State of On-chain Lending
Lending first entered the cryptocurrency arena in 2017–18 with two distinct experiments. MakerDAO launched the collateralized debt position (CDP) model, allowing users to lock ETH and mint DAI stablecoins against it. At the same time, custodial desks, such as Genesis, issued bilateral loans against deposited BTC and USDC.
Both shared a goal of turning dormant crypto assets into working capital, but the former embedded the entire credit engine on a public blockchain, whereas the latter mirrored prime-brokerage structures off-chain, with its own range of problems and challenges.
Compound Finance introduced algorithmic money markets in 2018, pooling collateral and floating a variable-rate curve to offer more continuous liquidity. By abstracting away bilateral matching, Compound catalysed the yield farming era we now know so well. In short, this system revolved around lenders earning COMP incentives on top of base interest, and borrowers leveraging positions at a low cost because rates lagged behind token rewards.
Within 12 months, total supplied assets ballooned from less than $100 million to over $6 billion. Then came Aave in early 2020 with two major upgrades to Compound’s system. It tokenised deposits into interest-bearing aTokens, making yield composable across DeFi. Second, it debuted flash loans - instant, uncollateralized borrow-and-repay cycles within a single block - showcasing the trust minimization that only a synchronous, permissionless ledger permits. The combo propelled Aave to pole position; its V3 pool on Ethereum now holds $23.6 billion in collateral and $8.9 billion in active borrows, a larger book than many regional banks.
Unlike CeFi desks that negotiate covenants on an ad-hoc basis, on-chain protocols encode loan-to-value, liquidation bonuses, and oracle feeds directly into smart contracts.
If collateral‐to-debt ratios pass acceptable thresholds, MEV bots then compete to repay debt and seize collateral, ensuring the pool remains solvent without human intervention. This always-on margin clerk drastically shortens reaction time during market stress—a fact proven in the March 2020 COVID crash, when Maker, Compound, and Aave survived a 50% intraday drawdown without depositor losses.
Centralized lenders, such as Celsius, Voyager, and BlockFi, scaled rapidly through 2021, collectively holding $48 billion in outstanding loans at the cycle’s peak, followed by an equally rapid deleveraging. Their collapse in mid-2022, sparked mainly by maturity mismatches and poor collateral discipline, drove an 80% contraction in industry-wide credit. However, critically, DeFi protocols remained solvent, as transparent reserves allowed markets to re-price risk in real-time. By Q4 2024, DeFi accounted for 63% of the revived $30 billion loan book, reversing the 2021 balance.
Post-crisis growth has been fueled by liquid-staking tokens (stETH, rETH, eETH) and restaking receipts that embed a 3-4% real yield. These assets provide a type of self-repaying collateral, where staking income offsets part of the borrowing cost, enabling capital-efficient directional or basis trades.
As of March 2025, LSTs back more than $13.5 billion of borrows on Aave alone, two-thirds of which are looped leverage strategies.
Maker’s DAI still anchors the CDP segment, but competition from delta-neutral dollars such as Ethena’s USDe and Curve’s crvUSD has bifurcated liabilities. Protocols entice lenders by capturing funding rates: sUSDe has yielded double-digit APYs while maintaining a quasi-fiat peg. Maker responded by hiking its Enhanced DSR to 8% and launching Spark, a Maker-owned fork of Aave that recycles idle DAI into higher-yield RWA vaults.
Because positions are tokenized (cTokens, aTokens, sfrxETH, etc.), they can serve as collateral elsewhere, leveraged on Gearbox, traded on Pendle, or wrapped into structured products on Ribbon. This money lego property amplifies utility but also interconnects risk: a contract bug or oracle failure in one layer can propagate shocks across the stack, demanding rigorous audits and layered circuit-breakers.
Mature protocols have converged on DAO governance with specialist sub-committees (Aave-Risk, Compound-Gauntlet) proposing asset listings, LTV bands, and reserve factors. Fee-switches remain modest (Aave retains ≤ 10% of interest spread), but protocol treasuries increasingly diversify into RWAs or hedging vaults to stabilise income. Maker’s Endgame Roadmap, for example, targets 60% RWA-derived revenue by 2026.
Traditional broker-dealers are tiptoeing in: Cantor Fitzgerald unveiled a $2 billion bitcoin credit facility in January 2025, using segregated on-chain collateral accounts to satisfy auditors while tapping DeFi liquidity for rehypothecation. Meanwhile, Coinbase’s Base L2 offers whitelisted turbo USDC pools aimed at hedge-fund carry trades, signaling a blurring of the line between prime finance and permissionless rails.
Prospects brightened after the U.S. SEC proposed rescinding SAB-121 capital haircuts; banks may soon recognise customer crypto under custody without full balance-sheet surcharge, potentially unlocking cheaper funding lines for protocol-integrated tri-party repos. Europe’s MiCA, effective December 2024, already provides a passporting regime that several DAOs (Aave-Arc, Compound-Treasury) are using to onboard euro-zone corporations.
In April 2025, on-chain money markets boasted approximately $20 billion in active loans, surpassing their CeFi counterparts for the first time. They combine real-time transparency, programmable risk management, and composability unmatched in traditional credit. Yet they remain overwhelmingly over-collateralized—an elegant starting point, but capital-inefficient relative to real-world lending ratios.
The next frontier is extending crypto’s trust minimisation to under-collateralised credit, a challenge we will explore in Section 2.
The Future of Under-collateralized Lending
Uncollateralized or under-collateralized credit promises true capital efficiency: every on-chain dollar can be lent without first locking more than a dollar of collateral.
Yet, removing that over-collateral cushion means the protocol must answer three hard questions that banks normally handle: Who is the borrower? How do you price their default risk? And what happens if they do not pay? A handful of projects - including Maple, TrueFi, Goldfinch, and the newer Wildcat protocol - have spent the past four years designing radically different answers, but none has yet achieved the hands-off, trust-minimised ideal.
