Borrow, Loop, Repeat: On-Chain Lending Markets in 2026
On-chain lending now holds more than $50 billion in deposits. Here is how borrowing, CDP stablecoins, looping, and Aave V4's rebuild actually work in 2026.
On-chain lending stopped being a novelty a long time ago. By late September 2026, the money markets built on public blockchains hold north of $50 billion in deposits, most of it concentrated in a handful of protocols that now behave like the wholesale funding desks of crypto. They set the base rate for borrowing dollars against Bitcoin and Ethereum, they mint several of the largest stablecoins in circulation, and they route billions in leverage that never touches a centralized exchange.
2026 was the year the plumbing underneath all of it got rebuilt. Aave, the largest protocol, shipped a ground-up rewrite. Morpho turned lending into a marketplace of independently curated vaults, then ran headlong into the limits of that idea. Sky, the protocol formerly known as MakerDAO, pushed its USDS stablecoin past $11 billion. A fast-growing challenger called Fluid fused lending and trading into a single liquidity layer. For a borrower, the four numbers that matter (collateral, loan-to-value, the health factor, and the liquidation point) are unchanged. The machinery around them looks nothing like it did a year ago.
This is a field guide to that machinery: how a supply-and-borrow market is assembled, how interest rates are set and why they lurch, why so many users now borrow a stablecoin they mint themselves, how the looping trade manufactures leverage without a perpetual future, and where the whole structure cracked open in November 2025.
What an On-Chain Money Market Actually Is
A DeFi lending protocol is a set of smart contracts that pools capital from lenders and hands it to borrowers with no loan officer in between. Suppliers deposit an asset, say USDC, and receive a receipt token that accrues interest automatically; on Aave that receipt is an aToken, on Compound it is a rebasing balance. Borrowers post collateral worth more than they take out, then draw a different asset against it. Interest paid by borrowers flows to suppliers, minus a reserve cut the protocol keeps for itself.
Two features separate this from a bank. First, almost everything is overcollateralized: to borrow $100 you must lock more than $100 of another asset, because the protocol cannot chase a defaulter through a court. Second, positions are transparent and liquidatable in real time. Anyone can watch a wallet drift toward the danger zone and, the moment it crosses, repay part of its debt in exchange for its collateral at a discount. That combination, excess collateral plus instant liquidation, is what lets a pool of anonymous strangers lend to one another safely.
The state of a pool is summarized by utilization: the share of supplied capital currently borrowed. Low utilization means cheap borrowing and thin lender yield; high utilization means the opposite; and if utilization hits 100 percent, suppliers cannot withdraw until borrowers repay or new deposits arrive. Every interest-rate model in DeFi is, at its core, a rule for pricing that one number.
Because the receipt tokens are themselves transferable, a supplier’s position is composable: an aToken can be posted as collateral elsewhere, wrapped into a vault, or dropped into a structured product. That composability is what makes DeFi lending a base layer rather than a destination, and it is also what lets risk travel between protocols faster than anyone can track, a theme that runs through everything below.
The Mechanics of a Borrow: Collateral, LTV, and the Health Factor
When you open a borrow, three numbers govern your fate. The loan-to-value ratio (LTV) sets how much you can draw against a given collateral: an 80 percent LTV on wstETH means $10,000 of wstETH lets you borrow up to $8,000. The liquidation threshold sits a notch higher, say 83 percent, and marks the point at which your position can be seized. The health factor is the ratio traders actually watch, calculated as collateral multiplied by the liquidation threshold, divided by debt. Above 1.0 you are safe; touch 1.0 and you are liquidated.
