Where onchain yield actually comes from

Onchain yield comes from five places: interest paid by borrowers in lending markets, coupons passed through by tokenized Treasury bill funds, funding and basis captured between spot and derivative prices, fees paid by traders to liquidity providers, and tokens issued by a protocol as an incentive. The first four are payments from a counterparty. The fifth is dilution.

Key takeaways

• A US payment stablecoin cannot pay you anything. The GENIUS Act prohibits a permitted payment stablecoin issuer from paying holders any form of interest or yield in connection with holding the token. Every stablecoin return is therefore produced by something the holder did with the token, not by the token.
• Four of the five sources have an identifiable payer. Ask who is worse off when you are paid. If nobody is, the return is emissions and it stops when the budget stops.
• Two sources paying the same rate are not substitutes. They fail in different regimes, and a portfolio holding both is only diversified if you checked that the regimes differ.

Just the basics

Yield is somebody paying you. Borrowers in a lending market pay interest and part of it reaches suppliers; a tokenized Treasury fund receives the coupon the US government pays on the underlying bills and passes it through after fees; in a basis or funding trade, traders on one side of a perpetual futures contract pay traders on the other side; and in liquidity provision the fee comes from people trading against your inventory. An emissions programme is the exception, because the protocol prints its own token and hands it out, so nothing is being paid by a counterparty and nothing continues once the allocation runs out. Everything advertised as an onchain rate reduces to some mix of those five. Figuring out the mix is the work, because the mix is what tells you when the rate goes away.

What's in this article?

1. Where does stablecoin yield come from?
2. Why does the source matter more than the rate?
3. What are the actual generators of onchain yield?
4. The regime that kills each source
5. How do you decompose a reported APY?
6. How do you identify the emissions component?
7. Stress-testing a source before allocating
8. How does Railnet fit?
9. What should you do next?

Where does stablecoin yield come from?

From a counterparty who has a reason to pay, or from token issuance. There is no third category, and the distinction does most of the analytical work.

The stablecoin itself does nothing. Since the GENIUS Act was enacted on 18 July 2025, a permitted payment stablecoin issuer may not pay the holder "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." The issuer earns on the reserves and the holder does not, a statutory split rather than a market outcome, and it is the reason the phrase "stablecoin yield" is slightly wrong as written: what people mean by it is the return on something they did with a stablecoin.

Once the balance moves, four counterparties become available. A borrower who wants leverage or working capital pays interest. Inside a tokenized money market fund, the coupon on the bills comes from the US Treasury. Funding flows from the trader who wants leveraged long exposure to whoever takes the other side, and a swap fee flows from the trader who wants to move out of one asset and into another to whoever supplied the inventory, so in each case the money starts with someone who received something they wanted.

Emissions are the fifth category and behave differently. The protocol mints its own token and distributes it to depositors, and at the moment of payment nobody is worse off in cash terms. That is exactly the problem: the cost falls on existing token holders through dilution, and on the protocol's balance sheet when the allocation is exhausted. Emissions are a customer acquisition budget expressed as a rate, and taking one deliberately can be rational. It is not a rate you can underwrite for a year.

The five categories are the source of the payment, not the wrapper it arrives in. A vault, a lending market and a token can all deliver the same underlying payer, so the wrapper is a distribution choice rather than a distinct kind of yield.

Why does the source matter more than the rate?

Rate tells you what you are paid. Source tells you when you stop being paid, and only the second one is a risk input.

Two positions quoting the same number can behave in opposite directions in the same week: a lending market supply rate rises when borrower demand rises and when utilization tightens, a funding capture strategy earns when leveraged longs crowd in and inverts when they leave, and a tokenized Treasury fund's coupon tracks front-end policy rates while being almost entirely insensitive to both. Hold all three and there are three payers with three different reasons to stop. Three lending markets are one payer with one reason, distributed across three interfaces.

This is the part most rate tables cannot express, because a table sorted by APY is sorted by the least informative column. An allocator comparing sources needs the payer, the dependency and the failure regime alongside the number, and needs to know that the number itself may be a blend of two sources with different lifespans.

Diversification onchain therefore means diversification across payers rather than across protocols. Two positions on two protocols that both depend on leveraged borrower demand are one position. See real yield for the term the market uses, imprecisely, to describe the first four categories.