Most live designs fall into one of two buckets. Permissioned-pool protocols (Maple, TrueFi, Clearpool) require KYC’d borrowers, off-chain legal agreements, and dedicated pool delegates who underwrite risk, then tokenize the resulting loans so that anyone can supply liquidity. Pure on-chain reputation models (such as Wildcat and ongoing experiments with Ethereum attestations or soul-bound tokens) keep the credit contract entirely on-chain, allowing lenders to decide which pseudonymous identities they trust, albeit at the cost of significantly thinner enforcement powers in the event of issues. In practice, hybrids that mix legal recourse with automated liquidations are gaining traction.
After early pain with trader-credit defaults in 2022, Maple Finance rebuilt around over-collateralized but under-secured institutional credit.
Its Secured Lending line ballooned over 1,600% in 2024, from $32 million to $562 million TVL, while the active lender base grew over 14 times to 796 institutions. Borrowers post liquid BTC, ETH or SOL, but at lower haircuts than Aave requires, and Maple’s delegate stake plus margin-call engine handles enforcement. The average net annual percentage yield (APY) in the flagship pools ranged from 10% to 17%, outperforming passive stablecoin strategies throughout 2024.
MakerDAO’s Spark Protocol allocated an initial $50 million to Maple’s syrupUSDC vault in February 2025 - effectively turning stablecoin float into a senior tranche of under-collateralized loans without leaving DeFi rails. The deal illustrates how permissioned credit can be integrated directly into permissionless money markets, creating a yield ladder that begins with DAI deposits and ends with real-world receivables or liquid staking collateral.
TrueFi pioneered zero-collateral, term-loan pools in 2020 and has originated $1.7 billion in loans, with a lifetime default rate of 1-4%, comparable to lower-tier high-yield bonds. Its 2025 roadmap doubles down on Polaris, a yield-bearing stablecoin that sweeps idle capital into government-bill repos and TrueFi RWA vaults. The idea echoes Maple–Spark: package credit risk inside a token that can circulate through the rest of DeFi while funneling yield back to holders.
Warbler Labs’ Blueprint for Goldfinch (GRC-02) reframed the protocol as an operating system for private-credit businesses. Goldfinch has already financed $110 million of loans in 20+ emerging-market jurisdictions and now targets the $1.4 trillion global private-credit market through robo-advisor front-ends such as Heron Finance. Borrowers clear community-driven diligence, and lenders earn senior or junior tranche tokens whose exposure is crystal-clear on-chain, offering TradFi-style diversification without opaque fund wrappers.
Wildcat eschews delegates and legal docs. Its contracts let borrowers set their own collateral ratio (even 0%), maturity, and penalty terms; lenders opt in only if they like the risk-return profile. The protocol itself won’t liquidate, upgrade or seize assets once a market is deployed - it simply enforces whatever ex-ante rules the two sides codify. It is a pure experiment in contractual freedom, but critics note that real recovery still hinges on off-chain courts if a borrower defaults.
Clearpool’s 2024 roadmap introduced Credit Vaults and Clearpool Prime V2, adding collateralized corporate loans and exchange-traded pools across multiple chains. Centrifuge is going one layer deeper: Centrifuge V3 tokenizes a $230 million Janus Henderson Treasury fund and transfers it across chains via Wormhole, allowing those tokens to become borrow-side collateral in DeFi money markets.
Huma Finance pushes the frontier again by lending against tokenized receivables - PayFi - to give SMEs instant liquidity without traditional factoring fees.
For under-collateralized pools to scale, lenders need credibly private yet verifiable borrower data. Projects are experimenting with Ethereum Attestation Service, privacy-preserving credit scores, Chainlink proof-of-reserve attestations, and zero-knowledge income proofs. Clearpool already embeds off-chain credit reports in pool dashboards, whereas Maple requires pool delegates to stake MPL that is slashed in the event of default, aligning underwriting incentives with performance.
Most protocols now slice pools into senior and junior tranches - senior liquidity can withdraw first, while junior stakers absorb initial losses, mimicking CLO equity. Maple delegates stake 1–5% of the loan principal; Goldfinch backers do the same with GFI.
Independent insurers, such as Neptune Mutual, are piloting credit default swaps on Maple loans, and EigenLayer restaking is being explored as a potential universal guarantee fund. Together, these mechanisms substitute skin-in-the-game and portfolio diversification for hard collateral.
Enforcement remains the Achilles’ heel: without collateral, recovery relies on off-chain courts whose timelines dwarf blockchain settlement. Information asymmetry is another - borrowers know far more about their finances than the public does. Regulation looms as both risk and catalyst: KYC/AML mandates raise costs, but the SEC’s tentative retreat from punitive custody rules could open bank balance sheets to on-chain private credit.
A Centrifuge / Keyrock study pegs tokenized private credit TVL at $12–17.5 billion by 2027, depending on macro and regulatory tailwinds. The base-case path runs through permissioned but composable pools: Maple and TrueFi scale institutional books, Goldfinch onboards retail via robo-advisors, and Clearpool or Centrifuge bridges RWA collateral back into Aave-style money markets. If those rails prove robust through a default cycle, fully permissionless experiments like Wildcat could finally have the on-chain identity primitives they need to thrive, unlocking the capital-efficient, global credit layer DeFi has chased since its inception.
Next Steps
The past five years have shown that permissionless plumbing + permissioned underwriting beats either extreme. Maple’s 16-fold TVL expansion in 2024 proves that institutional lenders will move on-chain - if they can rely on professional delegates, real-time collateral monitoring ,and fast liquidation rails.
The lesson: replicate the parts of TradFi that work (credit committees, senior-junior tranching) and outsource the rest to immutable code.
Goldfinch Prime packages senior-secured loans from Ares, Apollo, and Golub Capital into a single ERC-4626 vault, providing on-chain capital holders with exposure to thousands of SME loans that they could not source themselves. Expect Maple’s Syrup.fi, Clearpool Prime V2, and TrueFi’s forthcoming tfBill integrations to copy-paste that structure: wrap a yield-bearing senior token that can plug straight into Aave, Morpho, or Pendle for rehypothecation.
