Crypto credit's quiet rewrite: custody, collateral, and the new plumbing of on-chain lending

On 17 July 2026 at 13:01 UTC, Cointelegraph's research desk circulated a single-page framing that does the work of a longer essay: it separates "crypto credit" into four operating stacks, each resting on a different answer to one question, namely who actually holds the asset, and what happens when a borrower does not pay.

The note is short, but the taxonomy it proposes is the clearest one now in circulation: centralised-finance lending, decentralised-finance lending pools, tokenised US Treasuries, and private credit that has been wrapped into a token. The split matters because every major loss in this cycle has come from confusing one stack for another.

Four stacks, four custody stories

Centralised lending still means a company takes the deposit. The borrower signs a loan agreement governed by the law of an identifiable jurisdiction, and the lender holds the collateral in a wallet or with a custodian it controls. Enforcement runs through lawyers and bankruptcy courts. Decentralised lending works the opposite way: a smart contract pools liquidity from depositors, the borrower posts crypto collateral on-chain, and liquidation is automated when price thresholds are breached. There is no counterparty in the legal sense. There is only code, a price oracle, and the assumption that liquidators will execute before the market gaps.

Tokenised Treasuries sit between the two. The underlying asset is a US Treasury bill or note held by a regulated custodian; a token issued on a public ledger represents a claim against that custodian. The yield comes from the bond, not from the token, and the credit exposure is to the US government and to the issuer's legal structure, not to the chain. Private credit that has been tokenised is the youngest stack. A fund originates a loan to a corporate borrower off-chain, and investors receive tokens that represent pro-rata shares of that loan. The yield is the borrower's interest rate minus the fund's fees, and the credit exposure is to a single company, not to a sovereign.

Why the lines blur, and why that matters

Markets compress what accounting separates. A desk that sells yield to clients will describe all four products in similar language: duration, APY, net asset value. A borrower with a balance sheet to manage will treat them as interchangeable sources of dollar funding. The Cointelegraph note is explicit that the four stacks rely on different forms of custody, collateral and enforceability; that is a polite way of saying the risks compound when they are mixed.

The clearest example sits in the treasury-management function at a crypto-native company. Idle stablecoins can be parked in a DeFi pool for a variable APY, swapped into a tokenised T-bill for a yield closer to the SOFR rate, or routed into an off-chain private-credit fund for a multiple of both. Each choice shifts the credit exposure: to a smart contract and an oracle, to a custodian and a US Treasury, or to a fund manager and a private borrower. The legal remedies differ on each leg. A liquidation event in DeFi is a transaction; a default on a tokenised loan is a workout.

What the cycle's losses actually revealed

The cleanest lessons from the past two years sit in the failures the market already paid for. Where CeFi lenders collapsed, the binding constraint was the lender's own balance sheet and the speed at which withdrawals could be processed against it. Where DeFi protocols were drained, the binding constraint was the oracle, the liquidation engine, or the design of the collateral itself. Where tokenised treasury products wobbled, the binding constraint was the issuer's legal structure and the willingness of a custodian to honour redemptions in stressed conditions. The losses did not come from one architecture being superior; they came from users, and from desks, treating the products as substitutes.

This is the deeper point the Cointelegraph research framing makes when it insists on the language of custody, collateral and enforceability. Those three words are not synonyms. Custody decides who can move the asset. Collateral decides what can be seized, and at what speed, when a position goes wrong. Enforceability decides which court, in which country, will hear the dispute if the protocol's automated machinery runs out of road.

The structural shift underneath the labels

Read together, the four stacks describe a credit system that is no longer waiting for permission. A loan that, in 2019, would have required a bank, a syndicated facility and a wire transfer can in 2026 be originated by a fund manager in one jurisdiction, custodied in a second, and subscribed to by retail wallets in a dozen more, with the loan itself travelling as a token on a public chain. The plumbing is new. The credit is the same.

That is the reason the taxonomy matters more now than it did a year ago. The aggregate size of tokenised private credit and tokenised Treasuries is small relative to the on-chain stablecoin float, but the direction of travel is set. The next cycle of losses, when it comes, will not be a single catastrophic protocol drain; it will be a stack-confusion event in which a fund treats a tokenised loan like a money-market claim, or a treasury desk treats a DeFi pool like a custodian.

What remains unsettled is whether regulators will write the distinctions into binding rules before the market does. The Cointelegraph note does not claim that they have. It does the more useful job: it names the variables, so that when the next failure lands, the post-mortem starts in the right place.

Desk note: Monexus treats the four-stack framework as a structural map, not as a forecast. Wire coverage continues to mix yield, risk and product type; the editorial task is to keep them apart.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://t.me/cointelegraph
• https://t.me/cointelegraph/