worlds at once, betting that the shared infrastructure underneath is what ultimately wins.

The US Regulatory Picture: SEC, CLARITY, and the Broker-Rule Repeal

A blunt fact frames everything: a DeFi lending protocol is not a bank. There is no deposit insurance from the FDIC, no lender of last resort, and when a vault takes on bad debt, as Fluid and others did, the loss falls on suppliers rather than on any government backstop. That is the core risk regulators in the United States are still deciding how to treat.The tone from Washington has shifted markedly. Under Chair Paul Atkins, the Securities and Exchange Commission has pursued a friendlier agenda it calls Project Crypto, including a proposed safe harbor for decentralizing projects and clearer token taxonomy, and Atkins has signaled the SEC will write crypto market rules itself if Congress stalls. The agency is “ready, willing, and able to come out with rules,” he said, framing SEC action as a bridge to legislation. That legislation, the CLARITY Act, which would split oversight of digital assets between the SEC and the Commodity Futures Trading Commission and carve out treatment for DeFi and software developers, passed the House in July 2025 but, as our regulatory countdown has tracked, remained stuck in the Senate through the August 2026 recess. Separately, the most concrete DeFi-specific action so far cut the other way: in April 2025, Congress and the President used the Congressional Review Act to repeal the IRS DeFi broker rule, sparing front-end DeFi interfaces from the tax-reporting and know-your-customer obligations that centralized exchanges still carry. For now, most on-chain lending occupies a gray zone: increasingly welcomed, not yet fully governed.

How to Read a Lending Market Before You Deposit

The lesson of the past year is that yield is not risk-adjusted for you. In a modular world, two vaults offering the same asset can carry wildly different danger depending on who set them up and how. Before supplying to any market, it is worth checking a short list of things that separated the victims of 2025 and 2026 from the survivors.

The oracle. Does the market read a live, reputable price feed, or does it hardcode a value for a supposedly stable asset? Hardcoded pricing is the single most repeated cause of catastrophic loss.
The curator. In a curator model you are trusting a firm, not just code. Look at who runs the vault, their track record, and whether they were exposed to Stream or Resolv.
Isolation. Is your collateral in an isolated market that cannot infect others, and can others not infect it? Isolation limits contagion but does not eliminate correlated failure.
The LLTV and liquidity. Aggressive loan-to-value ratios boost capital efficiency and raise liquidation risk. Thin liquidity can trap you when everyone withdraws at once.
Backing and audits. For yield-bearing collateral, ask what actually backs it and whether that backing is verifiable, and remember that an audit reduces risk without removing it.

None of this makes on-chain lending safe in the way an insured savings account is safe. It makes the risks legible, which is the most an honest guide can offer in a market that pays you precisely because it can lose your money.

Frequently Asked Questions

How does DeFi lending actually work?

You deposit a crypto asset into a smart-contract pool and earn interest paid by borrowers, or you lock collateral worth more than your loan and borrow against it. Interest rates adjust automatically with how much of the pool is borrowed, and automated liquidators close your position if your collateral falls too far relative to your debt. No bank, credit check, or intermediary is involved.

Is DeFi lending safe?

It carries risks a bank deposit does not. There is no FDIC insurance, and losses from bad debt fall on suppliers. The largest 2026 failures came from faulty oracles that mispriced collateral and from curators who allocated deposits to risky strategies, which drove the Stream Finance and Resolv losses. Overcollateralization and market isolation reduce risk but do not remove it.

What is a curator in DeFi lending?

A curator is a firm or team that runs a lending vault on top of a neutral protocol such as Morpho, choosing which markets to lend into and how much risk to take in exchange for a fee. Depositors delegate risk management to the curator, so the curator’s competence and track record matter as much as the underlying code.

What is the difference between Aave and Morpho?

Aave is an integrated protocol that pools liquidity and sets risk parameters itself, and its V4 uses a hub-and-spoke design to cap contagion. Morpho is a minimal, immutable base layer that lists nothing on its own, while independent curators build the actual lending markets on top. Aave is closer to a bank, Morpho closer to neutral infrastructure that others lend through.

How much can you earn lending crypto?

Yields are variable and come from the interest borrowers pay, so they rise and fall with demand. Stablecoin supply rates often sit in the low-to-mid single digits but can spike much higher when borrowing surges. Higher advertised yields usually signal higher risk, whether from aggressive leverage, thin collateral, or a curator taking bets that can end in loss.Written by the HOGE Wire markets desk, covering DeFi credit, on-chain risk, and the infrastructure that moves money without a bank.