Real Yield in 2026: Every Yield Is a Spread Over T-Bills

Two numbers framed decentralized finance in 2026. The first: crypto projects spent a record $638 million buying back their own tokens this year, and just two of them, the perpetuals exchange Hyperliquid and the Solana memecoin launchpad Pump.fun, accounted for nearly 90% of it, according to Cointelegraph. The second: tokenized US Treasuries on public blockchains grew to roughly $15.92 billion, paying an average of about 3.44%, per rwa.xyz. One number is crypto paying itself. The other is crypto plugging directly into the plumbing of the US government balance sheet.

Both fall under the banner of “real yield,” the phrase DeFi adopted after the 2022 bear market to separate money that comes from actual revenue from money that comes from printing tokens. In 2026 the label is everywhere, stamped on staking dashboards, stablecoin savings rates, exchange fee-sharing schemes and buyback trackers. The problem is that it has become a marketing word as much as a technical one. When a protocol pays out more than it earns, that is not real yield; it is a subsidy with a countdown timer.

There is a cleaner way to read every on-chain yield, and 2026 finally made it possible. Now that tokenized Treasuries give DeFi a genuine risk-free rate, any yield can be split into two parts: the base rate you could earn with almost no risk, and the spread you are paid on top for taking a specific, nameable risk. If you cannot name the risk behind the spread, you are not earning yield; you are making a bet. This guide walks through the base rate, the spreads, how to measure them, the buyback machines now dominating the numbers, and the macro and regulatory forces squeezing the whole stack. For the companion question of how to separate durable cash flow from disguised emissions, see our guide on telling cash flow from emissions.

What real yield actually means

Real yield is income paid to token holders or depositors that comes from money the protocol actually collected from users: trading fees, borrowing interest, perpetual funding payments, MEV, or the coupons on real-world assets. It is set against inflationary yield, where a protocol mints new tokens and hands them out as rewards. The second kind can look identical on a dashboard, a big annual percentage number, but it is funded by dilution rather than demand. Holders are paying themselves with their own supply.

The simplest test is one we keep coming back to: would the yield survive if the protocol own token went to zero? Fees paid in ETH, USDC or other stablecoins survive. Rewards paid in a freshly minted governance token do not; without the token there is nothing to pay. A related trap is confusing activity with cash flow. A token can trade billions in volume and still generate no revenue for anyone, as the drawn-out autopsy of one dead AI crypto fund showed: attention is not a business model, and a lively chart is not a coupon.

Real yield also has to clear a subtler bar in 2026: it has to be net of incentives. If a protocol distributes $23 million to its token holders but only earned $8 million in fees, the extra $15 million came from somewhere (its treasury, its investors, its next token unlock), and the yield is really a rebate on the marketing budget. More on that arithmetic below, because it is where most of the mislabeling happens.

Where the idea came from

The phrase hardened during the 2022 downturn, when the double-digit and triple-digit yields of the DeFi summer of 2020 and 2021 collapsed. Those returns had been paid almost entirely in emissions; when token prices fell, the yields evaporated and the mercenary capital chasing them moved on, leaving protocols with nothing. A handful of projects that paid holders in hard assets instead of their own inflation became the movement flagbearers.

GMX, the perpetuals exchange that launched on Arbitrum and Avalanche, was the canonical example. It split platform fees so that 30% went to GMX stakers and 70% to the liquidity providers backing its GLP pool, and crucially it paid stakers in ETH and AVAX, real assets earned from real trading, rather than in freshly minted GMX. Synthetix, Gains Network and dYdX pushed variants of the same idea. The pitch was simple and, after a year of collapses, radical: a yield you could actually keep.

Four years on, GMX has been dwarfed by newer venues, but the principle it popularized is now the organizing question of the entire sector. The difference in 2026 is that we finally have a yardstick to measure it against.

DeFi found a risk-free rate: tokenized Treasuries

For most of its history, DeFi had no risk-free rate. In traditional finance the yield on short-term US Treasury bills is the anchor that every other return is measured against, because the US government is treated as the closest thing to a certain payer. On-chain there was no equivalent, so every yield floated free, with nothing to compare it to. Tokenized Treasuries changed that.

These are shares in funds that hold cash, Treasury bills and repo, wrapped as blockchain tokens that pay their coupon on-chain. The market has grown fast: from about $8.9 billion at the start of 2026 to $10.8 billion by late February, per Cointelegraph, and to roughly $15.92 billion by early September, spread across more than 67,000 holders, according to rwa.xyz. The average yield sits near 3.44%, which is no accident: it is the short-term Treasury rate passed through, minus a management fee.

