How to Evaluate RWA Yield: Five Questions to Ask Before You Allocate

Summary

• Income from a payer you can name. RWA yield is income paid by a real-world asset, such as T-bill coupons, borrower interest, reinsurance premiums or rent.
• A benchmark plus a premium. Read any RWA yield as a benchmark such as SOFR or T-bills, plus a premium for credit, liquidity and structural risk.
• Verification decides. Judge each yield by whether you can verify the asset, your legal claim on it and your route out.

Introduction

RWA yield sits on dashboards beside every other onchain rate, often with no word on where the income comes from. A rate funded by newly issued reward tokens can look identical to one funded by borrower interest. Only the second is real yield from a real-world asset (RWA), and the tokenized RWA market now runs to tens of billions of dollars.

This guide gives you five questions to ask before you allocate. They cover the cash source, the benchmark and premium, the asset, your legal claim and your exit. USD.AI sees the GPU-backed loans behind sUSDai from the deals it reviews. We use that credit-side view to show what each question looks like in practice.

What RWA Yield Is and How Large the Market Has Become

RWA yield is the income a real-world asset produces, such as Treasury interest, loan interest, insurance premiums or rent, passed to holders of a token. The token represents a claim on that asset or its cash flows, recorded onchain while the asset stays offchain. How reliable the income is depends on who pays it and how firmly the token connects you to it.

A real-world asset is any asset that exists outside a blockchain, such as a government bill, a loan or a building. A tokenized asset is a digital token recorded onchain that represents a claim on one of those assets. It usually sits in a fund or a special-purpose vehicle (SPV), a separate legal entity that ring-fences the assets from their originator.

Market-size figures vary because trackers count different things. The data provider rwa.xyz separates distributed and represented assets: distributed assets can be moved to wallets outside the issuing platform, and represented assets can't. On November 11, 2026, rwa.xyz showed about $40.01 billion distributed and $402.73 billion represented, both excluding stablecoins.

The IMF put public tokenized RWAs, excluding repos, at about $65 billion as of July 31, 2026. Its figure leaves out stablecoins and private markets and values each token at supply times issuer-reported unit value. Quote each measure with its own definition.

Forecasts run larger. Citi forecasts tokenized assets could reach $5.5 trillion by 2030 in its base case, with stablecoins projected separately. Its bear and bull cases are $2.7 trillion and $8.2 trillion. Yield-bearing stablecoins are a related category, and this guide leaves them aside.

Question 1: Where Does the Cash Come From?

Every RWA yield traces back to someone paying cash: a government, a borrower, an insurance buyer or a tenant. The payer sets the risk, so start by naming the payer and the asset that obliges it to pay.

Tokenized Treasuries: the government pays

Tokenized Treasuries earn the coupon or discount income on short-dated US government debt, held through a fund that issues the token. rwa.xyz put tokenized Treasuries at $14.04 billion distributed on November 10, 2026, down 5.58% over 30 days.

The mechanics differ by token. Ondo Finance states that OUSG holders accrue daily interest. Circle states that USYC accrues through its token price, with no staking or claiming required.

S&P assigned the Franklin OnChain U.S. Government Money Fund an "AAAm" principal stability fund rating on June 24, 2026. The rating rests on the creditworthiness of the fund's investments and counterparties. S&P reported total net assets of $813.5 million at May 31, 2026.

Tokenized Treasuries are the benchmark the rest of RWA yield is measured against. Every other class should explain what it pays above them and why.

Tokenized and onchain credit: the borrower pays

In credit RWAs, borrower interest is the cash, and credit quality and collateral set the risk. rwa.xyz groups this class under Tokenized Credit, spanning private credit, onchain lending, corporate, structured and specialty credit. On rwa.xyz's figures, Figure's tokenized home equity lines of credit (HELOCs) make up most of the category's represented value in November 2026.

A headline credit total therefore largely reflects one issuer's lending. rwa.xyz's March 17, 2026 classification framework added Corporate Credit, Asset-Backed Credit, Diversified Credit and Specialty Finance.

Credit RWAs differ most in what secures the loan. Maple Institutional makes overcollateralized loans to institutions, backed by liquid digital assets in qualified custody. Each Centrifuge pool is tied to an SPV, which keeps the pool's assets apart from the originator.