Work a real example. You deposit $10,000 of wstETH with an 83 percent liquidation threshold and borrow $7,000 of USDC. Your health factor is (10,000 x 0.83) divided by 7,000, or about 1.19. If Ethereum falls far enough that your collateral is worth $8,500, the health factor slides to roughly 1.008, one bad candle from liquidation. Nothing in that calculation cares who you are; it is pure arithmetic, re-run every block.
| Metric | Value | What it means |
|---|---|---|
| Collateral deposited | $10,000 wstETH | The asset you lock |
| Maximum LTV | 80% | Ceiling on borrowing ($8,000) |
| Amount borrowed | $7,000 USDC | Your actual debt |
| Liquidation threshold | 83% | Where collateral can be seized |
| Health factor | 1.19 | (10,000 x 0.83) / 7,000; 1.0 triggers liquidation |
Different assets carry different parameters, and that is where risk management lives. Blue-chip collateral (ETH, wrapped Bitcoin, major stablecoins) earns high LTVs; long-tail tokens get low ones, or are ring-fenced in isolated markets where a bad price cannot infect the rest of the pool. Setting those numbers well is the whole job, and in 2026 it became a job with a title: the risk curator.
How Interest Rates Are Set (And Why They Move)
Most DeFi borrow rates are algorithmic, not quoted by a desk. A protocol defines an interest-rate curve as a function of utilization: the rate climbs gently up to an optimal utilization point (the kink, usually somewhere around 80 to 90 percent), then rockets upward beyond it to force repayment and protect withdrawals. Borrowers pay that variable rate; suppliers earn it, scaled by how much of the pool is actually lent out.
The design makes on-chain rates jump around far more than a bank’s. A single large borrow can push utilization past the kink and double the rate in one block; a wave of repayments can halve it just as fast. It also tethers DeFi loosely to the wider cost of money. When the Federal Reserve lifted its target range to 3.75 to 4 percent in September 2026, tokenized Treasury yields and stablecoin lending rates moved with it, because capital that can earn roughly 4 percent risk-free in a money-market fund will not lend USDC on-chain for less unless something extra is on the table. HOGE Wire covered how crypto rallied even as the Fed hiked, and the same rate gravity now shapes on-chain credit.
It helps to separate the two rates a protocol quotes. The borrow APY is what borrowers pay; the supply APY is what lenders earn, and it is always lower, because it equals the borrow rate multiplied by utilization and then reduced by the reserve factor the protocol skims. A pool advertising a fat supply yield is really telling you that utilization is high, which is the same as saying withdrawals are tight. Yield and liquidity are two readings of one dial, and confusing them is how depositors get stuck.
The genuine novelty of the last two years is that the variable rate is no longer the only choice. Fixed-rate, fixed-term lending, long the missing piece of DeFi, arrived through intent-based markets such as Morpho V2 and purpose-built venues that let a borrower lock a coupon for a defined period. That matters most to institutions, which cannot underwrite a loan whose rate might triple overnight, and it is the precondition for an on-chain yield curve that resembles the one traditional fixed income runs on.
Liquidations: The Backstop That Keeps Lenders Whole
When a health factor drops to 1.0, the position is open season. Liquidators, almost always bots, repay a slice of the debt and receive the borrower’s collateral plus a bounty, typically 5 to 15 percent, as their incentive. In normal conditions the mechanism is brutally effective, and it is the reason suppliers can sleep: bad positions are usually cleared within seconds of crossing the line.
It is also a major source of MEV. Liquidation bounties are among the most contested opportunities in the mempool, and the priority-gas auctions that decide who wins them are a textbook case of maximal extractable value. Modern wallets increasingly automate the borrower’s side too; account-abstraction upgrades such as EIP-7702 let a smart account bundle a top-up-collateral action that fires the instant a health factor slips, sparing the owner a 3 a.m. liquidation.
Liquidations only fail when prices gap faster than bots can react, or when the oracle feeding the price is wrong. Either way, a position can slide below 100 percent collateralization before it is closed, leaving the protocol with bad debt: a loan that can never be fully repaid. Bad debt is the cardinal sin of a lending market, and avoiding it is why parameters are kept conservative, why oracle choice is obsessive, and why the events of November 2025 landed as hard as they did.
CDP Stablecoins: Borrowing a Dollar You Mint Yourself
The most consequential kind of on-chain loan does not draw from anyone’s deposit at all. In a collateralized debt position (CDP), you lock collateral and the protocol mints a brand-new stablecoin against it, then burns that stablecoin when you repay. You are not borrowing a lender’s USDC; you are issuing a dollar into existence, backed by your ETH, and retiring it when the loan closes.
