Undercollateralized Lending in 2026: DeFi’s Credit Frontier

The wall every crypto loan hits

Almost every dollar borrowed in decentralized finance is backed by more than a dollar of something else. To take out $100 of a stablecoin on Aave or Morpho, a borrower usually locks $150 or more in Bitcoin, Ether, or another token, then hopes the price does not fall far enough to trigger a liquidation. It is a strange kind of credit: you can only borrow if you already have the money. Industry trackers estimate that more than 90% of DeFi loans work this way, which means the sector’s roughly $50 billion in outstanding on-chain credit sits on top of a much larger pile of collateral doing nothing else, a point made repeatedly by researchers studying on-chain credit scoring and confirmed by the category data on DefiLlama.

That design is why on-chain lending walked through every crisis of the last cycle while centralized lenders like Celsius and BlockFi collapsed. Overcollateralization is brutally effective. It is also brutally inefficient. The two giants of the category show the scale of what has been built on the assumption that a lender never has to trust a borrower: AAVE trades near $128 with a market capitalization around $1.98 billion (CoinGecko), and Morpho’s MORPHO sits near $2.34, worth about $1.6 billion (CoinGecko). Both are large businesses, and both mostly refuse to lend a cent that is not already covered.

The frontier, and the far harder problem, is lending where the borrower does not post more than they take. Traditional banks do this every day; it is most of what a bank is. On-chain, it means swapping collateral for something else: an identity, a credit score, a real cash flow, or a pool of junior capital willing to eat the first loss. With the secured overnight financing rate, the US benchmark for cash borrowing, sitting at 3.66% in early September, per SOFR data, the yields on offer in undercollateralized crypto credit (often 8% to 18%) are large enough to pull in serious money, especially with the Federal Reserve’s path still the biggest variable hanging over the market this autumn. The question every lender should ask is what, precisely, they are being paid to take on.

What undercollateralized actually means

The words get used loosely, so it helps to lay out a spectrum. Overcollateralized lending requires more than 100% collateral; that is nearly all of DeFi today. Undercollateralized lending accepts less than 100%, so a borrower might post $70 to draw $100. Uncollateralized or unsecured lending posts nothing at all and rests entirely on a promise to repay. In practice, people say “undercollateralized” to cover the whole lower half of that range, and this piece uses it the same way.

Whatever the label, the lender is giving up the one thing that makes DeFi lending safe: the ability to seize and sell collateral automatically the moment a position goes bad. Remove that, and something has to take its place. In 2026 there are four candidates, and the interesting protocols are really just different bets on which one works. The first is identity and legal recourse: the borrower is a known entity that can be sued. The second is a credit score, on-chain or off, that prices the odds of repayment. The third is a real, verifiable cash flow, such as an invoice or a cross-border payment already in transit, that repays the loan by itself. The fourth is subordinated capital: a junior tranche of lenders who agree to absorb the first losses so that senior lenders can feel safe.

None of these is new. Banks, factoring firms, and private-credit funds have used all four for a century. What is new is trying to run them on public infrastructure, where the loan is a smart contract but the borrower is still a human or a company that can simply stop paying. You cannot repossess a reputation, and code cannot garnish a bank account. That gap between what the chain can enforce and what a real loan requires is the whole story of this category.

2022: the first wave that broke

This is not the first attempt. The 2021 and 2022 cycle produced a whole class of on-chain lenders built to extend uncollateralized credit to crypto trading firms: Maple, founded in 2021, alongside TrueFi, Goldfinch, and Clearpool. The pitch was clean. Market makers and funds needed working capital, they were reputable counterparties, and lenders could earn double-digit yields by funding them through on-chain pools. For a while it worked beautifully, which is exactly when this kind of lending is most dangerous.

Then FTX failed, and the contagion ran straight through the uncollateralized books. On December 5, 2022, Orthogonal Trading defaulted on roughly $36 million of loans on Maple, spread across eight loans that made up close to 30% of the protocol’s active credit at the time, as The Block reported. Maple said Orthogonal had “misrepresented its financial position” for weeks and had been “operating while effectively insolvent,” per CoinDesk; Bloomberg noted the firm was later put into provisional liquidation. Lenders had no collateral to sell, so they simply took the loss.