Spark’s $25m anchor position in Maple pools (with an option to double) hints at a broader trend: DAO treasuries redeploy idle stablecoin float into under-collateralized credit, then borrow against the resulting receipt to keep liquidity on hand. The loop converts dead capital into stacked yield without leaving the cryptographic trust boundary - the first institutional-grade example of money-legos in credit.
Wildcat’s free-banking experiment - let borrowers publish any mix of reserves, attestations or social proofs and let lenders decide - shines a spotlight on the missing piece: portable, privacy-preserving reputation. Ongoing pilots with Ethereum Attestation Service, zero-knowledge income proofs, and Chainlink Proof-of-Reserve feeds should be prioritized by every credit DAO; without credible private disclosure, under-collateralized lending will remain capped at KYC-gated pools.
Nearly every protocol now stakes delegate capital to first-loss tranches, but coverage remains shallow relative to loan books. Two levers can harden the stack: (a) external underwriters such as Neptune Mutual issuing CDS-style covers on Maple loans, and (b) EigenLayer restaking modules that slash validator yield to recapitalise pools in tail events. DAOs should treat these as non-negotiable expenditures, not optional nice-to-haves.
TrueFi’s 2025 roadmap divorces day-to-day underwriting from token-holder voting; instead, fixed mandates and time-boxed KPIs guide pool managers, while TRU holders review only escalation events. Copying this model reduces decision latency and dampens political risk - a prerequisite if regulated institutions are to trust DAO courts.
Europe’s MiCA regime already recognizes tokenized funds and loan notes, while the U.S. is tilting friendlier after the SEC signaled a retreat from punitive custody guidance. Protocols should localize legal wrappers in MiCA jurisdictions now, then utilize passporting to reissue yield tokens on global L1 and L2 networks once U.S. clarity is established. KYC-compliant sub-pools (Maple Insto, Goldfinch Prime) are the low-friction wedge product.
Centrifuge and Clearpool are already tokenising trade receivables and T-bills; next comes equipment leases, carbon credits, and even power-purchase agreements. Lenders get uncorrelated cash flows, borrowers get cheaper capital via crypto’s robo-infrastructure. Every credit DAO should earmark bandwidth for oracle partnerships and legal enforceability audits now, before the asset class floods in.
Default management is still painfully manual. Smart-contract escrow of collateral (or revenue share) plus automated payment routing can shrink resolution timelines from months to minutes. Pool delegates should push for programmatic step-downs: if the coverage ratio is less than 100%, a predefined settlement contract seizes escrowed cash flows, sells collateral, or triggers CDS payouts without boardroom drama. Borrowers shouldn’t need to use a wallet like MetaMask, and lenders should not need to read Solidity repos.
Maple already white-labels its APIs to exchanges and OTC desks; Goldfinch is shipping robo-advisor front-ends; TrueFi is overhauling UI/UX ahead of the Polaris stablecoin launch. The winner will be the stack that on-boards a CFO in three clicks, not the one with the cleverest yield engine.
Going forward, protocols must publish loss-adjusted APY, default frequency, recovery lag, and utilisation-weighted liquidity. Maple’s 61 margin-calls in 2024 were cured in three hours on average - a statistic more persuasive to institutions than the industry standard APR banners.
Achieving that curve requires (a) at least one transparent default cycle that proves automated workouts, (b) two or more regulated broker-dealers to warehouse senior tranches, and (c) settlement of U.S. custody rules. Developers should therefore prioritize audit-grade liquidation logic and stable UX. Investors should focus on protocols with skin-in-the-game delegates and external insurance. Policymakers should monitor the robust data trails these systems emit and craft disclosure-first, not permission-first, regimes.
The bottom line is that under-collateralized lending is no longer a moon-shot idea; it is a small but rapidly compounding asset class with credible product-market fit. The next steps are mostly operational, including identity, insurance, governance hygiene, and data standards. The protocols that execute these details will capture the bulk of the coming $40 billion on-chain credit boom.
State of On-chain Lending
Lending first entered the cryptocurrency arena in 2017–18 with two distinct experiments. MakerDAO launched the collateralized debt position (CDP) model, allowing users to lock ETH and mint DAI stablecoins against it. At the same time, custodial desks, such as Genesis, issued bilateral loans against deposited BTC and USDC.
Both shared a goal of turning dormant crypto assets into working capital, but the former embedded the entire credit engine on a public blockchain, whereas the latter mirrored prime-brokerage structures off-chain, with its own range of problems and challenges.
Compound Finance introduced algorithmic money markets in 2018, pooling collateral and floating a variable-rate curve to offer more continuous liquidity. By abstracting away bilateral matching, Compound catalysed the yield farming era we now know so well. In short, this system revolved around lenders earning COMP incentives on top of base interest, and borrowers leveraging positions at a low cost because rates lagged behind token rewards.
Within 12 months, total supplied assets ballooned from less than $100 million to over $6 billion. Then came Aave in early 2020 with two major upgrades to Compound’s system. It tokenised deposits into interest-bearing aTokens, making yield composable across DeFi. Second, it debuted flash loans - instant, uncollateralized borrow-and-repay cycles within a single block - showcasing the trust minimization that only a synchronous, permissionless ledger permits. The combo propelled Aave to pole position; its V3 pool on Ethereum now holds $23.6 billion in collateral and $8.9 billion in active borrows, a larger book than many regional banks.
Unlike CeFi desks that negotiate covenants on an ad-hoc basis, on-chain protocols encode loan-to-value, liquidation bonuses, and oracle feeds directly into smart contracts.
If collateral‐to-debt ratios pass acceptable thresholds, MEV bots then compete to repay debt and seize collateral, ensuring the pool remains solvent without human intervention. This always-on margin clerk drastically shortens reaction time during market stress—a fact proven in the March 2020 COVID crash, when Maker, Compound, and Aave survived a 50% intraday drawdown without depositor losses.