Reinsurance and real estate: premiums and rent

Reinsurance RWAs earn premiums for taking on insurance risk. Documentation from OnRe states that it earns premiums for assumed insurance risk, plus yield on its collateral. It also states that its token, ONyc, is priced at net asset value (NAV) with no peg, so its value moves with claims, and that redemptions run in set windows. Reinsurance income generally depends on claims experience, so a heavy claims year can bring losses.

Tokenized real estate pays rent minus property costs such as maintenance, taxes and management. Vacancy, repairs and property values all move that income.

GPU-backed loans: an emerging credit class

In GPU-backed credit, AI infrastructure operators pay interest on loans secured by their servers. The loans are repaid from offtake contracts, the agreements under which customers rent that compute. None of the trackers above has a GPU-backed credit category yet, so treat this as an emerging class.

For loans originated through USD.AI, an SPV holds the servers, the offtake contract and the revenue accounts. Capital is held in escrow until installation is independently verified, and each loan amortizes monthly.

The USD.AI loan history lists 20 publicly documented GPU-backed financings as of November 2, 2026. Those financings total $531.4 million committed and $296.8 million funded. USD.AI publishes the underwriting and risk standards every deal is assessed against.

Question 2: What Benchmark Is It Beating, and by How Much?

Any RWA yield can be read as a benchmark rate plus a premium, and the premium pays for a specific risk. The New York Fed published the Secured Overnight Financing Rate (SOFR) at 3.89% for November 6, 2026. At TreasuryDirect, the 4-week bill issued November 10, 2026 had a high rate of 3.890%. The 13-week and 26-week bills issued November 12, 2026 had high rates of 4.060% and 4.145%.

These rates change weekly, so use them as reference points. Whatever sits above them is the premium, which usually pays for a mix of four risks:

• Credit risk. The chance the payer defaults, and how much the collateral recovers if it does.
• Liquidity risk. Redemption windows, queues and thin secondary markets that can delay your exit.
• Structural and legal risk. Whether the SPV, fund or note gives token holders an enforceable claim.
• Smart contract risk. Bugs or failures in the code that issues, prices or redeems the token.

A yield close to the T-bill rate should behave like Treasuries. A yield well above it should have an identifiable risk behind it, because a risk you can't name is a risk you can't price.

As an illustrative example with no real product behind it, take a hypothetical credit token quoting the 13-week T-bill rate plus X percentage points. Ask what each slice of X pays for: borrower default, a redemption window, the legal structure or the code. The token's documents should let you trace each slice to a named risk.

Questions 3 to 5: Can You Verify the Asset, the Claim and the Exit?

The IMF flags risks for tokenized assets including legal enforceability, liquidity, operational and oracle risk. Each maps to a question you can check before you allocate. An oracle is the data feed that brings offchain information, such as a fund's NAV, onchain. In November 2025, IOSCO found that tokenized assets carry the risks of the underlying asset, plus technology risks.

Question 3: Can you verify the asset exists and performs?

Operational and oracle risk come down to who reports on the asset. If the issuer is the only source, you are trusting its records. Look for onchain reporting of loans and collateral, and for independent attestations that state what was checked and when.

USD.AI's Proof of Loans attestation is one example. A licensed independent CPA firm performed agreed-upon procedures and published the report on September 1, 2026, using a July 15, 2026 snapshot. The firm compared onchain disbursements with loan documents and recalculated each loan's LTV, the loan amount against the value of the GPUs.

In an agreed-upon procedures engagement, the firm performs specific tests agreed in advance and reports what it found. It gives no audit opinion, so read the findings for exactly the scope they cover.

Question 4: What legal claim do you actually hold?

A token can represent fund shares, a note or a pool share, and none of these automatically gives a direct claim on the asset. Legal enforceability is the risk that your claim fails when tested, usually in a default. Three checks cover most of it:

• The holding entity. Which SPV or fund holds the asset, and whether it is ring-fenced from the issuer's creditors.
• The security interest. Who holds the claim on the collateral, and whether that claim ranks first.
• The default path. What happens if the payer stops paying, and who can enforce.

Holding a share of a pool isn't ownership of any one loan, and your position moves with the pool as a whole. For financing through USD.AI, lenders hold a first-priority claim on the GPUs, contracts and revenue. If a loan defaults, the lenders can step in and enforce against the collateral.

Question 5: How do you get out, and how fast?

A token can be more liquid than the asset beneath it, and in our view that mismatch is the most underrated risk in RWA yield. Short-dated Treasuries mature within months, while loans repay over years and reinsurance contracts run until claims settle.

