Two Ways to Build a Delta-Neutral Dollar, and Why Flying Tulip Chose the Harder One

Key takeaways

• DeFi stablecoin yields have compressed to roughly what a brokerage pays on idle cash, which leaves where the yield comes from as the only question worth asking.
• Synthetic dollars build their hedge in one of two ways, and almost everything that matters follows from which one a protocol picked.
• The two constructions do not carry the same risks. They fail under completely different conditions, and holders are choosing between those conditions whether they know it or not.
• A quoted APY on a stablecoin does not tell you what you will actually be paid in.

In April, CoinDesk ran the numbers on the founding promise of DeFi and found it had quietly inverted. Aave USDC was paying 2.61%, its USDT 1.84%, Lido stETH 2.53%, Ethena sUSDe 3.47%. Interactive Brokers was paying 3.14% on idle cash. The headline wrote itself: the era of easy money in crypto was over, and a brokerage sweep account had started beating a smart contract.

Where a dollar of idle capital was earning, and from what

Every figure above is a single snapshot, as reported on 7 April 2026. All of them have moved since, and the rightmost column is the only one that has not.

Individual rates have moved a long way, and sUSDe was paying 4.12% in the Aave market as of 11 August on roughly 1.51 billion USDe supplied there. 

But the compression is the durable part, and it changes which question is worth asking. When a whole category of yields lands in the same range as a savings account, how much stops being interesting and from what starts being the entire story. Two products advertising the same number can be taking completely different risks to produce it.

For synthetic dollars, that question has a specific answer, and it is more useful than any rate.

Where the carry actually comes from

A synthetic dollar is not backed the way USDC is backed. There is no bank account holding a matching dollar. Instead the protocol holds a portfolio built so its net exposure to price is close to zero, and collects the income that portfolio throws off. That is what "delta-neutral" means: a long leg and a short leg, sized to cancel each other's directional risk, leaving the carry behind.

There are two ways to build the short leg.

The funding-rate construction. Hold spot collateral, such as ether, staked ether or bitcoin, and open an offsetting short position in perpetual futures. Price exposure nets out, because when the collateral falls the short gains roughly the same amount. The income is the funding rate that longs pay shorts in perpetual markets, plus any staking yield on the collateral. This is the construction that created the category, and Ethena is its reference implementation, described in its own terms as "a delta-neutral strategy using funding rates of perpetual futures positions."

The borrow-and-stake construction. Supply stablecoin collateral to a lending market and earn interest on it. Borrow a native network asset against that collateral, which is the short leg, because you now owe that asset. Stake the borrowed asset to earn staking rewards, which is the long leg. Net exposure to the asset price is close to zero, and the carry is what survives the borrowing cost: staking yield plus rewards, minus borrow cost.

Where a dollar of idle capital was earning, and from what

Both are genuinely delta-neutral. Neither is exotic. And they break under conditions that have nothing to do with each other.

The two failure modes

The funding-rate build needs funding to stay positive. When perpetual markets flip and shorts start paying longs for a sustained stretch, the short leg stops producing income and starts costing money. It also depends on where the hedge is executed, which makes the venues holding those positions counterparties, and on the operator keeping the hedge correctly sized through volatility. CryptoSlate catalogued the set for the whole category in 2025: "counterparty risk with exchanges, the funding rate turning negative for a prolonged period, or the underlying assets de-pegging."

The borrow-and-stake build has no funding exposure and no hedging venue, because there is no perpetual position. What it has instead is a debt. If the cost of borrowing the network asset climbs above the yield from staking it, the carry inverts, which is the same arithmetic that has caught looping strategies repeatedly. It also inherits the risks of the lending market and the staking provider it uses, on top of the ordinary hazards of any onchain system, because smart-contract risk and oracle risk do not disappear when the hedge moves.

So the choice is not between a risky construction and a safe one. It is between exposure to funding rates and trading venues, or exposure to borrowing costs and lending markets. A holder is picking one of those whether they have thought about it or not, and the useful version of that decision is asking which failure you would rather be exposed to, and which you would notice coming.

Which one Flying Tulip built

Flying Tulip built ftUSD on the borrow-and-stake construction, deploying what it describes as "the first fully on-chain delta-neutral architecture for ftUSD, first on Sonic and then on Ethereum," and it is now running on both, pairing a network asset with its staked form: ETH and staked ETH on Ethereum, S and staked S on Sonic.

The protocol deployed what it describes as "the first fully on-chain delta-neutral architecture for ftUSD, first on Sonic and then on Ethereum," and it is now running on both, pairing a network asset with its staked form: ETH and staked ETH on Ethereum, S and staked S on Sonic.

The mechanics are set out in the ftUSD documentation. Deposit stablecoin collateral, borrow the native asset against it, stake that asset, and let the borrowed position hedge the price risk of the staked one. Target leverage is configurable and deliberately modest. The same document names the other pairs the design contemplates, including stETH/ETH, stBNB/BNB, stAVAX/AVAX and stS/S, as ways to "capture staking yield while minimizing directional exposure."

One consequence is worth stating precisely, because it is the actual differentiator and it would be easy to overclaim. This construction does not depend on perpetual funding rates, and it does not require a hedge to be executed and maintained at a trading venue. Both are real properties of the design as it stands. Neither is a permanent philosophical commitment: the Flying Tulip documentation lists funding-based and options-based strategies as later stages, to be deployed if and when the engines behind them ship. What is true today is that the carry comes from a lending-and-staking spread, and the failure mode is a borrowing cost, not a funding flip.