The league table reads like a Wall Street roll call. BlackRock BUIDL fund, tokenized with Securitize and launched in March 2024, holds about $2.75 billion; Circle USYC, the money-market token it acquired with Hashnote in early 2025, is neck and neck at roughly $2.69 billion; Ondo USDY is around $2.19 billion; Franklin Templeton on-chain government money fund adds more than $1.71 billion through its iBENJI token; and WisdomTree WTGXX rounds out the top tier near $1.22 billion, all per rwa.xyz. These are not crypto-native experiments; they are some of the largest asset managers in the world putting their money-market funds on public chains.

Why does this matter for real yield? Because it sets the floor. A tokenized T-bill pays around 3.44% for taking essentially no credit risk and no token risk. That is the number every other on-chain yield now has to beat to justify the extra risk it carries. The base itself is a policy variable: the effective federal funds rate sat at 3.63% in early September, with three-month bills yielding about 3.75%, per the Federal Reserve H.15 release. When the Fed moves, DeFi risk-free rate moves with it, which is why the September policy meeting (covered below) matters to yield farmers who have never watched a Fed statement in their lives.

Every yield is a base rate plus a spread

Once you have a base rate, the whole field snaps into focus. Any yield you are offered can be split into the base rate (about 3.4% to 3.5% right now, in dollars) plus a spread, the extra return you collect for accepting some risk the T-bill does not carry. The discipline is to insist that the spread has a name. Is it funding-rate risk? Smart-contract risk? Liquidity or duration risk? Counterparty risk? If the answer is that you are not sure but the number is high, the spread is not compensation; it is a warning.

The table below decomposes the main sources of on-chain yield this way. Note the two that pay less than the base rate: pure ETH staking, whose reward is denominated in ETH and currently sits below the dollar T-bill rate, and most governance staking, which often pays a negative real spread once you account for token inflation.

The specific numbers behind that table are sourced in the sections that follow: the Sky Savings Rate is governance-set at 3.75% (per Sky), sUSDe pays around 4% (per DefiLlama), and Ethereum staking runs about 3% to 3.8% all-in in ETH terms (per ethereum.org). The pattern is the tell: the safest dollar yields cluster just above the base, and the wildest numbers come with the least nameable risks.

How to measure it: fees, revenue and holders revenue

The single most common mistake in DeFi analysis is quoting the wrong revenue number. Data providers such as DefiLlama and Token Terminal separate three figures that sound alike and mean very different things. Fees are the total paid by users. Revenue is the slice the protocol keeps for its treasury or token holders. Holders revenue is the portion that actually reaches token holders, through a distribution, a buyback or a burn. The gap between them is where most of the confusion lives.

Uniswap is the textbook case. Traders paid it roughly $875 million in swap fees over the trailing year, but the protocol itself kept only about $40 million, according to DefiLlama. The rest went to the liquidity providers who supply the pools. For years Uniswap kept nothing at all: its protocol fee switch was dormant, and 100% of fees flowed to LPs. That changed in December 2025, which we come back to in the buyback section. The lesson for now: a protocol can be enormous by fee volume and small by the revenue that supports its token.

Aave shows the same split from the lending side: borrowers paid it about $800 million in interest and fees over the trailing year, of which roughly $104 million was revenue kept by the protocol, per DefiLlama; the rest went to depositors as their yield. Once you have the revenue figure, you can value a token the way you would value a company, with a price-to-fees or price-to-revenue multiple, and you can ask whether the market is paying a reasonable price for the cash flow or a speculative premium for the story.

The Uniswap and Aave figures above are from DefiLlama. The third row is the trap the first two help you avoid: a farm advertising a headline APY with no fee revenue behind it is not underpaying its token holders, it is paying them with printed supply.

The subsidy test, and DeFi concentration problem

The most useful thing you can do with those three numbers is subtract. Take what a protocol paid its token holders and subtract what it actually earned. If the payout is larger than the revenue, the difference is a subsidy, and subsidies end. The perpetuals venue edgeX was the cleanest 2026 example: it ranked among the top payers of holders revenue at about $23.3 million over 30 days, yet its actual protocol revenue over the same window was only around $8 million, according to figures on DefiLlama. It was paying out roughly three times what it took in. That is not real yield; it is customer acquisition financed by the treasury, and it flatters every APY on the dashboard until the money runs out.

