On-chain finance for financial advisors: a field guide

What the category contains

On-chain finance includes crypto-native assets such as ETH, decentralized lending and trading protocols, stablecoins, tokenized fund shares, tokenized Treasuries, private-credit instruments, and the chains that settle them. These products share a rail but not a risk model.

An advisor should resist treating “digital asset” as one allocation. The more useful first question is: what economic exposure exists before the token wrapper is considered?

The institutional side is no longer small. On 2026-09-30 RWA.xyz counted $14.65 billion in tokenized U.S. Treasury funds, across 25 funds and 85,521 holders,¹ and $38.60 billion in tokenized real-world assets of every kind, beside $294.67 billion in stablecoins.² Size is not an endorsement. It is the reason these questions now reach ordinary advisory practice.

Four distinct exposures

What settlement on a public chain changes

Three properties change when an instrument settles on a public chain, and each cuts both ways.

Finality. A transfer settles in minutes and does not reverse. There is no T+1 and no failed-trade break, and also no chargeback, no recall, and no help desk that can unwind a mistaken or induced transaction. Operational error turns straight into loss, which is why the approval workflow belongs in the custody review.

The custody boundary. A conventional security exists on an intermediary’s books, and possession follows the account agreement. An on-chain asset obeys its key: whoever can sign controls the position. For tokenized securities a transfer agent stands above the key with allowlist and reissue powers, so “who can move this” has a different answer at every layer. The Atlas records those answers.

Composability. A token in a wallet can be posted as collateral, lent, or pooled in any compatible protocol the same hour, with no transfer-out request. That mobility is the real institutional draw. It is also the path contagion takes: when a widely accepted collateral token fails, the failure spreads into every protocol that accepted it.

DeFi versus tokenized securities

The same rail can carry software-native positions and conventional securities, and their diligence files should not look alike. DeFi describes financial functions performed by smart contracts. A tokenized security is a security whose ownership or transfer is recorded on a blockchain. A tokenized fund may interact with DeFi, and a DeFi receipt may wrap a tokenized asset, but the two words are not synonyms.

Three published assessments show how differently the two files read. BlackRock’s BUIDL is a tokenized security, so the diligence ran through the offering documents: a private fund offered to qualified purchasers under Rule 506(c) and the Investment Company Act §3(c)(7) exemption, a $5 million initial minimum, and transfers only between whitelisted wallets.³ The research is favorable with conditions, and the model client may not hold it; the code barely matters to the answer. Venus is a DeFi position, so the diligence ran through the deployed system’s record: a May 2021 collateral event that cost the protocol about $77 million by its own post-mortem,⁴ a shortfall of about $14.2 million when the LUNA price feed froze in May 2022,⁵ and a price-manipulation attack in March 2026. Rejected on the record, with no offering document to read. And Binance Staked ETH looks like DeFi while being neither: an exchange claim that moves on-chain. Rejected in a paragraph.

The stack can combine both, and then the files add up rather than replace each other. On 2026-09-25 Aave v4 on Base began taking tokenized stock certificates as collateral for USDC loans, the first lending market where the collateral is a security. A USDC supplier there depends on the certificate’s issuer and custodian, a price feed that runs only on weekdays, a reserve that pauses for splits and dividends, a liquidation that sells the certificate rather than the share, and an offer made only to non-US persons. The tokenized-assets report walks each link.

Write every position as one sentence: “The client owns [legal or economic claim], represented by [token], held through [wallet or custodian], deployed in [protocol], settling on [chain].” Any blank becomes a diligence task.

Where the return comes from

Every displayed percentage has a payer. Lending rates are paid by borrowers and float with how much of the pool is lent out. Staking yield is protocol issuance plus transaction fees, paid in the network’s own token. Tokenized cash yield is the underlying portfolio’s income, less fees. Reward yield is a subsidy paid in the protocol’s own token, and it declines on a schedule. Naming the payer is the diligence act. Comparing percentages across different payers is how a 5% credit position gets sized like a 5% Treasury position.

Stablecoin yield: who pays it, and what can break

A stable price target does not make an instrument cash, and a displayed rate does not say what risk earns it. Start with the law. The GENIUS Act, signed 2025-07-18 and in force by 2027-01-18 at the latest, bars a permitted payment stablecoin issuer from paying holders “any form of interest or yield (whether in cash, tokens, or other consideration) solely in connection with the holding, use, or retention of such payment stablecoin.”⁶ So yield on a dollar token comes from somewhere else: a lending market, a savings wrapper, a fund, or a trading strategy. Each is a separate instrument, with its own issuer, contracts, and exit, stacked on the stablecoin.