The lesson was blunt and it still governs the category. Uncollateralized lending is not a technology problem, it is a trust problem, and trust fails fastest precisely when everyone needs their money back at once. The protocols that survived spent the next three years rebuilding around that reality rather than pretending it away.

On-chain credit scores: can a wallet have a FICO?

To lend without collateral, you have to price the borrower, and pricing a borrower means scoring them. A cluster of projects is trying to give wallets the on-chain equivalent of a credit file, built from repayment history, wallet age, transaction patterns, and past liquidations. Cred Protocol produces decentralized on-chain credit scores; Spectral markets a Multi-Asset Credit Risk Oracle, or MACRO, score derived from on-chain behavior; and RociFi, built on Polygon, uses machine learning to assign scores and issues them as non-transferable, soulbound NFTs so a reputation cannot be sold or handed off. Blockchain Bureau runs a similar on-chain analysis and is one of the inputs used by newer credit protocols.

Two problems hang over all of it. The first is the empty-file problem: a brand-new wallet has no history, and the borrowers most likely to want an unsecured loan often have the thinnest record. The second is the whitewash problem: if a bad score is just an address, throw it away and mint a new one. Soulbound tokens and off-chain identity binding are the answers on offer. Tools like zkTLS and Chainlink’s DECO let a borrower prove a real-world credit number, such as a VantageScore or a FICO figure, without publishing their name and financial life on a public ledger. The scoring works; the open question is whether the track record is deep enough to trust with real size yet.

Maple in 2026: institutional underwriting, rebuilt

Maple is the survivor, and it survived by getting more conservative, not less. The protocol still centers on pool delegates who underwrite loans with human judgment and stake their own reputation and capital on each decision, but most of its book is now overcollateralized, typically at 105% to 130% in Bitcoin, Ether, or Solana, rather than the pure unsecured lending that broke in 2022. The blend of real underwriting, hard collateral, and legal agreements is the point: Maple stopped treating reputation as a substitute for security and started treating it as a supplement.

The numbers show the rebuild working. Maple closed the first half of 2026 with about $4.6 billion in assets under management, up roughly 81% from a year earlier, and was earning interest on close to $2 billion in active loans, according to The Cryptonomist; the firm says it originated more than $11 billion in loans in 2025 alone. Its tokenized lending positions, syrupUSDC and syrupUSDT, were added to the 1inch aggregator in late July, widening their reach across DeFi, and Maple has partnered with Cantor Fitzgerald, the Wall Street firm pushing into digital assets, on a bitcoin-backed credit line reported at up to $2 billion, per The Big Whale. The SYRUP token trades near $0.22 with a market cap around $262 million, up more than 17% on the week (CoinGecko).

Chief executive Sid Powell has turned the whole shift into a slogan. “DeFi as a standalone category is dead,” he told CoinDesk in December, arguing that institutions will stop distinguishing DeFi from traditional markets as capital-markets activity settles on-chain. In the same breath he forecast a high-profile on-chain credit default in 2026, a reminder that the man building the biggest institutional lender expects someone in the category to blow up this year.

3Jane: an algorithmic credit line for a wallet

Where Maple retreated toward collateral, 3Jane is running straight at the unsecured problem. Backed by a $5.2 million seed round led by Paradigm, with Coinbase Ventures, Robot Ventures, and others joining, per The Block, the Ethereum-based protocol describes itself as a peer-to-pool credit-based money market offering real-time, unsecured USDC credit lines to yield farmers, traders, businesses, and even AI agents. That last group matters more than it sounds: as autonomous software starts to transact, the demand for machine-speed credit is one of the live questions hanging over the whole on-chain AI economy.

The engine is an off-chain underwriting algorithm called 3CA, which produces a borrower’s Jane Score. Per the protocol’s documentation, 3CA blends on-chain scores from Cred Protocol and Blockchain Bureau with off-chain VantageScore 3.0 data attested through zkTLS, then sets a credit line, a default-risk premium, and a repayment rate. Capital is split into two tranches: USD3, the senior token that earns a credit-enhanced share of yield, and sUSD3, the junior, first-loss token that earns a levered slice and absorbs losses before USD3 holders feel anything.