What are the actual generators of onchain yield?

There are five, each with a different payer and a different thing that has to stay true.

Lending market interest

Suppliers deposit an asset, borrowers post collateral and draw it, and borrowers pay a rate set by a published formula rather than negotiated. Aave's model uses two slopes with an optimal usage ratio as the inflection point, and derives the supply rate from the variable borrow rate after subtracting the reserve factor that goes to the protocol treasury. Utilization is the variable that matters: the share of supplied assets currently borrowed.

Because the supply rate is a function of utilization, a rising supply rate is a report that withdrawal capacity is falling. Above the optimal ratio the second slope is steep by design, to pay new suppliers to arrive and to pressure borrowers to repay. A supplier who chases that rate is being paid for exit difficulty at the exact moment exit is getting difficult. See how DeFi lending works and what is an interest rate model.

Tokenized Treasury bill coupons

A fund holds short-dated US government paper and issues a token representing a share. The return is the coupon on the underlying, less the manager's fee, passed to holders either through a rising share price or through a rebasing balance. Payment originates with the US Treasury, and what can interrupt it is front-end policy rates on one side and the issuer's operational integrity on the other.

These instruments carry the access and settlement conditions of the offchain fund underneath. Ondo documents OUSG as restricted to accredited investors who are qualified purchasers and who have completed onboarding, with instant mint and redemption from a $5,000 minimum and a separate non-instant path carrying $100,000 investment and $50,000 redemption minimums. The yield is the least complicated part of the instrument; the gating and the redemption path are what create the operational work. See what are tokenized treasuries.

Basis and funding capture

A perpetual futures contract has no expiry, so venues use a periodic payment to keep its price near spot. That payment, the funding rate, moves directly between holders of long and short positions rather than being charged by the venue, and it is typically built from an interest rate component plus a premium measuring the gap between the contract price and an index price. Intervals and default parameters differ by venue, so read the specification of the venue a strategy actually trades on rather than assuming a market convention.

A delta-neutral basis strategy holds spot and shorts the perpetual, so price movement cancels and the position collects funding when it is positive. The payer is leveraged longs. What the position depends on is directional demand for leverage, a market sentiment variable rather than an interest rate. Funding is the one source of the five where the payment itself can reverse and charge you for holding the position.

Liquidity provision

An automated market maker holds inventory in two assets and quotes continuously against it, and traders pay a fee per swap that accrues to the depositors who supplied the inventory. The payer is whoever wanted to trade.

The complication is that fee income is gross, not net. When the pool's assets diverge in price, arbitrageurs rebalance the pool at prices worse than the market, and the resulting shortfall against simply holding the two assets is a real cost that the advertised fee APR does not net out. A stablecoin-to-stablecoin pool minimises it because the two assets are supposed to track each other, and for the same reason its fee income is thin unless one of the pair is under stress. Volume is the dependency, and volume is not stable.

Protocol emissions

The protocol distributes its own token to depositors on a schedule set by governance. There is no external payer. The rate is a function of the emission schedule and the token's market price, so it moves when either moves and it ends when the allocation ends.

Points programmes are the same mechanism with the payment deferred and the terms unstated. A points balance is an expectation of a future distribution at a conversion rate that has not been published. Treat it as a return of zero with option value, and if the allocation only clears the hurdle once points are valued at a number you chose, the allocation does not clear the hurdle.

The regime that kills each source

Each generator has one dominant failure regime, and the reason to hold several is that the regimes differ.

Two entries in the last column deserve attention. Funding capture and liquidity provision can both produce a negative net return while the interface continues to display a positive gross rate, because the cost sits in a different accounting line from the income. That is not deception. It is the difference between a fee APR and a total return, and it is the single most common reason a reported onchain yield does not arrive in the accounts.

How do you decompose a reported APY?

Split the headline into components until every component has a named payer, then treat the components without one separately. The method takes about an hour per source and does not require any data the venue does not publish.