Issuers bridge the gap with redemption windows, as OnRe does, with queues or with secondary-market trading, whose depth varies by venue. Credit pools pay exits from loan repayments and reserves, so your exit depends on the cash coming in. sUSDai withdrawals are processed in order, and the page on how sUSDai works covers the rest.

Conclusion

The five questions apply to any RWA yield: the cash source, the benchmark and premium, the asset, the claim and the exit. A yield that answers all five gives you something to price.

sUSDai is one worked example. It is received by staking USDai, a stablecoin backed 1:1 by PYUSD with no yield on its own, in Earn. Its value reflects GPU loan interest, income on capital committed to loans and held in escrow, and Treasury yield on PYUSD reserves.

sUSDai deploys at most 90% of its assets into new financings, a maximum utilization set by CHIP governance. sUSDai isn't a stablecoin, and its value can rise or fall; yield isn't guaranteed and can be lower, zero or negative. Holding sUSDai isn't ownership of any one GPU or loan.

Put it through the same five questions as any other RWA yield, starting with the USD.AI documentation. To see what backs a position, visit Earn.

FAQ

What does RWA stand for?

RWA stands for real-world asset, an asset that exists offchain, such as a Treasury bill, a loan, real estate or an insurance contract. A token recorded onchain represents a claim on that asset or its cash flows.

What are tokenized assets?

Tokenized assets are onchain tokens that represent a claim on an offchain asset or its cash flows, usually through a fund or SPV. rwa.xyz splits them into distributed assets, which can move to outside wallets, and represented assets, which can't.

What RWA yield is there beyond tokenized Treasuries?

Beyond tokenized Treasuries, RWA yield comes from tokenized and onchain credit, reinsurance premiums, real estate rent and emerging GPU-backed loans. Each class carries its own mix of credit, liquidity, structural and smart contract risk, and any premium it pays should reflect that mix.

Is RWA yield correlated with interest rates?

It depends on the class: Treasury-based yield moves with short-term rates, and floating-rate credit tends to move with SOFR plus its spread. Reinsurance income depends mainly on premiums and claims, though its collateral yield still moves with rates.

Editor Notes

Held back by brand rules or reviewer instruction

• sUSDai APY figures. Omitted under the Vocabulary rule (no yield or APY figures) and the Blog posts rule (no view on returns on the digital-assets track).
• sUSDai premium over T-bills (Lighthouse). Omitted per reviewer instruction; no sentence implies an sUSDai spread over SOFR or T-bills.
• Per-loan rates, borrower names and largest loan. Omitted under the Two tracks rule (no rate figures on the digital-assets track) and the Borrower names rule.
• Deposit, supply and cumulative yield figures. Omitted per reviewer instruction.
• McKinsey and Deloitte capex figures; BCG figures. Omitted per reviewer instruction.
• USYC AUM. Omitted per reviewer instruction because of an unexplained discrepancy between sources.
• Attestation firm name. Omitted under the Service providers rule; the article says "a licensed independent CPA firm".
• Residual-value warranty. Omitted per reviewer instruction (under the Compliance vocabulary rule it isn't on every loan).
• Withdrawal cycle length, utilization mechanics and operator loan terms (80% loan-to-value, 20% equity, debt-service reserve). Omitted under the brief's content angle; sUSDai withdrawals get one sentence and a link to the sUSDai page.
• How USDai is obtained. OPEN, pending Samuel; the article omits it.

Sibling posts not yet live (add links once published)

• Yield bearing stablecoin. Link from the yield-bearing stablecoins sentence in "What RWA yield is and how large the market has become".
• Compute as an asset class. Link from "GPU-backed loans: an emerging credit class".
• GPU private credit. Link from "GPU-backed loans: an emerging credit class".
• Invest in gpu loans. No placement assigned in the outline; place once live.

Publish-day checks

• Re-read live figures. Re-read all rwa.xyz values (distributed and represented totals, tokenized Treasuries total and fund AUM, Figure's share of Tokenized Credit), SOFR and TreasuryDirect rates at publish, and update dates to match.
• Loan history link. The gpuloans.com loan history sits on the GPU-financing track; it is used only as the source for the cited loan totals, as the outline approved. Confirm its "data as of" date at publish.
• Blocked link checks. Confirm the S&P rating notice link (Treasuries section) and the Centrifuge docs link (credit section) resolve; both blocked automated checks.