This is how the oldest DeFi stablecoin, DAI, has always worked, and in 2026 it is how three of the sector’s growth engines run. Sky, the rebranded MakerDAO, mints DAI and its successor USDS; USDS has grown into one of the largest stablecoins anywhere, past $11 billion in supply, on the back of a 2024 rebrand that split the token in two. Aave mints GHO, which pushed past $500 million in market value in early 2026 after a 245 percent climb and now sits at the center of the new V4 design. Curve mints crvUSD, a few hundred million in supply, backed by a soft-liquidation engine that sells collateral gradually as a position weakens rather than dumping it all at a single trigger price.
The appeal is structural. A CDP borrower controls their own rate (set by governance rather than by a twitchy utilization curve), faces no risk that lenders will withdraw the liquidity underneath them, and mints a dollar on which the protocol earns a durable spread. Applied across billions in supply, that spread has become the profit center of the biggest lending DAOs. The yield-bearing savings versions, sGHO and sUSDS, pass part of it back to holders and compete directly with the Sky Savings Rate and with money-market funds.
| Stablecoin | Issuer | Mechanism | Approx. supply | Savings product |
|---|---|---|---|---|
| DAI / USDS | Sky (ex-MakerDAO) | Overcollateralized CDP plus real-world assets | USDS past $11B | sUSDS / Sky Savings Rate |
| GHO | Aave | CDP, mintable from V4 spokes | Past $500M | sGHO |
| crvUSD | Curve | CDP with LLAMMA soft-liquidation | A few hundred $M | scrvUSD |
The Looping Trade: Leverage Without a Perpetual
If borrowing against collateral is the base case, looping is the power-user version, and it has quietly become one of the largest sources of borrow demand on-chain. The recipe is simple: deposit a yield-bearing asset such as wstETH, borrow a stablecoin against it, buy more of the yield-bearing asset, redeposit, and repeat. Each turn of the loop multiplies exposure. Four loops on wstETH can lift a 3 percent staking yield into low-double-digit territory, as long as the borrow rate stays below the staking rate and Ethereum does not fall out of bed.
Protocols now package the loop so users never touch the individual steps. Fluid, the Instadapp-built venue that merges lending and a DEX into one liquidity layer, ships pre-built vaults such as a wstETH/USDC position that loops to roughly 4x automatically; Fluid quadrupled its deposits over the past year on exactly this demand. Aave, Morpho, and Contango offer one-click leverage too, and smart-account standards collapse the whole multi-step bundle into a single signature.
Looping is also where leverage risk concentrates. A looped position is long its collateral and short the borrowed dollar with amplified sensitivity to both; a sharp drop in the collateral, or a spike in the borrow rate above the underlying yield, can unwind the entire stack through a chain of liquidations. The same mechanic that manufactures yield in calm markets manufactures forced selling in violent ones, which is precisely how a single bad stablecoin metastasized into a system-wide event in late 2025.
Aave V4 and the Hub-and-Spoke Redesign
On March 30, 2026, Aave shipped V4 to Ethereum mainnet after more than two years of development, the largest architectural change in the protocol’s history, as The Block reported. V4 retires the single shared pool of V3 in favor of a hub-and-spoke model: a central Liquidity Hub holds the capital and the accounting, while individual spokes plug in as separate borrowing environments, each with its own collateral set and risk parameters.
The goal is to end the old trade-off between deep liquidity and safe isolation. “The hub gives a credit line to each spoke,” Aave founder Stani Kulechov told The Block. “Every single use case is capped by the exposure of the credit line, so that’s like a risk mitigation feature.” V4 launches with three risk tiers Kulechov labels Prime (low risk), Core (risk-adjusted), and Plus (risk return), and with major staking and restaking names (Lido, EtherFi, Kelp, Ethena, and Lombard) lined up to run spokes at launch. The protocol documentation and the open-source codebase are both public.
