ksUSD is a carry-backed dollar for Base. Its yield comes from carry that Base's capital markets already generate — perp funding, staking yield, and lending spreads — not from token emissions or off-chain Treasury bills. ksUSD collects that carry whichever way funding runs and pays it through the share price, with every position verifiable on-chain.
Deposit USDC, hold ksUSD. The share price drifts up as Synthetix ETH-PERP funding, wstETH staking, and Aave USDC lending accrue into one Anchor vault — one program · one vault · one share mint.
Base's markets generate carry continuously. A perp market pays funding from one side to the other every hour. A liquid-staking token (LST) like wstETH compounds every epoch. A lending market reprices credit every slot. The carry is real and ongoing, but it goes uncollected by the dollars parked on top of it. The few dollars that do pay yield source it externally — from token emissions or off-chain Treasury bills — not from the chain's own activity.
ksUSD collects that carry, hedged so ETH price moves cancel, and pays it through the share price. One asset, one share class, three carry sources.
ksUSD is a rules-based engine that allocates capital across three regimes — NORMAL, REVERSE, and PARKED — based on the funding rate alone. No discretion, no manually-timed trades. A single signal drives the state, gross ETH exposure stays ≈ 0 in every regime, and transitions gate on a 7-day rolling funding mean (the primary whipsaw filter), confirmed by a 12-hour minimum hold per mode.
| Regime | Trigger (f) | Collateral | Perp leg | Carry sources |
|---|---|---|---|---|
| NORMAL | ≥ +2% | N wstETH on Synthetix | Short N ETH-PERP @ 1× | wstETH staking + funding received |
| REVERSE | ≤ −12% | USDC on Aave → borrow N·L wstETH, sell | Long N·L ETH-PERP | USDC lending + funding received − wstETH borrow cost |
| PARKED | otherwise | All capital in USDC lending (Aave, Morpho) | — | USDC lending only |
N = ETH-equivalent notional · L = 0.45 = Aave wstETH loan-to-value (LTV); N·L is the leveraged amount in REVERSE · mode switches cost ≈ 20–40 bps round-trip (swap + perp open/close).
Triggers are cost-anchored, not tuned. The +2% NORMAL threshold clears Synthetix's round-trip perp fee (~10 bps) plus the amortized mode-switch cost before the basis pays. The −12% REVERSE threshold has to clear all of that plus the reverse leg itself — the wstETH borrow rate (the lender's foregone ~5.8% staking yield plus Aave's spread), scaled by leverage L = 0.45, roughly 2–3% APR drag. That extra cost is why the negative threshold sits so much deeper than the positive one. Designs that ignore it work in only one regime.
The dead zone is a feature. When funding sits between −12% and +2%, the basis trade doesn't clear its costs, so ksUSD doesn't run it — it parks in USDC lending and earns the floor. A protocol that forces the trade in every regime is choosing to lose money in some of them. Refusing to trade is a position.
NORMAL. Deposit comes in; the vault swaps USDC → wstETH, posts it as Synthetix collateral, opens an ETH-PERP short at 1× notional. Funding accrues to the short hourly; wstETH staking accrues to the collateral every epoch. Gross ETH exposure ≈ 0. NAV rises.
REVERSE. Vault posts USDC on Aave, borrows wstETH against it at 0.45 LTV, sells the wstETH, and opens an ETH-PERP long on the leveraged notional. With funding negative, the long receives funding. Net carry = funding received + USDC lending − wstETH borrow cost. NAV rises when that net clears costs; otherwise the engine falls back to PARKED.
PARKED. All capital sits in USDC lending at ~4–5% APR until funding clears its costs again.
ksUSD is a non-rebasing share token — your balance is constant; the share price rises as carry accrues. Redeemable against vault NAV: instant from the liquidity buffer, queued for larger size. Backed by carry from Base market structure, every position verifiable on-chain.
ksUSD is not a $1-pegged stablecoin · a fixed-yield product · a Treasury-bill or RWA wrapper · an emission-subsidized token · a fiat-backed bank deposit · a directional bet on ETH.
The architecture is small on purpose — one program, one vault, a rule set you can read in an afternoon. Every dollar of attack surface is a dollar you have to defend; every discretionary lever is a place a human can be wrong or compromised. Complexity belongs in the market structure, not the protocol.
Fees hit net-new gains above the high-water mark only — capital pays no rent, exits pay no toll.
| Source | Active when | Contribution |
|---|---|---|
| Synthetix funding | Normal & reverse modes | 5–15% APR |
| wstETH staking | Normal mode collateral | ~5.8% APR |
| USDC lending | Buffer + reserve + parked | ~4–5% APR |
| Fee | Rate |
|---|---|
| Management | 0% |
| Performance | 20% > HWM |
| Reserve skim | 5% of perf |
| Withdrawal | 0% |
The design needs five things to sit on the same chain at once. Base is the only chain where they all do.