Subtraction also reveals how concentrated real yield has become. Across all of DeFi, the ten largest earners account for roughly 87% of holders revenue, per DefiLlama, and the top of that list is dominated by a handful of trading venues. Hyperliquid alone generated about $53.5 million over 30 days, close to 38% of the sector total, with Pump.fun and edgeX behind it. When one or two protocols produce most of the industry genuine cash flow, “DeFi real yield” is really “a few exchanges real yield,” and the fragility that implies is a risk in itself.

Figures from DefiLlama. Read the last column first: two of the three biggest payers are funded by real fees, and one is funded by its own balance sheet. Only one of those is a yield you can count on next quarter.

Staking: the base layer of crypto-native yield

Staking is where most people first meet real yield, and it is the clearest example of a yield that is real but not necessarily generous. When you stake ETH, you earn newly issued ETH for helping secure the network, plus a share of priority fees and MEV. That is genuine income, paid in the asset itself rather than in a side token. But the all-in rate is modest: Ethereum base consensus reward runs around 2.7%, and about 3% to 3.8% once priority fees and MEV are included, with roughly a third of all ETH now staked, per ethereum.org.

Here the base-rate lens bites. That 3% to 3.8% is denominated in ETH, while a T-bill pays about 3.44% in dollars. A staker is therefore earning a yield roughly equal to or below the risk-free rate, and taking ETH price risk and slashing risk on top. Staking makes sense if you want ETH exposure anyway and would rather earn on it than hold it idle, but it is not a way to beat the risk-free rate; it is a way to be paid a little for a risk you have already chosen. The wider trade-off between what security costs and what it pays is the subject of our comparison of hashprice and staking yield.

Liquid staking tokens such as Lido stETH wrap this yield so it stays liquid and composable, taking a cut (Lido keeps 10% of staking rewards, split between node operators and its DAO, per Lido) in exchange. That fee is itself real revenue for the protocol, which is why liquid-staking governance tokens are among the more defensible real-yield plays: the fee is charged on a service people demonstrably want, and it is paid in staking rewards, not in emissions.

Stablecoin savings: Sky and Ethena

If ETH staking pays roughly the base rate for taking ETH risk, dollar-denominated stablecoin savings pay roughly the base rate for taking much smaller risks, which makes them the closest crypto-native cousins of a Treasury fund. Sky, the protocol formerly known as MakerDAO, offers the Sky Savings Rate on its USDS stablecoin. It is set by governance and currently sits at 3.75%, per Sky, funded largely by the interest Sky earns on the real-world assets and over-collateralized crypto loans backing USDS. That is about as clean as on-chain real yield gets: a dollar in, a dollar out, plus a spread a hair above the T-bill rate for taking protocol and collateral risk.

Ethena sUSDe is the more interesting, and more debated, case. Its yield comes from a delta-neutral trade: hold staked ETH and other collateral, short an equivalent amount of perpetual futures, and collect the staking yield plus the funding rate that longs pay shorts. Ethena founder Guy Young has marketed the product as a crypto-native “Internet Bond,” a phrase now widely used to describe it (see Messari). The catch is that the funding-rate component is not a fixed coupon; it swings with market sentiment and can turn negative. That is why sUSDe yield has compressed from the double digits it printed in 2024 to around 4% in 2026, per DefiLlama. The yield is real, but the spread is variable, and the risk you are paid for, a funding-rate reversal, is exactly the one that shows up when you least want it.

Trading fees and the buyback era

The biggest change to real yield in 2026 was not a new source of revenue but a new way of returning it. Instead of paying holders a cash-like distribution, the largest protocols started using their revenue to buy back and destroy their own tokens. The mechanics differ, but the intent is the same: turn fee income into value for holders without cutting them a check that a regulator might call a dividend.

Uniswap led the shift. After years of a dormant fee switch, its DAO passed the “UNIfication” proposal in late December 2025 with more than 125 million UNI in favor and just 742 against, per DL News. It turned protocol fees on, set Uniswap Labs front-end fees to zero, and burned 100 million UNI from the treasury in one shot, worth close to $600 million at the time. Rather than distribute fees, the design routes them into an on-chain contract called TokenJar, from which value can only be withdrawn by burning UNI through a second contract, the Firepit, a continuous supply-reducing loop described in Uniswap own announcement. In July 2026 the switch was extended to Uniswap v4 pools across seven chains, roughly tripling protocol revenue to about $325,000 a day, according to The Defiant.