Centralized lenders, such as Celsius, Voyager, and BlockFi, scaled rapidly through 2021, collectively holding $48 billion in outstanding loans at the cycle’s peak, followed by an equally rapid deleveraging. Their collapse in mid-2022, sparked mainly by maturity mismatches and poor collateral discipline, drove an 80% contraction in industry-wide credit. However, critically, DeFi protocols remained solvent, as transparent reserves allowed markets to re-price risk in real-time. By Q4 2024, DeFi accounted for 63% of the revived $30 billion loan book, reversing the 2021 balance.
Post-crisis growth has been fueled by liquid-staking tokens (stETH, rETH, eETH) and restaking receipts that embed a 3-4% real yield. These assets provide a type of self-repaying collateral, where staking income offsets part of the borrowing cost, enabling capital-efficient directional or basis trades.
As of March 2025, LSTs back more than $13.5 billion of borrows on Aave alone, two-thirds of which are looped leverage strategies.
Maker’s DAI still anchors the CDP segment, but competition from delta-neutral dollars such as Ethena’s USDe and Curve’s crvUSD has bifurcated liabilities. Protocols entice lenders by capturing funding rates: sUSDe has yielded double-digit APYs while maintaining a quasi-fiat peg. Maker responded by hiking its Enhanced DSR to 8% and launching Spark, a Maker-owned fork of Aave that recycles idle DAI into higher-yield RWA vaults.
Because positions are tokenized (cTokens, aTokens, sfrxETH, etc.), they can serve as collateral elsewhere, leveraged on Gearbox, traded on Pendle, or wrapped into structured products on Ribbon. This money lego property amplifies utility but also interconnects risk: a contract bug or oracle failure in one layer can propagate shocks across the stack, demanding rigorous audits and layered circuit-breakers.
Mature protocols have converged on DAO governance with specialist sub-committees (Aave-Risk, Compound-Gauntlet) proposing asset listings, LTV bands, and reserve factors. Fee-switches remain modest (Aave retains ≤ 10% of interest spread), but protocol treasuries increasingly diversify into RWAs or hedging vaults to stabilise income. Maker’s Endgame Roadmap, for example, targets 60% RWA-derived revenue by 2026.
Traditional broker-dealers are tiptoeing in: Cantor Fitzgerald unveiled a $2 billion bitcoin credit facility in January 2025, using segregated on-chain collateral accounts to satisfy auditors while tapping DeFi liquidity for rehypothecation. Meanwhile, Coinbase’s Base L2 offers whitelisted turbo USDC pools aimed at hedge-fund carry trades, signaling a blurring of the line between prime finance and permissionless rails.
Prospects brightened after the U.S. SEC proposed rescinding SAB-121 capital haircuts; banks may soon recognise customer crypto under custody without full balance-sheet surcharge, potentially unlocking cheaper funding lines for protocol-integrated tri-party repos. Europe’s MiCA, effective December 2024, already provides a passporting regime that several DAOs (Aave-Arc, Compound-Treasury) are using to onboard euro-zone corporations.
In April 2025, on-chain money markets boasted approximately $20 billion in active loans, surpassing their CeFi counterparts for the first time. They combine real-time transparency, programmable risk management, and composability unmatched in traditional credit. Yet they remain overwhelmingly over-collateralized—an elegant starting point, but capital-inefficient relative to real-world lending ratios.
The next frontier is extending crypto’s trust minimisation to under-collateralised credit, a challenge we will explore in Section 2.
The Future of Under-collateralized Lending
Uncollateralized or under-collateralized credit promises true capital efficiency: every on-chain dollar can be lent without first locking more than a dollar of collateral.
Yet, removing that over-collateral cushion means the protocol must answer three hard questions that banks normally handle: Who is the borrower? How do you price their default risk? And what happens if they do not pay? A handful of projects - including Maple, TrueFi, Goldfinch, and the newer Wildcat protocol - have spent the past four years designing radically different answers, but none has yet achieved the hands-off, trust-minimised ideal.
Most live designs fall into one of two buckets. Permissioned-pool protocols (Maple, TrueFi, Clearpool) require KYC’d borrowers, off-chain legal agreements, and dedicated pool delegates who underwrite risk, then tokenize the resulting loans so that anyone can supply liquidity. Pure on-chain reputation models (such as Wildcat and ongoing experiments with Ethereum attestations or soul-bound tokens) keep the credit contract entirely on-chain, allowing lenders to decide which pseudonymous identities they trust, albeit at the cost of significantly thinner enforcement powers in the event of issues. In practice, hybrids that mix legal recourse with automated liquidations are gaining traction.
After early pain with trader-credit defaults in 2022, Maple Finance rebuilt around over-collateralized but under-secured institutional credit.
Its Secured Lending line ballooned over 1,600% in 2024, from $32 million to $562 million TVL, while the active lender base grew over 14 times to 796 institutions. Borrowers post liquid BTC, ETH or SOL, but at lower haircuts than Aave requires, and Maple’s delegate stake plus margin-call engine handles enforcement. The average net annual percentage yield (APY) in the flagship pools ranged from 10% to 17%, outperforming passive stablecoin strategies throughout 2024.
MakerDAO’s Spark Protocol allocated an initial $50 million to Maple’s syrupUSDC vault in February 2025 - effectively turning stablecoin float into a senior tranche of under-collateralized loans without leaving DeFi rails. The deal illustrates how permissioned credit can be integrated directly into permissionless money markets, creating a yield ladder that begins with DAI deposits and ends with real-world receivables or liquid staking collateral.
TrueFi pioneered zero-collateral, term-loan pools in 2020 and has originated $1.7 billion in loans, with a lifetime default rate of 1-4%, comparable to lower-tier high-yield bonds. Its 2025 roadmap doubles down on Polaris, a yield-bearing stablecoin that sweeps idle capital into government-bill repos and TrueFi RWA vaults. The idea echoes Maple–Spark: package credit risk inside a token that can circulate through the rest of DeFi while funneling yield back to holders.