Ethereum mainnet is structurally weaker here: gas makes frequent rebalancing uneconomic; the deepest perp venues are off-chain-matched or live on separate domains, so they aren't atomically composable from one contract; and LST collateral, perp funding, and lending aren't co-located. The edge is the co-location — and Base has it.
| Variant | Gross APY | Net APY | Modeled max DD |
|---|---|---|---|
| V2 standard-only (V1 product scope) | 19.82% | 15.57% | −0.57% |
| V2 both-ways | 20.04% | 15.73% | −0.96% |
| Public conservative claim (after haircuts) | — | ~11% | — |
| sUSDe benchmark | 6.7% | 5.4% | bleeds in low-funding |
| USDC lending benchmark | — | ~4% | flat floor (alternative) |
The model nets the 20% performance fee. The ~11% public claim further deducts wstETH's ~80% Synthetix collateral haircut and execution friction not in the model. USDC lending is the relevant comparison — the actual alternative for the same capital — and ksUSD's parked floor is roughly that rate. Reverse adds only ~16 bps in backtest because it fires on just 2–4% of days; its value is regime coverage, not headline alpha. The sUSDe figure is a directional benchmark from public dashboards, not a like-for-like backtest.
A serious allocator's first question is 'how do I get out.' The second is 'what happens when the world breaks.' This section answers both, in that order. Nothing is left out because it looks bad.
| Risk | Mitigation |
|---|---|
| Smart-contract | Pre-mainnet audit (OtterSec / Sec3 / Neodyme) scheduled. Devnet only until audit complete. Do not deploy significant capital pre-audit. |
| Perp-venue dependency | Synthetix is the primary Base perp venue that supports this design. A Synthetix exploit, outage, or socialized-loss event hits ksUSD directly. This single-venue dependency is our edge and our risk. We own it. |
| Swap router | Uniswap V3 routes every USDC↔wstETH swap. Slippage is bounded on-chain. An outage stalls mode rotation; buffer redemptions stay open. |
| Counterparty | Synthetix, Aave, Morpho, Uniswap. Position-mode design bounds exposure — no single counterparty holds all NAV at once. Reserve fund absorbs first-loss within its size. |
| Oracle | Chainlink ETH/USD reads gated by 5-minute staleness and 2% confidence checks. Outage past those bounds can still cause loss. |
| Funding compression | Parked regime earns only ~4–5% USDC APR — still positive, below headline target. The ~11% target is not guaranteed in dead-zone regimes. |
| Liquidity withdrawal | Buffer + queue absorb normal flow; an extended liquidity drought lengthens the queue and widens unwind slippage. |
| Redemption queue | Queued redemptions settle against realized unwind proceeds, not marked NAV — stressed markets can reduce the payout. |
| wstETH depeg | NAV marks wstETH at the stake-pool rate — depeg hits NAV only on realized sales. A 5% depeg triggers auto-pause, arming permissionless emergency_close; the 5% reserve fund absorbs typical depeg losses (~50–150 bps). Slashing risk is unhedgeable — we accept it as structural. |
| Borrow-rate spike | A wstETH borrow-rate spike compresses or inverts REVERSE economics; the engine falls back to PARKED rather than running a negative-carry leg. |
| Liquidation | Synthetix leverage and Aave LTV carry liquidation risk in extreme rapid moves. Continuous health monitoring, auto-deleverage, and the drawdown guard cap but do not eliminate the tail. |
| Regime-transition | Mode switches cost 20–40 bps and can mistime fast funding flips; the 7-day mean + 12-hour minimum hold trade a little latency for far less whipsaw. |
| Keeper outage | Buffer withdrawals stay open; mode rotation and queue processing halt until cranking resumes. Any wallet can step in as keeper. |
emergency_close unwinds the position; buffer redemptions stay open.| Collateral | Yield source | Regime coverage | Transparency | Venue deps | |
|---|---|---|---|---|---|
| ksUSD | wstETH / USDC | Perp funding + staking + lending | Both directions + parked floor | Fully on-chain, every position verifiable | Synthetix, Aave — Base-only |
| Ethena (sUSDe) | ETH/BTC + stables | Delta-hedged perp funding + staking | Long-basis only | Off-chain CEX positions, attested | Multiple CEXs + custodians |
| MakerDAO (DAI) | Crypto + RWA | Stability fees + RWA / savings rate | n/a | On-chain + off-chain RWA | RWA counterparties |
| Sky (USDS) | Crypto + RWA | Sky savings rate (RWA / T-bills) | n/a | On-chain + off-chain RWA | RWA, multichain |
| Perena (USD*) | Stablecoin basket | Swap fees + stable yields | n/a | On-chain | Base AMM/DEX |
Ethena proved the demand for a yield-bearing dollar — and exposed the limit of a one-directional design. Through 2025, sUSDe rode positive funding to ~15% yields; as funding compressed, supply fell from roughly $15B to $5.4B and yield to ~3.7%.1 Long-basis carry is real but cyclical: it pays in a bull and starves in the flat. ksUSD runs both ways and earns the floor when neither pays.
The longer arc is monetary infrastructure: a dollar whose yield is sourced from, and verifiable against, the market structure of the chain it lives on. Keystone Finance makes that carry collectible; ksUSD is the asset that earns it. ksUSD is productive collateral for Base — other protocols hold it, lend against it, build on it.
Simulated performance does not predict future results.
1 Ethena figures are directional, per public dashboards from the same era as the backtest window.