Diligence both parts. The stablecoin has an issuer, a reserve or collateral model, a redemption path, a freeze policy, and a depeg history. The venue adds borrower, contract, oracle, governance, bridge, curator, and liquidity risk. The chain and the custody are further layers, set out in the four-layer test.

In a lending market a high rate can be an exit warning. Rates climb steeply as a pool empties, and the condition that produces the attractive rate is the one that stops suppliers from withdrawing; the liquidity section of the due-diligence guide shows the mechanism.

The stablecoin itself can fail, and not only through fraud. On the night of 2023-03-10 Circle said it could not withdraw $3.3 billion of USDC reserves, about 8% of the total, from Silicon Valley Bank. Its minting and redemption ran on U.S. banking hours, so over the weekend redemptions all but stopped, holders who wanted out had to sell, and USDC traded as low as 86 cents.⁷ Dai, which let holders swap it one-for-one with USDC, lost its peg too, though it held no deposit at the bank.⁷ The peg returned after Treasury, the Federal Reserve, and the FDIC said on 2023-03-12 that all SVB depositors would be protected; by 2023-03-13 Circle reported the $3.3 billion fully available and moved to new banks.⁸ Full reserves did not stop a weekend at 86 cents. The redemption path decides the price, and it keeps business hours.

The yield venue fails in two classic ways. Maple Finance shows the first: credit. Its pools screened near 5%, and the premium paid for lending to trading firms without posted collateral. On 2022-12-05 one borrower, Orthogonal Trading, was sent a notice of default on eight loans totaling $36 million, about 30% of all active loans on Maple, after telling the pool’s manager for weeks that its FTX exposure was about $2.5 million.⁹¹⁰ Press reports put lenders in the worst-hit pool near an 80% loss. Borrower honesty is not visible on-chain, and no dashboard priced it. The rejection is published in full.

Usual’s USD0++ shows the second: the terms themselves. On 2025-01-09 Usual announced that it would replace the unconditional early one-for-one exit from USD0++ with a floor of 0.87 USD0, plus a one-for-one route that required a contribution of its USUAL token.¹¹ Its 2025-01-14 update acknowledged that exit liquidity had been heavily impaired and that pool liquidity had stayed below $20 million.¹² No exploit, no oracle failure. Governance can still set the early-exit floor. A redeemable token is worth its redemption terms, and the terms are a governance variable unless they are fixed in code. That memo is published too.

The minimum monitoring record for a stablecoin yield position:

• base yield and reward yield, shown separately;
• the period measured and the time of the reading;
• withdrawable liquidity and utilization;
• the issuer and who holds the freeze;
• venue, oracle, bridge, and curator dependencies;
• the depeg and redemption conditions that reopen the file.

The control layer decides more than the technology

Two dollar tokens on the same chain mark the poles. Circle’s USDC is administered: Circle can block addresses under its terms and legal obligations, and it has. Liquity’s LUSD is the opposite: immutable contracts, no admin key, no governance, no blocklist. Both are called stablecoins. They are different instruments in a compliance file. Neither finding is a verdict by itself; issuer control is exactly what a tokenized security requires. The point is that the control finding must be read from the contract and the documents, not assumed from the word “decentralized.” The Atlas publishes this layer for every asset and chain graded so far.

What tokenization changes, and two instruments the first question catches

Tokenization can change settlement speed, programmability, minimums, distribution, collateral mobility, and recordkeeping. It does not change the legal nature, credit quality, duration, fee load, or suitability of the underlying investment. A tokenized Treasury fund remains a fund. A tokenized private-credit vehicle remains private credit.

Binance Staked ETH, roughly $7.1 billion at the July 2026 review, is settled by the first question in about a minute. The underlying exposure is a claim on a centralized exchange. Binance holds the staked ETH; the token moves on-chain and so appears in dashboards beside non-custodial assets. The wrapper suggests self-custody and protocol-native risk; the claim delivers neither. Rejected.

USD.AI runs the other way. The name suggests a dollar; the yield token, sUSDai, is a share in a credit fund making non-recourse loans to AI infrastructure operators, secured on GPU hardware, with exits through 30-day redemption windows. A client who sizes it like cash has mistaken the wrapper for the exposure. Rejected, with the full reasoning published. Neither case needed a technical audit. Both needed the first question answered honestly.

The advisor’s sequence

1. Identify the underlying economic exposure.
2. Identify the legal claim and who may hold it.
3. Record who controls custody, transfer, and redemption.
4. Test the protocol, chain, and custody layers.
5. Decide what evidence and monitoring the recommendation needs.

For tokenized products, continue with tokenized assets for RIAs. For DeFi positions, the due-diligence guide takes each layer in turn.