Where crypto yield actually comes from
Ask one question before you look at any rate: who is paying you. The payer differs across the six common approaches, and who pays determines what the risk looks like.
When nobody can name the payer, the payer is you
Every yield has a source. Someone borrows your money and pays interest for it, or a protocol issues new supply by rule, or a share of trading fees gets routed to you. All of those are legitimate, and each carries its own risk. The dangerous case is the fourth one — where the payer cannot be named. In that case the money usually comes from whoever deposited after you, and the reported yield is typically the highest one on offer right up until it stops.
So the second column of the table below matters more than the first. It is also a question you can put to any platform, this one included — if they cannot answer it, the headline rate is not worth reading.
Six approaches, and where the money comes from
This table does not compare headline rates. That column expires, and it cannot be evidenced down to any particular venue. It compares the source of the yield and the risk that comes with it — those are structural, and they hold for years.
| Approach | Who pays the yield | Main risk | What makes it move |
|---|---|---|---|
| Exchange flexible savings | Whoever borrows the coin to open a leveraged position | Platform credit; a run if borrowers default together | Tracks borrowing demand in the derivatives market — thin when the market is quiet |
| Proof-of-stake staking | The protocol, by issuing new supply, plus a share of on-chain fees | Paid in the same coin, so a price drop can exceed the yield; unbonding delays; slashing | Tracks the network staking ratio and on-chain activity |
| Stablecoin lending | Whoever borrows the stablecoin, or the short-term treasuries behind it | Issuer and custodian; de-pegging under stress | Follows USD rates — when they fall, it falls |
| DeFi liquidity mining | Trading fees, plus incentive tokens minted by the project | Impermanent loss; contract exploits; the incentive token going to zero | Tracks the incentive budget, which usually decays fast |
| Fixed-rate lending platforms | The borrower, through interest | Borrower default; a run once the platform is maturity-mismatched | Fixed in name; in practice it depends on whether the risk desk holds up |
| Funding rate arbitrage | The long side of the perpetual contract, settled every 8 hours | Funding turning negative; exchange risk; platform credit risk from custody | Tracks market sentiment — rich in a bull market, thin in chop and bear |
None of the six rows is the safe one. They place the risk in different places: staking puts it on the coin price, stablecoin lending on the issuer and on USD rates, liquidity mining on contract code and on how long the incentives last, and funding rate arbitrageremoves directional price exposure and takes on funding turning negative plus counterparty risk instead. Which one suits you depends on which risk you can live with, not on whose number is biggest.
Gross and net: four things behind any "X% APR"
Change the denomination, change the compounding convention, leave out two cost lines, and the same real result can be reported at more than double the rate — with none of those four versions being a lie. So "is 8% realistic" cannot be answered before these four are separated out.
Denominated in what
A coin-denominated "5% APR" means you end up with more coins, and you are still down if the coin fell 30%. Only a USD-denominated figure speaks to preserving value.
How to check: Ask directly: at maturity do I get back more coins, or more USDT?
Summed or compounded
Compounding assumes every payment is immediately redeployed. The number looks better, but it requires the reinvestment to actually happen. Reporting compounded figures without compounding is inflation.
How to check: Ask whether it auto-compounds. If it does not, the cumulative figure should be a sum.
Which costs were deducted
This is where gross and net part company. A figure net of one fee line sits materially above a figure net of all of them, and nothing about it looks wrong.
How to check: Make them list every cost line: fees on both legs, cost of capital, slippage, exit cost.
A commitment or a backtest
A historical result and a contractual floor are different things that look identical on a marketing chart.
How to check: Ask whether the number goes into the contract. If it does not, it is history, not a promise.
On the funding-rate route, who is the payer
A perpetual contract has no delivery date. A funding payment settled every 8 hours is what pulls its price back toward spot. Most of the time the contract trades above spot, sothe long side pays the short side — and the short leg of a delta-neutral position is the side collecting. The payer has a name. That is the structural difference between this and anything whose source cannot be traced.
How often "most of the time" is can be checked directly. Of the 878,142 settlement records in the database, BTC accounts for 7,344 periods, of which85.95% were positive, going back to 2020-01-01. The year-by-year breakdown, including the stretches where it turned negative, is onfunding rate history data.
A high share of positive periods does not mean every period pays.The same dataset contains pairs that fund negative for long stretches; ending up on the receiving side is not automatic. That is also why this strategy only takes positions on the most liquid pairs rather than spreading across the whole market.
How far this sits from the word "stable"
Funding follows sentiment. In a bull market longs crowd in and the payment is both frequent and rich; in chop and in bear markets it thins out, and in some stretches it inverts.This is not a deposit product, and it has none of the steadiness of an interest rate.
So before treating it as "stable crypto yield", pin down what the word is doing:stable here means not betting on direction — price moves are offset between the two legs — not that each period pays the same amount. Those two get spoken about interchangeably, and the gap between them is the reason this page exists.
This site puts the worst year and the good years in the same table at the same weight: across 2021–2025, ETHUSDT in 2022 returned 7.873% for the entire year (scaled 10x), against an average of 124.296% over the same span. The year-by-year detail as of 2026-08-25 is onperformance — read that year first, then the average.
This product, by the same four questions
From 5,000 USDT, terms of 1-5 years, and one flat 2% annual management fee across all three plans, taken from returns. What differs is only how much protection you take and how much of the upside you give up.
Answering our own checklist: USD-denominated; cumulative figures aresummed, not compounded (the system does not auto-reinvest); costs deduct fees on both legs, cost of capital and slippage, not just the futures leg; and the7% floor applies only to the principal-and-yield protected plan, and it is written into the contract. Of the other two plans, one protects principal only, and one protects neither. The management fee is 2% a year.
One label worth being precise about: "fixed yield" is accurate for exactly one of the three plans. That plan writes the rate into the contract, which earns the word. The other two move with the market, and calling them fixed would be saying something off-site that the site itself already denies.
Funds are held in a platform account, which means you carry platform credit risk. Nothing above removes it; it is set out line by line on risk and boundaries. If you want to compare against running the strategy yourself first, seeself-hosted bot or managed.