NFT lending lets holders borrow crypto against NFTs that they lock as collateral. This guide compares the protocol models, the parts a lending protocol needs, the risk limits and the audits, and it shows how small the NFT lending market has become in 2026.
The main design choice is who prices each loan. In the peer-to-peer model, lenders price loans themselves, while in the pool model they deposit into a shared pool and an oracle prices the loans. This choice sets the protocol's risk, its liquidity and its liquidation process.
This page covers protocol design. The guide to NFT loans covers the product for borrowers and lenders: loan offers, the loan lifecycle and operations.
Why NFT lending is harder than token lending
In a protocol such as Aave, the collateral is a liquid token with a reliable price feed, and liquidators can sell it at once on large markets.
NFT collateral lacks these features. Each item is different, its price changes only when a similar item trades, and selling a seized NFT can take days. Every NFT lending design deals with collateral that is hard to price and hard to sell. Readers who know token lending can compare with Aave-style lending protocols.
NFT lending protocol models
| Model | How loans are priced | What happens on default | Protocols that used it |
|---|---|---|---|
| Peer-to-peer, fixed term | A lender makes an offer for one NFT or for any NFT in a collection. | The lender takes the NFT if the borrower misses the due date. | NFTfi, Arcade, Gondi |
| Peer-to-pool | An oracle's floor price times a loan-to-value ratio | A low health factor starts an auction. | BendDAO, early ParaSpace |
| Perpetual callable loans | Lenders make offers, and the market sets the rate. | The lender calls the loan, and a refinancing auction runs. The lender gets the NFT if nobody refinances. | Blend (Blur) |
| Vault-backed | The price of the vault token on a DEX | The protocol sells the vault token. | Lending against fractionalized collections |
Peer-to-peer lending
A peer-to-peer protocol needs no oracle, because a lender decides what a specific NFT is worth. Defaults are simple: after the due date, the lender can claim the collateral from escrow.
Matching is slower, because a borrower waits for an acceptable offer. Peer-to-peer protocols handled market crashes better than pool protocols, because they have no automatic chain of liquidations.
NFTfi was the best-known peer-to-peer protocol. It launched in 2020 and arranged about $737 million in loans. NFTfi stopped new loans in June 2026 and ended operations on 31 August 2026, citing the smaller NFT market, according to Blockchain Reporter.
Peer-to-pool lending
Lenders deposit ETH into a pool and earn interest, and borrowers borrow at once against collections that the protocol approved. The protocol needs a floor-price oracle and a liquidation process.
BendDAO showed the risk of this model in August 2022. Floor prices of large collections fell, and many loans became eligible for liquidation at the same time. The auctions found few bids above the debt, and depositors rushed to withdraw. BendDAO survived by changing its parameters during the crisis, and its case is the standard example of risk in the pool model.
Perpetual callable loans
Blur launched Blend in 2023. Blend loans have no fixed term and use no oracle, and a loan runs until the lender calls it. The call starts a Dutch auction, in which other lenders can refinance the loan. Lenders set prices through offers, as in peer-to-peer lending, and can still exit a loan by calling it.
Blur announced in October 2026 that Blend will stop matching new loans on 12 November 2026, when the blur.io website closes. Borrowers must repay or let their loans default by that date, according to Unchained.
NFT lending platforms in 2026
NFT lending shrank with NFT trading. The table shows the value locked in several NFT lending platforms on 19 October 2026, from DefiLlama's NFT lending category.
| Platform | Model | Chains | Value locked | Status |
|---|---|---|---|---|
| Blend (Blur Lending) | Perpetual callable | Ethereum | about $3.0 million | New loans stop on 12 November 2026 |
| Gondi | Peer-to-peer with refinancing | Ethereum, Hyperliquid L1, Robinhood Chain | about $3.0 million | Active |
| BendDAO (v1) | Peer-to-pool | Ethereum | about $0.5 million | Listed |
| Arcade | Peer-to-peer | Ethereum | about $0.2 million | DefiLlama marks the website as dead |
| Sharky | Peer-to-peer | Solana | about $0.14 million | Listed |
| Citrus | Peer-to-peer | Solana | about $0.08 million | Listed |
| NFTfi | Peer-to-peer | Ethereum | about $0.02 million | Operations ended 31 August 2026 |
Value locked counts the collateral and funds held in each protocol's contracts. Contracts stay on-chain after a website closes, so a closed platform can still show some value.
Components of an NFT lending protocol
- Escrow or lien contract. This contract holds the collateral or records a lien that blocks transfers. It must handle ERC-721 tokens, and often ERC-1155 tokens, safely, including the
onERC721Receivedcallback. - Loan terms engine. The engine stores the principal, the interest rate or fixed repayment, the duration and the origination fee, using integer math with explicit rounding.
- Oracle, for the pool model. A floor-price feed should use time-weighted averages and ignore wash trades and outliers, and it should pause when its data is stale. A floor price from one marketplace is easy to manipulate.
- Liquidation module. The module runs auctions of a set length and gives the borrower a grace period to repay. It also needs a plan for bad debt when no bid covers the loan.
- Collection allowlist and risk limits. Each collection gets a maximum loan-to-value ratio, a liquidation threshold and a cap on total loans, which governance changes through a timelock.
- Indexer and front end. The interface shows offers, health factors and auction status.
Risk limits for a pool
In a pool model, the chosen numbers matter more than the code. Keep the loan-to-value ratio well below the ratios that token lending uses. Set the cap on total loans for each collection from its real daily trading volume.
Test a case in which the floor price halves within a week and no buyers appear; if the pool cannot survive that case, the limits are wrong. The protocol should be able to remove a collection quickly, and the process for removal should be public.
Security and audits
NFT lending contracts hold valuable collateral and move through many states, so attackers target them. Common bugs include:
- re-entrancy through token receiver hooks
- wrong interest calculations
- liquidation logic that others can front-run
- replay of signed off-chain offers
Plan for at least one independent audit and a bug bounty. See smart contract audits.
When a new NFT lending platform makes sense
The figures above show a small market. NFTfi has closed, Blend stops new loans in November 2026, and the active platforms hold a few million dollars. A new protocol needs a specific reason to exist, such as an underserved chain, a specific asset class such as tokenized physical goods, or a link to a marketplace that already has users. The DeFi lending and borrowing guide covers pooled lending in general.
Frequently asked questions
What is NFT lending?
NFT lending is borrowing crypto against an NFT that the borrower locks as collateral. If the borrower does not repay, the lender or the protocol takes the NFT.
What happened to NFTfi?
NFTfi stopped new loans in June 2026 and ended operations on 31 August 2026. It had arranged about $737 million in loans since 2020, and its contracts remain on Ethereum.
Which NFT lending model is safest for lenders?
Peer-to-peer lending with careful offers carries less shared risk, because each lender sets terms and no shared pool depends on an oracle. A lender can still end up with an NFT worth less than the loan.
How do NFT lending platforms price collateral?
On peer-to-peer platforms, lenders decide the price. Pool platforms use oracles built from recent sales and floor listings on several marketplaces, usually averaged over time to resist manipulation.
How long does it take to build an NFT lending protocol?
A small senior team may need three to five months for a peer-to-peer protocol with a front end, plus audit time. This is an estimate, not a quote. Pool models with oracles and auctions take longer and need more risk work.