GHO sits at the heart of the redesign. It becomes natively mintable from any spoke rather than from a single market, and the yield-bearing sGHO functions as an on-chain savings account wired straight into the hub. The ambition is openly institutional. Chainlink co-founder Sergey Nazarov, whose oracles feed the protocol’s prices, called the launch “a major milestone toward connecting onchain finance directly to broader global capital markets.” Whether the credit-line caps hold up in a real drawdown is the question the next bout of market stress will answer.
Morpho and the Curator Model
Where Aave rebuilt one integrated system, Morpho went the opposite direction and unbundled lending entirely. Morpho Blue is a minimal, immutable base layer: anyone can spin up an isolated market defining a single collateral, a single loan asset, an LTV, and an oracle, and that market cannot touch any other. On top of that base sit curated vaults, where professional curators decide which markets a depositor’s funds flow into and tune the risk-and-yield mix on the lender’s behalf.
The design pushed Morpho into second place among lending protocols, past $10 billion in deposits by DefiLlama’s count, and turned risk curation into a business. Firms like Gauntlet, the quantitative shop that has set parameters for Aave, Compound, and Maker since 2020, and Steakhouse Financial run competing vaults and market their track records the way fund managers market returns. The pitch is seductive: keep the transparency and self-custody of DeFi, but hand the hard part, deciding which markets are actually safe, to a specialist.
It is a fundamentally different bet from Aave’s. Aave’s DAO curates one large book conservatively; Morpho lets a thousand books bloom and lets depositors choose their curator. Both approaches have real merits, and both were about to be tested by the same question: what happens when a curator gets it wrong? In November 2025, the market found out.
The Curator Reckoning: What Stream Finance Broke
On November 4, 2025, a yield protocol called Stream Finance imploded and took a chunk of the curator model’s credibility with it, in what Protos called a risk-curator reckoning. Stream issued a stablecoin, xUSD, whose eye-catching 18 percent yield came from a looped strategy: curators on Morpho and Euler had funneled real stablecoins into isolated markets against xUSD collateral, and the borrowed dollars were looped back into still more xUSD, manufacturing footprint without adding backing. When the strategy broke, xUSD fell 77 percent from around $1.00 to $0.26 within a day, froze roughly $160 million in deposits, and exposed some $285 million of interconnected debt across protocols.
The damage spread through the collateral graph. Elixir’s deUSD, roughly 65 percent backed by loans to Stream, collapsed close to 98 percent and was wound down; Stables Labs’ USDX slipped below $0.60, according to a Chorus One post-mortem. It was the 2025 version of a bank run: the same rehypothecated collateral pledged in several places at once, unwinding all at once.
Crucially, the base protocols mostly held. On Morpho, only one of roughly 320 curated vaults, run by MEV Capital, had direct xUSD exposure, and it absorbed around $700,000 in bad debt. Morpho co-founder Paul Frambot argued the model worked as designed, noting that the isolated-market structure left the other 319-plus vaults, each with its own risk profile, with zero exposure. Critics drew the opposite lesson: curators chasing the highest advertised yield had recreated the exact maturity-and-collateral mismatch DeFi was supposed to abolish, and depositors had no realistic way to audit what their chosen curator was doing. The episode also reopened the fight over emergency controls, the subject of HOGE Wire’s look at crypto’s off-switch reckoning: who gets to freeze a market, and how fast.
A Map of the Market: Who Holds the Deposits
For all the architectural diversity, deposits stay concentrated. Aave is still the anchor tenant, with the better part of $20 billion across chains; Morpho is the clear number two; and Sky’s Spark lending arm, Compound, and Fluid round out the top tier. The table below sketches the shape of the market in late September 2026, with deposit figures drawn from DefiLlama.