1. Write down what the number measures. APR or APY, gross or net of fees, trailing or forward-looking, and over what window. A seven-day trailing APY on a source with lumpy income is a different claim from a thirty-day one, and a compounded figure is a different claim again from a simple one.
2. Find the fee stack and subtract it. ERC-4626 specifies that totalAssets must be inclusive of any fees charged against assets in the vault, so a compliant vault's share price is already net of fees taken from assets. That is not a guarantee that a displayed APY figure is. Check whether the number on the page is derived from share price or from an underlying rate before the fee.
3. Identify every distinct payer. For a vault allocating across several markets, that means listing each market and its own payer. One line item per payer.
4. Separate the native-asset component from the incentive-token component. If the position pays USDC and also distributes a governance token, those are two returns with two lifespans, and only the first is denominated in the thing you deposited.
5. Value the incentive component at what it is, not at what it prints. Apply the token's actual liquidity and any lockup or vesting to the notional. An emission worth its face value only if sold at size into a thin market is worth less than its face value.
6. Compute the base rate. Headline, less fees, less the incentive component, less any cost that sits outside the income line such as rebalancing losses in a liquidity position. What remains is what a counterparty is actually paying you.
7. Restate the position as base rate plus incentive, with two separate durations. The base rate carries a regime dependency. The incentive carries an end date, whether or not the end date has been announced.

The output is a sentence rather than a number. "This position pays a base rate from borrower demand in one market, plus an incentive component of unstated duration, and the base rate is the part that survives a deleveraging." That sentence, rather than the headline APY, is what belongs in an investment memo.

Running the same decomposition continuously across a live portfolio, rather than once per underwriting, is a data problem rather than an analytical one, and it is where most allocators stop. To discuss attribution against a portfolio you already hold, talk to the Railnet team.

How do you identify the emissions component?

Four checks, in order. Any one of them can resolve it, and the first is usually enough.

• Check the denomination. If part of the return arrives in a token that is not the asset you deposited and is not a claim on an external cash flow, that part is emissions. This settles most cases in a minute.
• Check the governance record. Emissions are approved. There is a proposal, an allocation, a schedule and usually an end date. If a forum post authorised the rate, the rate is a budget line.
• Ask who is worse off. Trace the payment to a party who received something in exchange. A borrower got liquidity. A trader got execution. If no such party exists, the payment is dilution.
• Check the sensitivity. A base rate moves with utilization, policy rates or funding. An emissions rate moves with the incentive token's price and with nothing else. Watch the quoted rate against both and the correlation is usually obvious within a few weeks.

The purpose is not to avoid emissions. A short allocation into a well-collateralized market during an incentive window can be a sound decision, taken deliberately, with an exit date written down at entry. The failure is booking an emissions rate into a forecast, or comparing an emissions-inclusive number against a base rate elsewhere and concluding the first venue is better. That comparison is between a return and a marketing budget.

Vaults add a layer here because the person setting the allocation is not the depositor. Morpho's vault role model makes the split explicit: the curator configures adapters, caps, fees and interest rate limits with most actions timelocked, the allocator moves assets between enabled adapters, and a sentinel can reactively deallocate, cut caps or revoke pending timelocked actions, with performance fees capped at 50% and management fees at 5%. A depositor reading a vault APY is reading the output of decisions made by those roles, and the attribution work has to be done at the level of the underlying markets rather than the vault. See what a curator controls and what they do not.

Stress-testing a source before allocating

Run the position through five checks: the source stops paying, it becomes hard to exit, its dependency inverts, its own holdings sit a layer deeper than you priced, and the number you are quoted comes from a manipulable method. All five need written answers before the first deployment.

1. Set the base rate to zero and keep the position. If the allocation only makes sense at the current rate, it is a trade with a stop, not an allocation, and it needs an exit rule rather than a mandate line.
2. Model an exit at the worst published path, not the fast one. For a lending market, exit when utilization is above the optimal ratio and the second slope is active. For a tokenized fund, exit through the non-instant route with its stated minimums and cutoffs. For a funding trade, exit when the leg you need to close is the crowded one.
3. Invert the dependency. Funding goes negative, utilization collapses, or the stablecoin pair in a liquidity position stops tracking. Each of these has happened repeatedly and none requires a contract failure to occur.
4. Trace the dependency chain one level down. A tokenized fund can hold other tokenized funds. As of 24 July 2026 Ondo published OUSG's holdings as State Street's SWEEP, BlackRock's BUIDL, Franklin's BENJI, Fidelity's FYOXX, and a small cash balance. That is a disclosed structure and a reasonable one. It also means a concentration limit written against the instrument you bought is measuring the wrong thing, and has to be applied to what sits underneath.
5. Check what the price you are quoted is derived from. ERC-4626 warns that the standard's preview methods "are manipulable by altering the on-chain conditions and are not always safe to be used as price oracles." A valuation or a rate computed off those methods inherits that property.