Aave built a more explicit buyback engine. After approving a weekly buyback in 2025, its DAO moved in mid-2026 to an automated program under the “Aavenomics 3.0” overhaul, routing protocol and GHO revenue into continuous on-market AAVE purchases. Founder Stani Kulechov described the update as delivering “immutable and automated buybacks of AAVE,” per The Defiant. Notably, the DAO also trimmed the buyback budget from about $50 million a year to $30 million in March 2026, per Aave governance, after its own accounting showed spending outrunning revenue, a reminder that even the best-run buyback is only as durable as the fees behind it. By that point Aave had bought back more than 205,000 AAVE, around 1.28% of supply.

GMX, the original model, still splits fees to stakers in ETH and AVAX rather than buying back, proof that the older distribution approach has not disappeared. The buyback wave is best understood as a second option that became fashionable once a friendlier US regulatory climate (more on that below) made returning cash to holders look less legally fraught.

Hyperliquid: the biggest payer, and the cracks

No protocol embodies the promise and the strain of the buyback era better than Hyperliquid, the perpetuals exchange whose HYPE token became the poster child for on-chain cash flow. Hyperliquid recycles roughly 97% of its trading fees into open-market HYPE purchases through a mechanism it calls the Assistance Fund, per OAK Research. When volumes are high, that is an enormous, visible bid for the token, funded entirely by real usage. For most of 2025 it was the single largest source of holders revenue in all of DeFi.

Then the cracks showed. The Assistance Fund bought about $290 million of HYPE in the third quarter of 2025 but only around $149 million in the second quarter of 2026, roughly half as much, as protocol revenue fell about 43% from its peak to near $202 million, again according to OAK Research. Records were still being set on volume and open interest, yet the revenue backing the buyback was shrinking. Part of the reason is a shift toward Hyperliquid HIP-3 real-world-asset perpetuals, which share more of their fees with third-party builders, eating into the take that funds HYPE, as CoinDesk reported.

The symbolic blow came from Pump.fun. The Solana memecoin launchpad overtook Hyperliquid in monthly revenue in 2026, and across the year the two together carried out nearly 90% of a record $638 million in token buybacks, with Hyperliquid at roughly $370 million and Pump.fun near $200 million, per Cointelegraph. That a memecoin factory can out-earn the most sophisticated exchange in DeFi is either proof that real yield can come from anywhere, or a warning that the revenue is only as durable as the speculative appetite feeding it. Both readings are defensible, which is the point.

Buyback, burn or distribution: how value reaches you

Not all returning value to holders is equal, and the mechanism changes both the risk and the tax and regulatory profile. There are four broad models in use, and knowing which one a protocol uses tells you a lot about how, and whether, you actually get paid.

Two practical points follow. First, a buyback only helps you if the buying pressure is real and sustained; a one-quarter burn during a revenue slump is a press release, not a yield. Second, the mechanism has tax consequences. A distribution is usually income the moment you receive it, while a buyback or a wrapper that simply grows in value may not be taxed until you sell, and the treatment varies by where you live. US readers should treat any of these as potentially taxable and check the specifics; our state-by-state guide on how crypto taxes differ across the country lays out why the same yield can leave two people with very different after-tax returns.

The macro squeeze: a base rate that may rise

Because DeFi risk-free rate is now the Treasury rate, DeFi yields have become hostage to the Federal Reserve. Through most of 2026 the Fed held its policy rate at 3.50% to 3.75%, keeping the tokenized-Treasury base steady near 3.44%. The debate heading into the September meeting was whether the next move would be a cut, a hold, or, unusually, a hike.

Fed Chair Kevin Warsh tilted that debate hawkish. In his August 28 Jackson Hole address, titled “In Our Time,” he warned that “price stability is not self-executing, nor is inflation necessarily mean-reverting,” and reaffirmed the 2% target as “a firm, fixed target,” per the Federal Reserve. Markets moved quickly: the implied odds of a September rate hike jumped from around 37% a week earlier to a coin flip or higher across prediction venues, with CME FedWatch reading near 70% at one point, as Cryptonews and others reported.

For real yield, a rising base rate is a squeeze from both ends. It lifts the risk-free floor, so a 4% on-chain yield that looked attractive against a 3.4% base looks thin against a 4% base; and it pulls capital back toward Treasuries, draining the liquidity that crypto-native yields depend on. The one category that benefits is the tokenized Treasuries themselves, which simply pass a higher rate straight through. The full calendar of policy dates and data prints that will decide how this plays out is laid out in our September countdown.

Where the SEC stands

Real yield has always carried a legal shadow: a token that pays its holders a share of profits looks a lot like a security under the US Supreme Court Howey test, which asks whether investors expect profits from the efforts of others. That fear is a big part of why Uniswap sat on its fee switch for years, and why the protocols that finally turned revenue on in 2026 mostly chose burns over dividends. Burning tokens raises the value of the rest without ever paying anyone directly, which is harder to characterize as a dividend.