| Protocol | Model | Approx. TVL | Native stablecoin | 2026 headline |
|---|---|---|---|---|
| Aave | Pooled, moving to hub-and-spoke (V4) | ~$18-20B | GHO | V4 mainnet launch |
| Morpho | Isolated markets plus curated vaults | ~$10B | None | Curator model stress-tested |
| Spark (Sky) | Pooled, funded by USDS | Several $B | USDS / sUSDS | Grove institutional arm |
| Compound | Isolated base-asset markets (Comet) | A few $B | None | Institutional pivot |
| Fluid | Lending plus DEX liquidity layer | Above $1B | None | Roughly 4x deposit growth |
| Euler | Modular vaults (v2) | Rebuilt | None | Back after its 2023 exploit |
The concentration matters, because it means the health of on-chain credit rides on a few dozen sets of risk parameters and a small number of oracle feeds. Set conservatively, the system is arguably safer than the traditional shadow banking it imitates: every position is visible, and every liquidation is automatic. Set greedily, as November proved, the failure arrives faster than anything in traditional finance, with no overnight window and no central bank to call.
Real-World Assets and the Institutional Tier
The other structural shift of 2026 is the arrival of Wall Street collateral. Tokenized Treasury bills, private-credit funds, and money-market tokens are increasingly posted as collateral or parked in protocol reserves, and the biggest DAOs have spun up dedicated vehicles to hold them. Sky’s Grove allocates institutional treasury capital; Aave, Compound, and Morpho have each shipped permissioned, identity-gated markets aimed at funds that cannot legally lend into an anonymous pool. Compound’s DAO steered a sizable budget toward institutional clients in 2026, part of a broader split of the market into a permissionless retail tier and a compliant institutional one.
That bifurcation drags in the compliance questions DeFi spent years sidestepping. A permissioned lending market has to know its counterparties, which means running the same identity and anti-money-laundering machinery that centralized exchanges run; HOGE Wire’s account of OKX’s $504 million KYC lesson is a preview of what regulators expect when that machinery is treated as a formality. Tokenized real-world collateral also imports off-chain risk that no amount of on-chain overcollateralization can liquidate away: a bond issuer can default, a custodian can freeze the assets, a legal wrapper can fail in a court that has never heard of a health factor.
For the protocols, the appeal of real-world collateral is obvious: it is a large, relatively stable source of yield uncorrelated with crypto’s own booms and busts, and it lets a lending DAO earn a Treasury-like return on reserves rather than letting them sit idle. For regulators, it is where the tidy story of a self-contained, code-is-law system runs into the messy reality of issuers, custodians, and courts. The two tiers of the market are, in the end, two different answers to the same question: who do you have to trust?
The Risks Nobody Prices Correctly
A borrower fixated on the health factor is watching only one of several ways to lose money in a lending market. The catalogue, roughly ordered by how often it actually bites:
- Oracle risk: if the price feed is wrong or can be manipulated, liquidations misfire and collateral is mispriced. Stream was partly an oracle failure, since xUSD was marked above its real value for far too long.
- Curator and parameter risk: an LTV set too high, or a market pointed at bad collateral, can create bad debt without any code bug at all. This is now the dominant risk in the vault model.
- Smart-contract risk: the code can simply be exploited. Euler lost close to $200 million in 2023 before recovering the funds; the risk never reaches zero, which is why audits and formal verification are a permanent line item.
- Depeg and contagion risk: rehypothecated stablecoins and looped collateral bolt protocols together, so one failure propagates, exactly as November 2025 demonstrated.
- Governance and rate risk: a DAO vote can change your borrow rate or your collateral’s status overnight, and a variable rate can spike past the yield that made your loop profitable in the first place.
None of these is hypothetical, and the users who lost money in 2025 mostly lost it to the second and fourth items, not to a cinematic hack. The uncomfortable truth is that the products with the highest advertised yields (curated high-APY vaults, exotic collateral, deep loops) are precisely the ones whose failure modes are hardest for a depositor to see coming.
Rules, Taxes, and the SEC
The regulatory picture is still unsettled, and 2026 sharpened it rather than resolving it. In the United States, the Securities and Exchange Commission has spent years arguing that some lending-and-yield products are unregistered securities, a stance set during the prior cycle’s cases against centralized lenders, while the industry counters that a permissionless, overcollateralized, non-custodial protocol is closer to software than to a broker. The market-structure bills that might have drawn a clean line have repeatedly stalled in Congress, leaving the SEC’s enforcement posture as the effective rulebook for now.