Warbler Labs’ Blueprint for Goldfinch (GRC-02) reframed the protocol as an operating system for private-credit businesses. Goldfinch has already financed $110 million of loans in 20+ emerging-market jurisdictions and now targets the $1.4 trillion global private-credit market through robo-advisor front-ends such as Heron Finance. Borrowers clear community-driven diligence, and lenders earn senior or junior tranche tokens whose exposure is crystal-clear on-chain, offering TradFi-style diversification without opaque fund wrappers.
Wildcat eschews delegates and legal docs. Its contracts let borrowers set their own collateral ratio (even 0%), maturity, and penalty terms; lenders opt in only if they like the risk-return profile. The protocol itself won’t liquidate, upgrade or seize assets once a market is deployed - it simply enforces whatever ex-ante rules the two sides codify. It is a pure experiment in contractual freedom, but critics note that real recovery still hinges on off-chain courts if a borrower defaults.
Clearpool’s 2024 roadmap introduced Credit Vaults and Clearpool Prime V2, adding collateralized corporate loans and exchange-traded pools across multiple chains. Centrifuge is going one layer deeper: Centrifuge V3 tokenizes a $230 million Janus Henderson Treasury fund and transfers it across chains via Wormhole, allowing those tokens to become borrow-side collateral in DeFi money markets.
Huma Finance pushes the frontier again by lending against tokenized receivables - PayFi - to give SMEs instant liquidity without traditional factoring fees.
For under-collateralized pools to scale, lenders need credibly private yet verifiable borrower data. Projects are experimenting with Ethereum Attestation Service, privacy-preserving credit scores, Chainlink proof-of-reserve attestations, and zero-knowledge income proofs. Clearpool already embeds off-chain credit reports in pool dashboards, whereas Maple requires pool delegates to stake MPL that is slashed in the event of default, aligning underwriting incentives with performance.
Most protocols now slice pools into senior and junior tranches - senior liquidity can withdraw first, while junior stakers absorb initial losses, mimicking CLO equity. Maple delegates stake 1–5% of the loan principal; Goldfinch backers do the same with GFI.
Independent insurers, such as Neptune Mutual, are piloting credit default swaps on Maple loans, and EigenLayer restaking is being explored as a potential universal guarantee fund. Together, these mechanisms substitute skin-in-the-game and portfolio diversification for hard collateral.
Enforcement remains the Achilles’ heel: without collateral, recovery relies on off-chain courts whose timelines dwarf blockchain settlement. Information asymmetry is another - borrowers know far more about their finances than the public does. Regulation looms as both risk and catalyst: KYC/AML mandates raise costs, but the SEC’s tentative retreat from punitive custody rules could open bank balance sheets to on-chain private credit.
A Centrifuge / Keyrock study pegs tokenized private credit TVL at $12–17.5 billion by 2027, depending on macro and regulatory tailwinds. The base-case path runs through permissioned but composable pools: Maple and TrueFi scale institutional books, Goldfinch onboards retail via robo-advisors, and Clearpool or Centrifuge bridges RWA collateral back into Aave-style money markets. If those rails prove robust through a default cycle, fully permissionless experiments like Wildcat could finally have the on-chain identity primitives they need to thrive, unlocking the capital-efficient, global credit layer DeFi has chased since its inception.
Next Steps
The past five years have shown that permissionless plumbing + permissioned underwriting beats either extreme. Maple’s 16-fold TVL expansion in 2024 proves that institutional lenders will move on-chain - if they can rely on professional delegates, real-time collateral monitoring ,and fast liquidation rails.
The lesson: replicate the parts of TradFi that work (credit committees, senior-junior tranching) and outsource the rest to immutable code.
Goldfinch Prime packages senior-secured loans from Ares, Apollo, and Golub Capital into a single ERC-4626 vault, providing on-chain capital holders with exposure to thousands of SME loans that they could not source themselves. Expect Maple’s Syrup.fi, Clearpool Prime V2, and TrueFi’s forthcoming tfBill integrations to copy-paste that structure: wrap a yield-bearing senior token that can plug straight into Aave, Morpho, or Pendle for rehypothecation.
Spark’s $25m anchor position in Maple pools (with an option to double) hints at a broader trend: DAO treasuries redeploy idle stablecoin float into under-collateralized credit, then borrow against the resulting receipt to keep liquidity on hand. The loop converts dead capital into stacked yield without leaving the cryptographic trust boundary - the first institutional-grade example of money-legos in credit.
Wildcat’s free-banking experiment - let borrowers publish any mix of reserves, attestations or social proofs and let lenders decide - shines a spotlight on the missing piece: portable, privacy-preserving reputation. Ongoing pilots with Ethereum Attestation Service, zero-knowledge income proofs, and Chainlink Proof-of-Reserve feeds should be prioritized by every credit DAO; without credible private disclosure, under-collateralized lending will remain capped at KYC-gated pools.
Nearly every protocol now stakes delegate capital to first-loss tranches, but coverage remains shallow relative to loan books. Two levers can harden the stack: (a) external underwriters such as Neptune Mutual issuing CDS-style covers on Maple loans, and (b) EigenLayer restaking modules that slash validator yield to recapitalise pools in tail events. DAOs should treat these as non-negotiable expenditures, not optional nice-to-haves.
TrueFi’s 2025 roadmap divorces day-to-day underwriting from token-holder voting; instead, fixed mandates and time-boxed KPIs guide pool managers, while TRU holders review only escalation events. Copying this model reduces decision latency and dampens political risk - a prerequisite if regulated institutions are to trust DAO courts.
Europe’s MiCA regime already recognizes tokenized funds and loan notes, while the U.S. is tilting friendlier after the SEC signaled a retreat from punitive custody guidance. Protocols should localize legal wrappers in MiCA jurisdictions now, then utilize passporting to reissue yield tokens on global L1 and L2 networks once U.S. clarity is established. KYC-compliant sub-pools (Maple Insto, Goldfinch Prime) are the low-friction wedge product.