The part that separates 3Jane from the 2022 wave is what happens on default. A mechanism called the Credit Slasher deters walking away through Jane-score slashing and a pooled-upside model, and when a loan does sour, the non-performing debt is sold through auctions where licensed US collections agencies bid for the right to recover it. In other words, 3Jane wires real-world legal enforcement into the protocol rather than hoping borrowers behave. To bridge into consumer lending, the team also set up a $10 million warehouse line and a $50 million forward-flow agreement, per Crypto Briefing. It is the most complete attempt yet to make an unsecured on-chain loan behave like a real one, and it is deliberately aimed at US users, which pulls it into a much heavier regulatory lane.

Huma Finance: lending against receivables

Huma takes the most conservative route to the same destination. Instead of lending against who a borrower is, it lends against a specific payment they are already owed. The firm calls the category PayFi, and it runs primarily on Solana after launching Huma 2.0 there in 2025. The core insight is that a verified receivable is a near-ideal form of collateral: it is short-dated, it is tied to a real transaction, and it repays the loan by itself when the payment lands.

The mechanics are simple and fast. In cross-border payments, Huma supplies stablecoin liquidity to licensed payment institutions so they can settle transfers instantly, with each advance repaid within one to seven days and backed by funds already sitting in safeguarding accounts, according to the company’s own write-up. In trade finance, it works with TradeFlow Capital to fund physical commodities in transit, which typically settle in 30 to 90 days. Because each cycle is so short, the same dollar can finance many transactions across a year, which is how a modest amount of capital supports a large flow of volume.

That flow is now large. Huma reports more than $17 billion in cumulative transaction volume with zero credit defaults to date, growing by over $1 billion a month. It leans on Chainlink’s CCIP for cross-chain movement, a design choice that puts cross-chain security at the center of its risk model, since a bridge failure would hit the plumbing that moves its liquidity. The HUMA token trades near $0.022 for a market cap around $39 million (CoinGecko), a fraction of the volume it processes, and the protocol first expanded to Solana specifically to scale trade financing, as The Block covered.

Wildcat: hands-off and borrower-defined

Wildcat is the philosophical opposite of everything above, and it is the most honest about what undercollateralized lending really is. Built by Crypto Twitter mainstay Laurence Day with Indexed Finance’s Dillon Kellar in 2023, and with Wintermute’s Evgeny Gaevoy as a silent backer, Wildcat lets borrowers parameterize almost every term of a loan themselves: the reserve ratio, the withdrawal cycle, the collateral level (which can be zero), and which lenders are even allowed in. The protocol does not underwrite anything, and once a market is deployed it cannot liquidate collateral, freeze the market, or touch the funds. It raised a $3.5 million seed extension led by Robot Ventures in 2025, per The Block.

Day is candid about the trade-off. “We’re not dictating what it is that people are allowed to borrow or what asset or what rate. This is a fairly kind of free market approach,” he told The Block, adding that reputable borrowers “can truly just coast on their reputation if they wish.” His warning to lenders is the sharpest line in the whole category: “If you see something that’s a deal that looks too good to be true, like 20% on your USDC from some party that perhaps you haven’t heard of, it is.”

Wildcat got its first real stress test when Kinto, an Ethereum layer-2 network, shut down after a summer exploit drained $1.55 million and left it unable to repay its Wildcat market. It was the platform’s first default since launch. Lenders to the Kinto Phoenix Facility were set to recover about 76% of principal, and Wildcat stressed that the design held: “By design, the loss is limited to the Phoenix Facility alone, with no risk of contagion or haircut for any other lender to any other borrower,” it said, per The Block, noting it had over $150 million in outstanding credit across the platform at the time. Isolation contained the damage; it did nothing for the individual lenders who lost roughly a quarter of their money. That is the model working exactly as designed.

The four models, side by side

The protocols above are not really competitors so much as four different answers to the same question: if not collateral, then what? Maple bets on underwriting and legal wrappers, 3Jane on algorithmic credit scoring and collections, Huma on self-repaying receivables, and Wildcat on radical borrower freedom plus isolation. The legacy of the first wave still runs underneath all of it, with Goldfinch, founded in 2020, pioneering uncollateralized real-world credit to emerging-market lenders.

Read down the “on default” column and the real design space comes into focus. Every one of these systems is an attempt to answer a question that overcollateralized lending never has to ask: when the money does not come back, who chases it, and how?