A source that survives all five is not safe, only understood, and understood is the only property an allocator can actually verify in advance. See a risk framework for onchain allocation.

How does Railnet fit?

Railnet is the operating layer for onchain asset management, and attribution is one of the things it produces. Positions across several venues are executed through one standard and reported with the source of each return separated rather than blended into a single portfolio APY.

In shape it is vault infrastructure, the layer that holds the strategy between the venues it allocates into and whatever channel distributes it. Aave, Morpho, Compound, Ondo and tokenized treasury issuers are sources that connect to it. None of this is an argument for using one venue over another.

The specific problem it addresses is that attribution done once during underwriting goes stale immediately. Utilization moves, emission schedules end, a curator reallocates, a fund's holdings change. A manager running positions across ten sources has ten rate models and ten reporting formats, and blending them into one number is what most reporting tools do because it is the only thing a common denominator allows. Railnet's approach is a single state model across sources, so the composition of a return can be reported continuously and a change in composition is visible when it happens rather than at the next review.

There are places it does not help. Railnet does not tell you whether an emissions programme is worth taking. It does not price a token you cannot sell. And any attribution is limited by what the source publishes, so an opaque venue stays opaque no matter what sits above it. The judgment in that method is yours.

What should you do next?

• If you are underwriting a single source, run the decomposition in how do you decompose a reported APY and keep the resulting sentence in the memo.
• If you are building an allocation, spread it across payers rather than across protocols, with the failure-regime table above in hand.
• If you need a baseline to measure against, use the component of the rate that survives a deleveraging, not the headline.
• If this is treasury cash rather than fund capital, start at how treasuries deploy capital onchain.
• Source integrations and the reporting surface are documented at docs.railnet.org.

See all questions on onchain yield sources

• What is real yield?
• What is an interest rate model?
• How does DeFi lending work?
• What are tokenized treasuries?
• What are yield-bearing stablecoins?

FAQ

Where does stablecoin yield come from?

From borrowers paying interest in lending markets, from Treasury bill coupons inside tokenized money market funds, from funding paid between traders in perpetual futures, from swap fees paid to liquidity providers, and from tokens a protocol issues as an incentive. The stablecoin itself pays nothing, because a US payment stablecoin issuer is prohibited from paying holders interest or yield.

What is the difference between real yield and emissions?

Real yield has an external payer who received something in exchange: liquidity, execution, or a loan. Emissions have no external payer, and the cost falls on existing token holders through dilution and on the protocol when the allocation is exhausted. The test is to trace the payment to a counterparty who is worse off. If you cannot, it is emissions.

Why does a lending rate spike?

Because utilization has risen past the model's optimal ratio and the second slope has activated. Aave's model derives the supply rate from the variable borrow rate after the reserve factor, and the steep upper slope exists to attract supply and pressure repayment. A rate spike is a report that withdrawal capacity is tight.

Is a basis trade a safe source of yield?

It is a source with a specific dependency: demand for leveraged long exposure. Funding is transferred between long and short holders at set intervals, and it can turn negative, leaving the position paying to stay open. It also carries exchange counterparty and margin risk that lending market positions do not.

Why is my realised return lower than the advertised APY?

Usually one of three reasons. Fees are taken below the displayed rate. A portion of the headline was an incentive token valued at a price you could not achieve. Or the position has a cost outside the income line, such as rebalancing losses in a liquidity pool, that a fee APR does not net out.

Can I compare two vault APYs directly?

Not without decomposing both. Two vaults quoting the same number may hold different payers, different fee structures and different incentive components. Compare base rates after attribution, and compare the failure regimes separately, because two positions that fail in the same regime are one position held twice.

Do points count as yield?

Treat them as zero with option value. A points balance is an expectation of a distribution at a conversion rate that has not been published, on a date that has not been fixed. If an allocation only clears its hurdle once points are valued at a number you chose yourself, it does not clear the hurdle.

Talk to the team

If you are attributing returns across several venues, or trying to report the composition of a portfolio's yield rather than one blended figure, talk to the Railnet team. There is no form to fill in first.