The regulatory climate has also shifted. Under Chair Paul Atkins, the SEC spent 2026 rolling out a friendlier framework it calls Project Crypto. In an August statement on regulation of crypto assets, the agency proposed tailored exemptions and a taxonomy that treats most crypto tokens as not themselves securities, and separate SEC staff guidance has indicated that certain protocol-staking activities are not securities transactions, per the SEC. That is a meaningful thaw from the enforcement posture of a few years earlier, when the agency extracted a settlement over a staking-as-a-service product. It is not a blanket clearance: guidance is not law, the carve-outs have conditions, and a token explicitly engineered to pass profits through to holders still invites scrutiny. But the direction of travel in 2026 gave protocols the confidence to return revenue in the open.

The risks a headline yield hides

Even genuine real yield carries risks that no APY number shows. Durability is the first: fees are cyclical, and a yield that depends on bull-market volume will shrink in a downturn, exactly as Hyperliquid buyback did. Funding-rate strategies like Ethena can see their spread go negative when sentiment flips. Smart-contract risk is ever-present; a yield paid by a contract is only as safe as that contract code and its keys. And concentration is a systemic risk, given how few protocols produce most of the sector cash flow.

Tokenized Treasuries, the safest source on paper, carry their own quieter risks: the credit of the issuer and custodian standing between you and the actual bills, the smart-contract layer, and, for anyone whose home currency is not the dollar, exchange-rate risk that can dwarf a 3.44% coupon in a single bad month. And because so much on-chain yield now sits inside custodial or semi-custodial wrappers, the security of keys and access matters as much as the yield itself. A high yield on an insecure foundation is not a high yield; it is a larger target.

A five-question checklist before you chase a yield

Before committing capital to any on-chain yield in 2026, run it through five questions. If it fails the first two, stop.

• Would the yield survive if the protocol token went to zero? If it is paid in the protocol own inflation, it is emissions, not revenue.
• Is the payout larger than the protocol revenue? Subtract one from the other; a negative answer means you are collecting a subsidy on a timer.
• What is the spread over the base rate, and what risk am I paid for? If the yield is below about 3.4% in dollars, you are being paid less than a T-bill; if it is far above, name the risk or assume it is the one you cannot see.
• How does value actually reach me? Distribution, buyback, burn and wrapper accrual each behave differently in a downturn and at tax time.
• Is the revenue durable or cyclical? A yield built on bull-market trading volume is not the same as one built on recurring demand.

Frequently Asked Questions

What is real yield in DeFi?

Real yield is return paid from a protocol actual revenue, such as trading fees, lending interest, perpetual funding or the coupons on tokenized real-world assets, rather than from newly minted tokens. The test is whether the income would survive if the protocol own token went to zero. Fees paid in ETH or stablecoins survive; rewards printed in a governance token do not.

How can I tell if a DeFi yield is real or just token emissions?

Compare what a protocol pays out to what it earns. Data sites like DefiLlama publish both fees, meaning what users pay, and revenue, meaning what the protocol keeps; if the payout to holders is larger than revenue, the difference is a subsidy funded by the treasury or by inflation. Also check the currency of the reward: real yield is paid in hard assets, not in the protocol own freshly issued token.

What is a good real yield rate in 2026?

Start from the base rate. Tokenized US Treasuries paid about 3.44% in early September 2026, and that is roughly the risk-free floor for a dollar yield. A real yield meaningfully above that is possible, but every extra percentage point should correspond to a specific risk you can name; a yield far above the base with no clear risk is usually a subsidy or a hidden danger.

Is Ethereum staking real yield?

Yes, but it is modest. Staking pays real income, newly issued ETH plus priority fees and MEV, for securing the network, running around 3% to 3.8% all-in. Because that yield is denominated in ETH and sits near or below the dollar T-bill rate, stakers are effectively paid roughly the base rate while taking ETH price and slashing risk, so it suits holders who want ETH exposure anyway rather than those chasing the highest return.

Are crypto token buybacks the same as dividends?

No. A dividend pays holders directly in cash or stablecoins, while a buyback uses protocol revenue to purchase and often burn the token, raising the value of the remaining supply without a direct payment. Protocols such as Uniswap and Aave favored buybacks and burns in 2026 partly because they are harder to characterize as securities-like profit distributions, and partly because the tax treatment can be deferred until you sell.

By Adaeze Okafor, HOGE Wire senior DeFi correspondent.