Centrifuge and Clearpool are already tokenising trade receivables and T-bills; next comes equipment leases, carbon credits, and even power-purchase agreements. Lenders get uncorrelated cash flows, borrowers get cheaper capital via crypto’s robo-infrastructure. Every credit DAO should earmark bandwidth for oracle partnerships and legal enforceability audits now, before the asset class floods in.
Default management is still painfully manual. Smart-contract escrow of collateral (or revenue share) plus automated payment routing can shrink resolution timelines from months to minutes. Pool delegates should push for programmatic step-downs: if the coverage ratio is less than 100%, a predefined settlement contract seizes escrowed cash flows, sells collateral, or triggers CDS payouts without boardroom drama. Borrowers shouldn’t need to use a wallet like MetaMask, and lenders should not need to read Solidity repos.
Maple already white-labels its APIs to exchanges and OTC desks; Goldfinch is shipping robo-advisor front-ends; TrueFi is overhauling UI/UX ahead of the Polaris stablecoin launch. The winner will be the stack that on-boards a CFO in three clicks, not the one with the cleverest yield engine.
Going forward, protocols must publish loss-adjusted APY, default frequency, recovery lag, and utilisation-weighted liquidity. Maple’s 61 margin-calls in 2024 were cured in three hours on average - a statistic more persuasive to institutions than the industry standard APR banners.
Achieving that curve requires (a) at least one transparent default cycle that proves automated workouts, (b) two or more regulated broker-dealers to warehouse senior tranches, and (c) settlement of U.S. custody rules. Developers should therefore prioritize audit-grade liquidation logic and stable UX. Investors should focus on protocols with skin-in-the-game delegates and external insurance. Policymakers should monitor the robust data trails these systems emit and craft disclosure-first, not permission-first, regimes.
The bottom line is that under-collateralized lending is no longer a moon-shot idea; it is a small but rapidly compounding asset class with credible product-market fit. The next steps are mostly operational, including identity, insurance, governance hygiene, and data standards. The protocols that execute these details will capture the bulk of the coming $40 billion on-chain credit boom.
State of On-chain Lending
Lending first entered the cryptocurrency arena in 2017–18 with two distinct experiments. MakerDAO launched the collateralized debt position (CDP) model, allowing users to lock ETH and mint DAI stablecoins against it. At the same time, custodial desks, such as Genesis, issued bilateral loans against deposited BTC and USDC.
Both shared a goal of turning dormant crypto assets into working capital, but the former embedded the entire credit engine on a public blockchain, whereas the latter mirrored prime-brokerage structures off-chain, with its own range of problems and challenges.
Compound Finance introduced algorithmic money markets in 2018, pooling collateral and floating a variable-rate curve to offer more continuous liquidity. By abstracting away bilateral matching, Compound catalysed the yield farming era we now know so well. In short, this system revolved around lenders earning COMP incentives on top of base interest, and borrowers leveraging positions at a low cost because rates lagged behind token rewards.
Within 12 months, total supplied assets ballooned from less than $100 million to over $6 billion. Then came Aave in early 2020 with two major upgrades to Compound’s system. It tokenised deposits into interest-bearing aTokens, making yield composable across DeFi. Second, it debuted flash loans - instant, uncollateralized borrow-and-repay cycles within a single block - showcasing the trust minimization that only a synchronous, permissionless ledger permits. The combo propelled Aave to pole position; its V3 pool on Ethereum now holds $23.6 billion in collateral and $8.9 billion in active borrows, a larger book than many regional banks.
Unlike CeFi desks that negotiate covenants on an ad-hoc basis, on-chain protocols encode loan-to-value, liquidation bonuses, and oracle feeds directly into smart contracts.
If collateral‐to-debt ratios pass acceptable thresholds, MEV bots then compete to repay debt and seize collateral, ensuring the pool remains solvent without human intervention. This always-on margin clerk drastically shortens reaction time during market stress—a fact proven in the March 2020 COVID crash, when Maker, Compound, and Aave survived a 50% intraday drawdown without depositor losses.
Centralized lenders, such as Celsius, Voyager, and BlockFi, scaled rapidly through 2021, collectively holding $48 billion in outstanding loans at the cycle’s peak, followed by an equally rapid deleveraging. Their collapse in mid-2022, sparked mainly by maturity mismatches and poor collateral discipline, drove an 80% contraction in industry-wide credit. However, critically, DeFi protocols remained solvent, as transparent reserves allowed markets to re-price risk in real-time. By Q4 2024, DeFi accounted for 63% of the revived $30 billion loan book, reversing the 2021 balance.
Post-crisis growth has been fueled by liquid-staking tokens (stETH, rETH, eETH) and restaking receipts that embed a 3-4% real yield. These assets provide a type of self-repaying collateral, where staking income offsets part of the borrowing cost, enabling capital-efficient directional or basis trades.
As of March 2025, LSTs back more than $13.5 billion of borrows on Aave alone, two-thirds of which are looped leverage strategies.
Maker’s DAI still anchors the CDP segment, but competition from delta-neutral dollars such as Ethena’s USDe and Curve’s crvUSD has bifurcated liabilities. Protocols entice lenders by capturing funding rates: sUSDe has yielded double-digit APYs while maintaining a quasi-fiat peg. Maker responded by hiking its Enhanced DSR to 8% and launching Spark, a Maker-owned fork of Aave that recycles idle DAI into higher-yield RWA vaults.
Because positions are tokenized (cTokens, aTokens, sfrxETH, etc.), they can serve as collateral elsewhere, leveraged on Gearbox, traded on Pendle, or wrapped into structured products on Ribbon. This money lego property amplifies utility but also interconnects risk: a contract bug or oracle failure in one layer can propagate shocks across the stack, demanding rigorous audits and layered circuit-breakers.
Mature protocols have converged on DAO governance with specialist sub-committees (Aave-Risk, Compound-Gauntlet) proposing asset listings, LTV bands, and reserve factors. Fee-switches remain modest (Aave retains ≤ 10% of interest spread), but protocol treasuries increasingly diversify into RWAs or hedging vaults to stabilise income. Maker’s Endgame Roadmap, for example, targets 60% RWA-derived revenue by 2026.
