A friend sent me a screenshot of a stablecoin vault paying somewhere in the low teens and asked the question everyone asks, is this safe. The honest answer is that the question cannot be answered in that form. Safety is a function of where the yield comes from, and where the yield comes from is knowable. Every stable yield I have ever traced lands on one of four engines. Treasury bills, overcollateralized lending demand, the perp funding basis, or token emissions. Sometimes a blend, but always those four. Once you can name the engine, you can look up its capacity, its failure mode, and whether the number on the screen is the engine actually running or a subsidy dressed up as one.
The four engines
The first engine is short-dated government debt. Tokenized treasury products, the savings rates attached to some of the bigger stablecoins, most of the yield-bearing dollar tokens aimed at institutions, they all hold bills or repo and pass the rate through. Capacity is effectively unlimited relative to anything in crypto, and the ceiling is hard, the T-bill rate itself minus fees. The ceiling is the useful part. If a product says its backing is treasuries and its yield sits meaningfully above what treasuries pay, then by definition there is a second engine or a subsidy in the mix, and your whole job is to find it.
The second engine is overcollateralized lending. Someone posts ETH or BTC as collateral on Aave or Morpho or a similar money market and borrows stablecoins against it, usually to buy more of the asset they already hold. Your deposit funds that loan and their interest is your yield. Capacity here is borrow demand, and borrow demand is market mood with an interest rate attached. When the market runs hot, people pay remarkable rates to lever long and lenders eat well. When things go quiet, utilization falls and the same pool can pay less than a treasury bill. Nothing is broken in either state. The engine is just cyclical, and the advertised APY is a snapshot of wherever the cycle happens to sit on the day you looked.
The third engine is the funding basis. Hold spot, short the perpetual future against it, collect the funding payments that leveraged longs make to keep the perp pinned to the spot price. Historically, crypto funding has been positive more often than not, because the marginal perp trader wants leveraged upside, so a delta-neutral position harvests a fairly steady stream in normal conditions. Two limits matter. Capacity is bounded by open interest, since the strategy earns what leveraged longs pay and there are only so many of them, and a product running this trade at a size that is large relative to the perp market starts compressing the very yield it is harvesting. The second limit is that funding flips negative in drawdowns, which means the engine runs in reverse at exactly the moment everything else in your portfolio is also having a bad month.
The fourth engine is emissions. A protocol prints its own token and hands it to depositors to bootstrap liquidity. The yield is spendable, you can sell the tokens, but nothing productive generated it, and the sell pressure from every farmer doing the same thing tends to grind the token price down until the APR quoted at the old price is fiction. Points programs are the same engine with the token deferred and the dilution unpriced. Emissions are fine as a bonus stacked on a real engine. As the entire yield, they are an invitation to be someone else's exit liquidity.
The risk you stack on top of the rate
A treasury bill held at a broker carries some interest rate risk and very little else. The same dollar routed through a stablecoin yield product picks up a stack. Smart contract risk on every contract in the path, and the path is often three or four protocols deep once you unwrap the vault. Depeg risk on the stablecoin itself. Oracle risk if lending is involved. Exchange and custodian risk if the funding basis is involved, because someone is holding the collateral for that short. Admin key and governance risk on whoever can upgrade the contracts. And exit risk, the question of whether you can actually redeem at par on the day everyone else wants to.
The spread over T-bills is your paycheck for carrying that stack, and pricing it that way does real work. A stable yield sitting modestly above bills means you are carrying contract risk and peg risk for a spread that rounds toward nothing, and the bill was probably the better trade. A stable yield at several times the bill rate means either the engine is running unusually hot, in which case the question is for how long, or the number leans on emissions, in which case the question is who buys your tokens.
How I classify an advertised yield
This takes maybe twenty minutes per product, and it is most of the game.
- Name the engine. Read the docs until you find the sentence that says where the interest is generated. If the docs never quite say, treat that as a finding on its own.
- Check the ceiling. T-bill engines cap at the bill rate. Lending caps at what borrowers will pay in the current mood. Basis caps at open interest. Emissions cap at the market's tolerance for dilution.
- Ask about the bad regime. What does this pay when borrow demand dries up or funding goes negative, and who eats the loss in the meantime? If a reserve fund covers the gap, find out how big it is, because a sticky advertised yield with a dead engine behind it is a countdown running on that reserve.
- Strip the emissions out. Recompute the yield with the token component set to zero and ask whether you would still deposit.
- Price the stack. Compare whatever is left against the bill rate and decide whether the spread pays you for the contracts, the peg, and the exit door.
When stables actually beat T-bills
They genuinely do, in stretches. In a proper bull market, lending rates and funding rates spike together, because both are downstream of the same crowd of leveraged longs, and organic double-digit stable yield shows up with everyone paid from real interest and real funding. Those windows are the good trade. You are effectively the house, lending chips to gamblers at the exact moment gamblers are most numerous, and the contract risk you carry is the same as in the quiet months while the pay is a multiple of it.
The mirror image is that in flat or falling markets, most on-chain stable yield compresses to below bills once you adjust for the stack, and the honest move is to rotate into bills or a tokenized version of them and wait. Funding rates and lending utilization are the tell for which regime you are in, and both are public. I watch aggregate funding across exchanges for exactly this reason, it doubles as a regime signal, and it is one of the feeds that goes into the market scorecards on Blockcircle.
None of this requires trusting anyone's marketing. The engines are visible in the rate data, the ceilings are arithmetic, and the risk stack is listed in every audit report that nobody reads. My own rule has gotten simple enough to be boring. If I cannot name the engine after twenty minutes of reading, I assume the engine is me, and the money stays in bills.