A dividend, not a share issue

Here is the part that decides what a holder is actually taking on, and it has nothing to do with the percentage.

The return has two sources. The first is yield on the stablecoin collateral. The second is the delta-neutral position itself, pairing ETH with staked ETH on Ethereum, or S with staked S on Sonic. That is the whole engine, and it runs entirely onchain, which is what makes the composition checkable rather than a matter of trust.

What arrives in the hands of a staker is the part worth pausing on, because it is the difference between a dividend and a share issue. Rewards are paid in FT, and the FT used to pay them is bought on the open market out of protocol fees and revenue, with the buybacks published as a running figure. No new supply is minted to fund the reward. An emissions programme pays yield in newly issued tokens, which dilutes every existing holder and lasts precisely as long as the budget behind it. A reward funded by buying the token back on the open market is paid out of money the protocol actually earned, and it scales with the business rather than against it.

It also costs the holder something specific, which is why it belongs in the open rather than in a footnote. Part of the return arrives as a volatile token rather than as dollars, and it has to be claimed by hand. A quoted APY on a stablecoin implies dollars; part of this one is not dollars. Anyone comparing it against a number from a different protocol should know which of those two things they are comparing.

Two protocols, two definitions of the same word

There is a second problem with comparing quoted rates, and almost nobody writing about stablecoin yield mentions it. APY is not computed the same way everywhere.

Flying Tulip publishes two figures and labels them separately. Current APY is the annual percentage yield based on the trailing thirty days. Projected APY is the current rate annualised. Those two numbers can sit a long way apart in a month when the underlying spread has moved, and they answer different questions: one tells you what the strategy did, the other tells you what it would pay if today lasted a year.

Put a trailing-thirty-day figure from one protocol next to a spot-annualised figure from another and the comparison is not measuring what it appears to measure. Before comparing two yields, it is worth checking whether they are the same kind of number. Frequently they are not.

What you can check

None of the above is worth much if it can't be verified, which is the one place this category is unusually testable.

Flying Tulip publishes a strategy breakdown that decomposes the ftUSD APY into its stablecoin-collateral component and its delta-neutral-strategy component, alongside the leverage employed, the borrow rate being paid, the staking rate being earned, and a profitability read on the delta-neutral leg. The current rate lives there rather than in this article, for the reason the previous section gives: it is variable, and a number printed here would be wrong by the time you read it. The July update carries an overview of the dashboards and a list of resources at the foot of the post. Attribution at that granularity is rare. Most products publish one number and leave the reader to trust it.

Redemption terms are worth checking directly rather than taking on trust from anyone, this article included. No lockup or cooldown is applied to ftUSD today, and the documentation is candid that exit parameters are shown in-app and remain subject to risk settings and network conditions, as they should be in a system that also runs circuit breakers on abnormal outflows. A system that can never gate an outflow is not safer. It is just less honest about what it would do in a crisis.

Rates move. Constructions don't.

Every rate quoted above is a snapshot, and snapshots age: yields shift with funding, with borrow costs, with demand for leverage. Anyone choosing a synthetic dollar on this month's highest number is choosing on the one input guaranteed to change.

What does not change is how the carry is built. A protocol earning it from perpetual funding stays exposed to funding and to the venues where it hedges. A protocol earning it from a lending-and-staking spread stays exposed to borrowing costs. Three questions survive every repricing: what instruments create the carry, what has to happen for it to invert, and what you are actually paid in.

Ask those three and the rate becomes a detail, one you can check yourself rather than take on trust. Ask only about the rate and you learn what a protocol earned last month, which is the one thing that cannot tell you what happens next.

Sources referenced

Flying Tulip

• ^{April update, the borrow-and-stake mechanics, the named staking pairs, and FT obtained through secondary market buybacks: https://blog.flyingtulip.com/flying-tulip-april-update/}
• ^{June update, "the first fully on-chain delta-neutral architecture for ftUSD," the three yield sources, and "no lockups": https://blog.flyingtulip.com/flying-tulip-june-update/}
• ^{July update, live on both chains, and the 3 August staking rates: https://blog.flyingtulip.com/flying-tulip-july-update/}
• ^{ftUSD documentation, funding-based and options-based strategies as later stages, claiming rewards as FT, treasury discretion, and in-app exit parameters: https://docs.flyingtulip.com/product-suite/ft-usd}
• ^{The ftUSD dashboard, APY decomposition, leverage, borrow and staking rates: https://flyingtulip.com/ftusd/dashboard}
• ^{The APY decomposition dashboard announcement, via @flyingtulip_, 9 July 2026: https://x.com/flyingtulip_/status/2075257118194541011}
• ^{Circuit breakers on abnormal outflows, via @flyingtulip_, 16 July 2026: https://x.com/flyingtulip_/status/2077868331638325556}

Third-party data and news

• ^{Aave USDC 2.61%, USDT 1.84%, stETH 2.53%, sUSDe 3.47% and Interactive Brokers 3.14%, CoinDesk, 7 April 2026: https://www.coindesk.com/business/2026/04/07/defi-yields-are-crashing-so-hard-that-they-can-t-compete-with-a-traditional-savings-account}
• ^{sUSDe at 4.12% on roughly 1.51 billion USDe supplied in the Aave market, and the Ethena description of its own strategy, Aavescan, 11 August 2026: https://aavescan.com/rates/ethena-susde}
• ^{Category risks for synthetic dollars, CryptoSlate, September 2025: https://cryptoslate.com/where-did-all-the-boring-dollars-go-how-synthetics-are-turning-stablecoins-into-green-lean-yield-machines/}