Credit delegation, tranches, and who eats the first loss

Underneath the branding, most undercollateralized lending is built from two old private-credit tools. The first is credit delegation. Aave has long allowed a depositor to delegate their borrowing power to a trusted third party, who can then draw an uncollateralized loan against the delegator’s collateral, with the relationship usually backed by an off-chain legal agreement. The lender is choosing to extend their own good standing to someone else and accepting that they, not the protocol, carry the risk if it goes wrong.

The second tool is tranching, and it is everywhere once you look. Splitting a pool into a senior slice and a junior, first-loss slice is how you make an unsecured book palatable to conservative money. The junior tranche earns a higher, often levered return in exchange for absorbing losses before the senior tranche loses a cent. 3Jane’s USD3 and sUSD3 are a textbook version; Goldfinch’s backer and senior-pool structure is another. The mechanism is elegant and genuinely useful, but it hides a trap for the unwary: yield and subordination move together. If a product is paying you far more than the safe rate, you are almost certainly in a junior position, which means you are the collateral. The single most important thing a lender can know about any of these markets is whether they sit senior or first-loss, because that determines who is protecting whom.

This is also where the tokenized-real-world-credit boom connects to the story. Tokenized private credit passed roughly $14 billion in active on-chain loans by mid-2026, according to counts drawn from RWA.xyz, with Maple, Goldfinch, and Centrifuge among the largest venues. Almost all of that capital sits in exactly these senior and junior structures, ported from Wall Street onto public rails.

The enforcement problem nobody can code away

Here is the hard truth that every project in this space is quietly organized around. A smart contract can liquidate collateral in a single block, with no lawyers and no borders. It cannot garnish a paycheck, place a lien on a building, or repossess a bank account in another country. The moment a loan is undercollateralized, recovery leaves the chain and re-enters the slow, expensive, jurisdiction-bound world of actual debt collection. In on-chain security the theft is often the easy part and the recovery is the hard part; the same asymmetry runs through on-chain loss more broadly, and credit is no exception.

Each protocol’s answer to enforcement is really its whole identity. 3Jane binds borrowers with legal agreements, geofences to the US, and routes bad debt to licensed collections agencies, effectively renting the traditional legal system. Huma sidesteps enforcement almost entirely by lending only against payments that repay themselves within days. Wildcat refuses to pretend it can enforce anything, isolating each market so a default cannot spread. Maple leans on underwriting, collateral, and courts all at once. What none of them can do is make a borrower who has left the jurisdiction and spent the money suddenly repay. Strategic default, where a solvent borrower simply calculates that walking away is cheaper than paying, is the permanent tail risk, and it is why the yields have to be high enough to pay for the ones who do walk.

The yield you are actually being paid for

Undercollateralized lending advertises returns that look spectacular next to a 3.66% risk-free benchmark. They are not free money; they are the market’s price for a specific bundle of risks. The table below sketches the rough hierarchy, with yields stated as indicative ranges rather than promises, because the real number swings with the borrower, the tranche, and the cycle.

The spread between a 4% Treasury and a 15% private-credit position is not a gift from the protocol. It is compensation for credit risk, for illiquidity if your capital is locked to a maturity, for the possibility that enforcement fails, and, in a junior tranche, for standing in front of the losses. Yields of 8% to 18% are real, and so is the reason they exist. When the spread looks too wide for the risk described, the risk is usually being described wrong.

Security or not, bank or not: the US rules

Real credit touches real law, which makes undercollateralized lending a much bigger regulatory target than the overcollateralized kind. Start with the plainest fact: a DeFi lending protocol is not a bank. There is no deposit insurance, no lender of last resort, and no regulator standing behind the pool. When a borrower defaults and the buffers are gone, the loss falls on the people who supplied the capital. That is true of every protocol in this article.

Then there is the securities question. A tokenized lending position that pays a yield from the efforts of an underwriter, a syrupUSDC or a junior sUSD3 tranche, looks a great deal like an investment contract under the Howey test, and first-loss tranches sold for their return look even more so. The current Securities and Exchange Commission under Chair Paul Atkins has been friendlier to the industry, pushing a “Project Crypto” agenda that floats a DeFi safe harbor and clearer token rules, and Congress has been inching a market-structure bill toward a split of oversight between the SEC and the CFTC. But none of that has become settled law, and a US-facing lender that scores borrowers, signs credit agreements, and uses collections agencies, as 3Jane does, steps into the world of consumer-lending and fair-lending rules, state licensing, and debt-collection law that has nothing to do with crypto. That is why so many institutional pools, including Maple, TrueFi, and Clearpool, gate access to accredited or professional investors in the first place.