For an ordinary user, the more immediate question is tax. Interest earned by supplying to a pool is generally treated as ordinary income, and depending on the jurisdiction, a borrow, a loop, or a liquidation can each be a taxable event. The reporting rules tightened again in 2026, which HOGE Wire tracked in its coverage of the digital-asset tax bill the House advanced. The permissioned institutional tier exists in large part because funds need exactly the compliance and reporting that the permissionless tier was designed to do without.
What to Watch in Late 2026
The near-term story is whether the rebuilt plumbing holds under load. Watch how quickly Aave V4’s spokes accumulate liquidity, and whether the hub’s credit-line caps behave as advertised the first time a major collateral gaps down. Watch Curve’s LlamaLend V2, which shipped in 2026 after an external audit, and the wider push into fixed-rate, fixed-term markets that would make on-chain credit legible to institutions accustomed to a yield curve. Watch, above all, whether the curator model tightens its standards (real disclosure, hard position limits, independent risk ratings) after November’s lesson, or whether the next high-yield vault simply runs the same play again.
The direction of travel is clear enough. On-chain lending is consolidating into a small number of large, increasingly institutional venues; it is splitting into a retail tier and a compliant tier; and it is pushing its own stablecoins as the settlement layer underneath everything else. The mechanics a borrower has to understand (collateral, loan-to-value, the health factor, and the liquidation point) are the same as they were in 2020. Everything wrapped around those mechanics is being rebuilt in real time, and the cost of not understanding the difference is measured in liquidations.
Frequently Asked Questions
What is a DeFi lending market, and how is it different from a bank?
A DeFi lending market is a set of smart contracts that pools lenders’ capital and lends it to borrowers automatically, with no bank or loan officer in the middle. Loans are overcollateralized, positions are liquidated instantly by bots if collateral falls too far, and every balance is transparent on-chain. There is no credit check and no recourse to a court, so excess collateral is the only protection lenders have.
How much can I borrow against my crypto in 2026?
It depends on the collateral’s loan-to-value ratio. Blue-chip assets like ETH, wrapped Bitcoin, and major stablecoins typically allow 70 to 90 percent LTV, so $10,000 of ETH might let you borrow $7,000 to $8,000. Volatile or long-tail tokens allow far less and are often ring-fenced in isolated markets. Your live health factor shows how close the position sits to liquidation.
What is a CDP stablecoin like GHO, USDS, or crvUSD?
A CDP, or collateralized debt position, stablecoin is a dollar you mint yourself by locking collateral in a protocol, rather than borrowing an existing coin from a pool. DAI and USDS from Sky, GHO from Aave, and crvUSD from Curve all work this way: deposit an asset, mint the stablecoin against it, then burn it to unlock your collateral. The borrow rate is set by governance and the protocol keeps the spread.
What is looping in DeFi lending, and is it risky?
Looping is recursive borrowing: you deposit a yield-bearing asset, borrow a stablecoin against it, buy more of the asset, redeposit, and repeat to multiply exposure. It can lift a 3 percent staking yield into low-double-digit returns, but it is leverage. If the collateral falls or the borrow rate rises above the yield, a looped position can unwind through cascading liquidations, which is what amplified the November 2025 contagion.
What happened with Stream Finance and curator risk?
In November 2025, the yield protocol Stream Finance collapsed when its xUSD token lost its peg, falling around 77 percent in a day and exposing roughly $285 million in interconnected debt across DeFi. Third-party curators on Morpho and Euler had funneled deposits into looped xUSD strategies chasing high yields. Isolated markets limited the direct bad debt to a small amount, but the episode exposed how much depositors were trusting curators they could not easily audit.
By the HOGE Wire markets desk, reporting on DeFi, on-chain credit, and market structure. This article is information, not financial advice.