Traditional broker-dealers are tiptoeing in: Cantor Fitzgerald unveiled a $2 billion bitcoin credit facility in January 2025, using segregated on-chain collateral accounts to satisfy auditors while tapping DeFi liquidity for rehypothecation. Meanwhile, Coinbase’s Base L2 offers whitelisted turbo USDC pools aimed at hedge-fund carry trades, signaling a blurring of the line between prime finance and permissionless rails.
Prospects brightened after the U.S. SEC proposed rescinding SAB-121 capital haircuts; banks may soon recognise customer crypto under custody without full balance-sheet surcharge, potentially unlocking cheaper funding lines for protocol-integrated tri-party repos. Europe’s MiCA, effective December 2024, already provides a passporting regime that several DAOs (Aave-Arc, Compound-Treasury) are using to onboard euro-zone corporations.
In April 2025, on-chain money markets boasted approximately $20 billion in active loans, surpassing their CeFi counterparts for the first time. They combine real-time transparency, programmable risk management, and composability unmatched in traditional credit. Yet they remain overwhelmingly over-collateralized—an elegant starting point, but capital-inefficient relative to real-world lending ratios.
The next frontier is extending crypto’s trust minimisation to under-collateralised credit, a challenge we will explore in Section 2.
The Future of Under-collateralized Lending
Uncollateralized or under-collateralized credit promises true capital efficiency: every on-chain dollar can be lent without first locking more than a dollar of collateral.
Yet, removing that over-collateral cushion means the protocol must answer three hard questions that banks normally handle: Who is the borrower? How do you price their default risk? And what happens if they do not pay? A handful of projects - including Maple, TrueFi, Goldfinch, and the newer Wildcat protocol - have spent the past four years designing radically different answers, but none has yet achieved the hands-off, trust-minimised ideal.
Most live designs fall into one of two buckets. Permissioned-pool protocols (Maple, TrueFi, Clearpool) require KYC’d borrowers, off-chain legal agreements, and dedicated pool delegates who underwrite risk, then tokenize the resulting loans so that anyone can supply liquidity. Pure on-chain reputation models (such as Wildcat and ongoing experiments with Ethereum attestations or soul-bound tokens) keep the credit contract entirely on-chain, allowing lenders to decide which pseudonymous identities they trust, albeit at the cost of significantly thinner enforcement powers in the event of issues. In practice, hybrids that mix legal recourse with automated liquidations are gaining traction.
After early pain with trader-credit defaults in 2022, Maple Finance rebuilt around over-collateralized but under-secured institutional credit.
Its Secured Lending line ballooned over 1,600% in 2024, from $32 million to $562 million TVL, while the active lender base grew over 14 times to 796 institutions. Borrowers post liquid BTC, ETH or SOL, but at lower haircuts than Aave requires, and Maple’s delegate stake plus margin-call engine handles enforcement. The average net annual percentage yield (APY) in the flagship pools ranged from 10% to 17%, outperforming passive stablecoin strategies throughout 2024.
MakerDAO’s Spark Protocol allocated an initial $50 million to Maple’s syrupUSDC vault in February 2025 - effectively turning stablecoin float into a senior tranche of under-collateralized loans without leaving DeFi rails. The deal illustrates how permissioned credit can be integrated directly into permissionless money markets, creating a yield ladder that begins with DAI deposits and ends with real-world receivables or liquid staking collateral.
TrueFi pioneered zero-collateral, term-loan pools in 2020 and has originated $1.7 billion in loans, with a lifetime default rate of 1-4%, comparable to lower-tier high-yield bonds. Its 2025 roadmap doubles down on Polaris, a yield-bearing stablecoin that sweeps idle capital into government-bill repos and TrueFi RWA vaults. The idea echoes Maple–Spark: package credit risk inside a token that can circulate through the rest of DeFi while funneling yield back to holders.
Warbler Labs’ Blueprint for Goldfinch (GRC-02) reframed the protocol as an operating system for private-credit businesses. Goldfinch has already financed $110 million of loans in 20+ emerging-market jurisdictions and now targets the $1.4 trillion global private-credit market through robo-advisor front-ends such as Heron Finance. Borrowers clear community-driven diligence, and lenders earn senior or junior tranche tokens whose exposure is crystal-clear on-chain, offering TradFi-style diversification without opaque fund wrappers.
Wildcat eschews delegates and legal docs. Its contracts let borrowers set their own collateral ratio (even 0%), maturity, and penalty terms; lenders opt in only if they like the risk-return profile. The protocol itself won’t liquidate, upgrade or seize assets once a market is deployed - it simply enforces whatever ex-ante rules the two sides codify. It is a pure experiment in contractual freedom, but critics note that real recovery still hinges on off-chain courts if a borrower defaults.
Clearpool’s 2024 roadmap introduced Credit Vaults and Clearpool Prime V2, adding collateralized corporate loans and exchange-traded pools across multiple chains. Centrifuge is going one layer deeper: Centrifuge V3 tokenizes a $230 million Janus Henderson Treasury fund and transfers it across chains via Wormhole, allowing those tokens to become borrow-side collateral in DeFi money markets.
Huma Finance pushes the frontier again by lending against tokenized receivables - PayFi - to give SMEs instant liquidity without traditional factoring fees.
For under-collateralized pools to scale, lenders need credibly private yet verifiable borrower data. Projects are experimenting with Ethereum Attestation Service, privacy-preserving credit scores, Chainlink proof-of-reserve attestations, and zero-knowledge income proofs. Clearpool already embeds off-chain credit reports in pool dashboards, whereas Maple requires pool delegates to stake MPL that is slashed in the event of default, aligning underwriting incentives with performance.
Most protocols now slice pools into senior and junior tranches - senior liquidity can withdraw first, while junior stakers absorb initial losses, mimicking CLO equity. Maple delegates stake 1–5% of the loan principal; Goldfinch backers do the same with GFI.
