Stablecoin yield and the four engines behind it

Stablecoin yield is the easiest part of crypto to analyse honestly, because there is no token price to hide behind. A dollar-denominated asset paying a dollar-denominated rate can be set directly against the risk-free alternative. On 15 September 2026 the three-month US Treasury bill paid 3.97%, the effective fed funds rate was 3.63%, and a one-year bill paid 4.16%. Every number on this page should be read against that line.

Do that and the market sorts itself immediately. Mainstream on-chain stablecoin yield sits at or slightly below Treasuries. Centralised platforms advertise more, sometimes much more, and the difference is always explained by one of four things — and sometimes by none of them, because the platform does not say. What follows is each engine in turn, what it currently pays, and whether it can keep paying it.

The four engines, and whether they last

Notice what is missing from that table: there is no engine that produces a high, stable, low-risk dollar return. If a product offers one, it is running one of these four with leverage, or it is not telling you which one it is running. The lending engine is dissected in detail in our guide to crypto lending; the rest are below.

Sky, and the proof that governance-set rates move

Sky — formerly MakerDAO — runs the largest on-chain savings product tied to a stablecoin. Its sUSDS pays the Sky Savings Rate, which was 3.60% on 16 September 2026. The rate is funded from several billion dollars of tokenised treasury exposure, the spread on Spark's borrow rates, and legacy stability fees from collateralised debt positions, with Sky Lending holding $5.35 billion in TVL.

What makes it worth studying is not the level. It is the history. This is an administered rate, changed by governance vote, and it has moved by a factor of three and a half in three years.

We could not retrieve a complete audited series for the intermediate points, so the timeline above is deliberately labelled as selected data points rather than a full history. The conclusion does not depend on filling the gaps: an administered rate reflects a treasury's needs, not a market's clearing price, and it can be cut the week after you deposit.

Ethena, and what a basis trade actually is

USDe is not a fiat-backed stablecoin and Ethena does not claim it is. Backing assets are deployed into delta-neutral basis positions in crypto perpetuals and futures, delta-neutral positions in non-crypto markets, overcollateralised on-chain lending, overcollateralised loans to institutional counterparties, tokenised real-world assets, and other stablecoin yield. The backing sits in off-exchange custody with multiple regulated providers.

The core trade is simple to state. Hold spot, short the perpetual future against it, and the position is indifferent to price. When perpetual funding is positive — which it usually is in a bull market, because more people want to be long with leverage than short — the shorts get paid, and that payment is the yield. Only staked USDe receives it. Unstaked holders earn nothing and effectively subsidise the stakers.

That compression — better than tenfold in two and a half years — is the honest headline for Ethena, and it is also the mechanism working exactly as described. Funding is a price. It fell because leverage demand fell. The supply figures track the same story: above $14 billion at the October 2025 peak, roughly $8.5 billion within weeks of the 10 and 11 October crash, $5.92 billion in March 2026, and $4.686 billion when we checked.

Tokenised treasuries: the compliant engine

The third engine is the only one that does not depend on somebody else's leverage. Tokenised money market funds and treasury products hold actual short-dated government paper and pass the coupon through on-chain. The sector held $15.65 billion of distributed value on 16 September 2026, down 3.46% over thirty days, with an aggregate seven-day APY of 3.74%.

The structural point is the one almost nobody makes: 3.74% is below the 3.97% available from the underlying three-month bill, and the individual funds are lower still — Superstate's USTB showed a 3.59% thirty-day yield and Ondo's OUSG 3.44%, the latter partly because it is a fund-of-funds carrying a second layer of fees. The wrapper costs roughly 25 to 50 basis points. What you are buying with that spread is settlement at any hour and composability with other contracts, not extra return. The eligibility maze — USDY restricted to non-US persons, OUSG to accredited qualified purchasers with a $5,000 minimum, USTB to accredited and qualified US residents — is the other half of the story, and it has its own page intokenised treasuries.

What the centralised platforms pay, and what they want for it

CeFi rates sit above everything above, which is exactly what you would expect from an unsecured claim on a company. The question is always what is attached.

The pattern is consistent enough to be a rule. Where the yield source is disclosed, the rate is moderate and the conditions are few — Ledn is the clean example. Where the rate is high, the conditions multiply: a portfolio minimum, a native-token tier, a subscription, a balance cap, or a discretionary component the provider can withdraw. We take the wrappers apart product by product in crypto savings accounts.

Bitpanda deserves singling out for the opposite reason to Nexo. Its own campaign page states that the 7% is a 3% fixed base reward plus a bonus granted voluntarily at Bitpanda's sole discretion and revisable in 14-day cycles, that ownership of the assets passes to Bitpanda GmbH in Vienna for the term, and that the offering is an unregulated product with no deposit protection and exposure to counterparty, insolvency and de-pegging risk. Every one of those sentences makes the product look worse and the company look better; the detail is inour Bitpanda review. It also demonstrates something readers routinely get wrong — a MiCAR licence covers the regulated services a firm provides, and does not convert an unsecured loan into a protected deposit. The failure mode when that goes wrong is documented in the CeFi lending collapses.

The regulation that is closing this market

Stablecoin yield is being squeezed from two directions at once, and neither squeeze is finished.