The tax treatment is its own maze. Borrowing against an asset instead of selling it can defer a taxable event, one of the durable attractions of crypto credit, but the rules on what counts as a disposal, and on how yield is taxed, are exactly the kind of detail the 2026 tax rulebook keeps reshaping. Anyone lending or borrowing size should assume the regulatory surface is still moving under their feet.

How to size up an undercollateralized position

The category is worth understanding because it is where on-chain credit has to go if it wants to be more than a leverage machine for people who are already rich in tokens. It is also where the next big losses will come from. Before supplying capital to any undercollateralized market, a lender can run through a short, unglamorous checklist.

• Who eats the first loss, and is it you? If the yield is unusually high, assume you are in a junior tranche until proven otherwise.
• What actually backs the loan: a legal claim, a verifiable receivable, a credit score, or just a good reputation and a promise?
• Can anyone enforce repayment? Look for a real jurisdiction, a named legal entity, and a defined collections or recovery process.
• What is the track record? A protocol reporting zero defaults through a benign period has told you very little about a bad one.
• Is the market isolated or pooled? Isolation limits contagion but does not protect the lenders inside a failed market.
• Can you exit? Check whether capital is redeemable on demand or locked to a maturity, and how it would behave in a rush for the door.
• Is it gated? Accredited-investor or geographic restrictions are a compliance signal, not a safety guarantee.
• Does the yield match the story? If the return is far above the risk being described, the description is the thing to distrust.

Undercollateralized lending is not a gimmick, and it is not going away. It is the part of finance that overcollateralized DeFi deliberately skipped, being rebuilt in the open by people who now understand why it is hard. The protocols that last will be the ones that treat trust as something to be priced, enforced, and isolated, rather than something to be assumed. The rest will teach the same lesson Orthogonal taught in 2022, at someone else’s expense.

Frequently Asked Questions

What is undercollateralized lending in DeFi?

Undercollateralized lending lets a borrower take out a loan worth more than the collateral they post, or with no collateral at all. Instead of relying only on locked assets, protocols price the borrower using identity, an on-chain or off-chain credit score, real cash flows such as invoices, or a pool of junior capital that absorbs losses first. It is how most traditional bank lending works, and it is the hardest problem in on-chain credit because trust is difficult to enforce without a way to seize assets.

Is uncollateralized crypto lending safe?

It carries more risk than overcollateralized lending. If a borrower does not repay, there is often no collateral to sell, so lenders can lose principal, as happened when Orthogonal Trading defaulted on $36 million on Maple in 2022. Newer protocols reduce risk with legal agreements, credit scoring, receivables backing, and isolated markets, but none remove it. There is no deposit insurance, so a lender who supplies to these markets is taking direct credit risk.

What is an on-chain credit score?

An on-chain credit score measures a wallet’s creditworthiness from its blockchain history, including repayment behavior, wallet age, and transaction patterns. Platforms such as Cred Protocol, Spectral, and RociFi generate these scores, and some combine them with off-chain data like a VantageScore or FICO number attested privately through zero-knowledge proofs. The scores let lenders price undercollateralized loans, though thin track records and wallet-swapping remain real challenges.

Can you get a crypto loan with no collateral in 2026?

Yes, in limited cases. Protocols such as 3Jane offer algorithmic USDC credit lines based on a combined on-chain and off-chain credit score, Huma Finance lends against verified receivables, and Wildcat lets vetted borrowers set their own terms. Most of these are geofenced, aimed at institutions or accredited investors, or backed by legal agreements, so a fully anonymous, no-collateral loan remains rare.

What happens if a borrower defaults on an undercollateralized loan?

It depends on the protocol. Some, like 3Jane, sell the bad debt to licensed US collections agencies and pursue borrowers through legal agreements. Others, like Wildcat, isolate each market so the loss falls only on that market’s lenders with no wider contagion. In pooled models, junior or first-loss capital is wiped out before senior lenders, and if losses exceed the buffer, suppliers absorb the rest. There is no central backstop.

Yuki Tanaka is a senior markets writer at HOGE Wire, covering DeFi credit, stablecoins, and on-chain infrastructure.