Independent insurers, such as Neptune Mutual, are piloting credit default swaps on Maple loans, and EigenLayer restaking is being explored as a potential universal guarantee fund. Together, these mechanisms substitute skin-in-the-game and portfolio diversification for hard collateral.
Enforcement remains the Achilles’ heel: without collateral, recovery relies on off-chain courts whose timelines dwarf blockchain settlement. Information asymmetry is another - borrowers know far more about their finances than the public does. Regulation looms as both risk and catalyst: KYC/AML mandates raise costs, but the SEC’s tentative retreat from punitive custody rules could open bank balance sheets to on-chain private credit.
A Centrifuge / Keyrock study pegs tokenized private credit TVL at $12–17.5 billion by 2027, depending on macro and regulatory tailwinds. The base-case path runs through permissioned but composable pools: Maple and TrueFi scale institutional books, Goldfinch onboards retail via robo-advisors, and Clearpool or Centrifuge bridges RWA collateral back into Aave-style money markets. If those rails prove robust through a default cycle, fully permissionless experiments like Wildcat could finally have the on-chain identity primitives they need to thrive, unlocking the capital-efficient, global credit layer DeFi has chased since its inception.
Next Steps
The past five years have shown that permissionless plumbing + permissioned underwriting beats either extreme. Maple’s 16-fold TVL expansion in 2024 proves that institutional lenders will move on-chain - if they can rely on professional delegates, real-time collateral monitoring ,and fast liquidation rails.
The lesson: replicate the parts of TradFi that work (credit committees, senior-junior tranching) and outsource the rest to immutable code.
Goldfinch Prime packages senior-secured loans from Ares, Apollo, and Golub Capital into a single ERC-4626 vault, providing on-chain capital holders with exposure to thousands of SME loans that they could not source themselves. Expect Maple’s Syrup.fi, Clearpool Prime V2, and TrueFi’s forthcoming tfBill integrations to copy-paste that structure: wrap a yield-bearing senior token that can plug straight into Aave, Morpho, or Pendle for rehypothecation.
Spark’s $25m anchor position in Maple pools (with an option to double) hints at a broader trend: DAO treasuries redeploy idle stablecoin float into under-collateralized credit, then borrow against the resulting receipt to keep liquidity on hand. The loop converts dead capital into stacked yield without leaving the cryptographic trust boundary - the first institutional-grade example of money-legos in credit.
Wildcat’s free-banking experiment - let borrowers publish any mix of reserves, attestations or social proofs and let lenders decide - shines a spotlight on the missing piece: portable, privacy-preserving reputation. Ongoing pilots with Ethereum Attestation Service, zero-knowledge income proofs, and Chainlink Proof-of-Reserve feeds should be prioritized by every credit DAO; without credible private disclosure, under-collateralized lending will remain capped at KYC-gated pools.
Nearly every protocol now stakes delegate capital to first-loss tranches, but coverage remains shallow relative to loan books. Two levers can harden the stack: (a) external underwriters such as Neptune Mutual issuing CDS-style covers on Maple loans, and (b) EigenLayer restaking modules that slash validator yield to recapitalise pools in tail events. DAOs should treat these as non-negotiable expenditures, not optional nice-to-haves.
TrueFi’s 2025 roadmap divorces day-to-day underwriting from token-holder voting; instead, fixed mandates and time-boxed KPIs guide pool managers, while TRU holders review only escalation events. Copying this model reduces decision latency and dampens political risk - a prerequisite if regulated institutions are to trust DAO courts.
Europe’s MiCA regime already recognizes tokenized funds and loan notes, while the U.S. is tilting friendlier after the SEC signaled a retreat from punitive custody guidance. Protocols should localize legal wrappers in MiCA jurisdictions now, then utilize passporting to reissue yield tokens on global L1 and L2 networks once U.S. clarity is established. KYC-compliant sub-pools (Maple Insto, Goldfinch Prime) are the low-friction wedge product.
Centrifuge and Clearpool are already tokenising trade receivables and T-bills; next comes equipment leases, carbon credits, and even power-purchase agreements. Lenders get uncorrelated cash flows, borrowers get cheaper capital via crypto’s robo-infrastructure. Every credit DAO should earmark bandwidth for oracle partnerships and legal enforceability audits now, before the asset class floods in.
Default management is still painfully manual. Smart-contract escrow of collateral (or revenue share) plus automated payment routing can shrink resolution timelines from months to minutes. Pool delegates should push for programmatic step-downs: if the coverage ratio is less than 100%, a predefined settlement contract seizes escrowed cash flows, sells collateral, or triggers CDS payouts without boardroom drama. Borrowers shouldn’t need to use a wallet like MetaMask, and lenders should not need to read Solidity repos.
Maple already white-labels its APIs to exchanges and OTC desks; Goldfinch is shipping robo-advisor front-ends; TrueFi is overhauling UI/UX ahead of the Polaris stablecoin launch. The winner will be the stack that on-boards a CFO in three clicks, not the one with the cleverest yield engine.
Going forward, protocols must publish loss-adjusted APY, default frequency, recovery lag, and utilisation-weighted liquidity. Maple’s 61 margin-calls in 2024 were cured in three hours on average - a statistic more persuasive to institutions than the industry standard APR banners.
Achieving that curve requires (a) at least one transparent default cycle that proves automated workouts, (b) two or more regulated broker-dealers to warehouse senior tranches, and (c) settlement of U.S. custody rules. Developers should therefore prioritize audit-grade liquidation logic and stable UX. Investors should focus on protocols with skin-in-the-game delegates and external insurance. Policymakers should monitor the robust data trails these systems emit and craft disclosure-first, not permission-first, regimes.
The bottom line is that under-collateralized lending is no longer a moon-shot idea; it is a small but rapidly compounding asset class with credible product-market fit. The next steps are mostly operational, including identity, insurance, governance hygiene, and data standards. The protocols that execute these details will capture the bulk of the coming $40 billion on-chain credit boom.