In the United States, the GENIUS Act, enacted in July 2025, prohibits permitted payment stablecoin issuers from paying "any form of interest or yield (whether in cash, tokens, or other consideration) solely in connection with the holding, use, or retention" of a payment stablecoin. Read carefully, that provision — §4(a)(11) — binds issuers only. It says nothing about affiliates or third parties, which is precisely why exchanges characterise their programmes as platform rewards rather than issuer-paid interest. Coinbase's decision to restrict USDC rewards to Coinbase One subscribers from 15 December 2025 sits squarely in that context, and we go through it in the Coinbase review.

The Office of the Comptroller of the Currency has proposed to close that gap. Its notice of proposed rulemaking, released in February 2026 and running to 376 pages, includes a provision creating a rebuttable presumption of violation where an issuer has any contract with affiliates or related third parties — defined broadly enough to cover anyone offering yield as a service and any entity whose branding appears on a white-labelled stablecoin. If it is finalised as proposed, exchange rewards programmes on branded stablecoins become presumptively unlawful and the exchange must affirmatively prove the arrangement is not a circumvention. The comment period closed on 1 May 2026, and Comptroller Jonathan Gould has said the intention is to get a final rule out by November so applications can be processed in the new year. As of 16 September 2026 it remains a proposal.

The statutory alternative failed. The Digital Asset Market CLARITY Act, whose negotiated text would have preserved transaction-based stablecoin rewards while giving the Treasury Secretary authority to restrict them if deposit flight became a problem, lost a Senate cloture vote 49–50 on 15 September 2026. There is no federal market-structure statute, so the question is being settled by rulemaking instead.

In the European Union, MiCA already bans it. The regulation prohibits granting interest on asset-referenced tokens and on e-money tokens, and the prohibition is written broadly enough to catch any benefit related to the length of time a holder holds the token, whatever it is called — rewards, cashback or bonus — and it extends to service providers, not only issuers. Compliant platforms withdrew stablecoin rewards for EU customers accordingly. Which products remain open to you therefore depends heavily on where you live, and we track that inavailability by country.

The lawful workaround, and what it costs you

Because MiCA's prohibition is on remuneration linked to holding, a reward linked tospending falls outside it. That is not a loophole so much as a different product, and the market has moved accordingly: stablecoin card volumes reached $759 million in July 2026, roughly two and a half times a year earlier, and MEXC launched a USDT Visa card on 31 August 2026 offering up to 10% transaction cashback alongside a separate yield product through its Earn feature.

The trade-off is worth stating plainly. Cashback pays on turnover, not on balance, so it rewards spending money rather than holding it — which is the opposite of what someone looking for yield is trying to do. A 10% rebate on a small monthly spend is worth far less than a 4% rate on a large balance, and comparing the two headline percentages is meaningless. The third route neither framework closes is self-custody: neither MiCA nor GENIUS directly regulates non-custodial DeFi, so a user can hold their own keys and supply to a lending market. That is the arbitrage both regimes currently leave open, and it substitutes smart-contract risk for regulatory protection rather than removing risk.

The token and the rate are two separate risks

One habit separates people who understand this market from people who get hurt by it: assessing the stablecoin and the yield as two independent exposures rather than one product. A fully-reserved fiat-backed token can hold its peg flawlessly while the platform paying you 9% on it stops answering emails. Conversely, a well-run yield strategy can be perfectly solvent while the token it is denominated in trades at a discount on the venue where your position is marked, which is exactly what happened to USDe on Binance in October 2025. The two failures have nothing to do with each other and they require different checks.

For the token, the question is what backs it and who can redeem. A fiat-backed token is a claim on reserves. A synthetic dollar like USDe is a claim on a portfolio of hedged positions. A governance-managed stablecoin depends on collateral parameters set by vote. For the yield, the question is the one this page began with — which engine, and what is attached. Ranking the yield-bearing options by what you are actually taking makes the point quickly: OUSG at 3.44% is US government credit behind two layers of fees; Aave's USDC at 3.57% is utilisation-driven borrower demand; a Steakhouse curated vault at 3.18% to 4.36% is crypto-collateral lending with a curator's judgement and a performance fee of 5% or 25% depending which vault you are in; Ethena's sUSDe at about 5.03% is perpetual funding plus exchange and custody risk; and Euler's pools averaging 9.98% are long-tail collateral and incentives, mostly on a single chain. Those are five different businesses wearing the same dollar sign.

It is also worth being honest about how volatile these figures are. Compound's Ethereum USDC market moved across a seven-day range of 3.29% to 5.85% in the week we checked, and two reputable trackers disagreed on its rate at the same timestamp by more than a percentage point. A stablecoin yield quoted to two decimal places without a date attached is decoration. That is why every figure on this site carries the day it was checked, and why we publish the ranges rather than a single tidy number.

Reading a stablecoin rate properly

Three questions settle almost every case. First, which engine — lending, basis, treasuries or subsidy? If the provider will not say, you are being asked to trust a brand, and you should price that accordingly. Second, what does it pay relative to 3.97%? A rate below the bill needs an excuse and a rate far above it needs an explanation. Third, what is attached — a balance cap, a native-token tier, a subscription, a discretionary component, or a jurisdiction that may lose the product to a rulemaking next quarter.

Applied consistently, that framework does most of the work. It tells you why Aave's 3.57% and Bybit's 12.00% are not comparable numbers, why Sky's 3.60% is a governance decision rather than a market price, and why Ethena's 5.03% is a position on funding rates rather than an interest rate at all. The same discipline, applied to the whole market, is therates comparison; applied to protocols that pay in emissions rather than cash flows, it is DeFi yield farming.