In September 2024, MakerDAO — the protocol that had quietly minted the first credible decentralised dollar for seven years — announced it was rebranding as Sky. DAI didn't disappear; it got a sibling called USDS, a new governance token (SKY, replacing MKR at 24,000:1), and a roadmap that eventually splits the project into smaller, focused "Stars" subDAOs. To users, it looked like a logo change. Underneath, it was the most consequential redesign in the history of crypto-collateralised stablecoins.
This matters because crypto-collateralised dollars sit on the philosophical opposite end of the stablecoin spectrum from USDC and USDT. Fiat-backed coins ask you to trust an issuer and its banks. Crypto-backed coins ask you to trust a smart contract and the assets locked inside it. After the GENIUS Act of 2025 federalised payment-stablecoin regulation in the US, that distinction stopped being academic. If you want a dollar that nobody can freeze, blacklist, or pull off-chain by court order, this is the category you live in.
A Short History of DAI
DAI launched in December 2017 as Single-Collateral DAI (SAI), backed only by ETH. In November 2019 it became Multi-Collateral DAI, accepting a basket of approved assets. The mechanism was elegant: lock collateral worth more than the DAI you minted, pay a Stability Fee, and get liquidated if your collateral ratio fell below the minimum. If you've read our piece on how on-chain lending works, DAI is essentially that pattern with a stablecoin as the borrowed asset.
For years DAI held its peg through a combination of over-collateralisation, the DAI Savings Rate (DSR), and Peg Stability Modules that let users swap USDC 1:1 for DAI to absorb volatility. By 2022, DAI's circulating supply peaked above $10B. Then came the slow pivot: real-world assets (US Treasuries via Monetalis, BlockTower, and others) crept onto the balance sheet, and by 2024 a meaningful share of DAI's backing was, ironically, off-chain Treasury bills earning yield for the protocol. That tension — a "decentralised" dollar increasingly backed by the same instruments as USDC — is part of what triggered the Sky redesign.
The Sky Migration
The rebrand wasn't cosmetic. Sky introduced:
- USDS, a new ERC-20 stablecoin, 1:1 upgradeable from DAI through an on-chain converter. DAI continues to exist; USDS is the new default.
- SKY, the governance token. MKR holders can convert at 24,000 SKY per 1 MKR.
- The Sky Savings Rate (SSR), USDS's equivalent of the DSR — deposit USDS, receive sUSDS, which accrues yield automatically.
- A freeze function on USDS (DAI does not have one). This was the most controversial change: USDS is technically capable of address-level freezes via governance, which Sky argues is necessary for institutional adoption and regulatory comfort. Purists kept their DAI.
- A planned split into Stars — semi-independent subDAOs with their own tokens and mandates. Spark is the first Star and operates SparkLend plus the Sky Savings Rate distribution.
As of April 2026, USDS supply sits in the multi-billion range — somewhere between $7B and $11B depending on which dashboard you ask — and DAI around $3–4B, for a combined ~$10–14B. Coinbase auto-converted eligible DAI balances to USDS at 1:1 between 4–6 May 2026. The SSR has ranged between 3.75% and 4.5% over the past year, tracking a blend of on-chain borrowing demand and Treasury yields.
How USDS Actually Works
Mechanically, USDS is not radically different from late-stage DAI. Users open Vaults (formerly CDPs), deposit collateral (ETH, stETH, wstETH, WBTC, RWA tokens, and more), and mint USDS up to a collateral-specific debt ceiling and liquidation ratio. Stability Fees accrue on the debt; if collateral value drops, vaults are liquidated through Dutch auctions.
The Sky Savings Rate is the demand-side lever. When USDS trades above $1, governance lowers SSR to discourage holding; when it trades below, SSR rises to pull demand in. Behind the scenes, the rate is funded by Stability Fees paid by borrowers plus yield from RWA holdings. It's the closest thing crypto has to a central-bank policy rate, except the "committee" is SKY token holders voting on-chain.
Price feeds for liquidations come from a network of oracles — and if you want to understand why that piece is load-bearing, the oracles and price feeds explainer walks through the failure modes. A bad oracle print is the single biggest existential risk for any over-collateralised stablecoin.
USDe: A Different Animal
Ethena's USDe is often lumped in with crypto-collateralised dollars, but it's a fundamentally different mechanism. Rather than over-collateralising, USDe is delta-neutral: each dollar minted is backed by spot ETH or BTC paired with an equivalent short perpetual futures position on a centralised exchange. The long and short cancel out price exposure, leaving only the funding rate, which has historically been positive (longs paying shorts).
That funding rate is then passed through to sUSDe stakers — yields hit 25-35% during the 2024 bull market, and have ranged 8-15% through early 2026 as funding compressed. By March 2026 sUSDe APY had compressed further to roughly 3.7–5%. Supply hovers around $5.9B after peaking above $14B during the 2025 cycle.
The trade-offs are real:
- Custody risk — collateral lives at OES providers (Copper, Ceffu) and exchange margin accounts, not on-chain.
- Funding-rate risk — if perpetual funding goes deeply negative for sustained periods, USDe's reserve fund absorbs the loss until it doesn't.
- Exchange risk — a Bybit or Binance failure mid-position is not a hypothetical.
USDe is a clever instrument, but calling it "decentralised" stretches the word. It's better understood as a tokenised cash-and-carry trade.
The Other Crypto-Collateralised Dollars
A handful of smaller projects round out the category:
- crvUSD (Curve) — uses a novel LLAMMA soft-liquidation mechanism that gradually converts collateral to crvUSD as price falls, rather than triggering hard liquidations. Around $293–372M in supply.
- GHO (Aave) — minted directly against Aave v3 collateral positions; ~$584M in April 2026, up roughly 245% since the start of 2025. Aave V4 introduces stkGHO via the Umbrella safety module paying ~8.4% APY.
- frxUSD — the December 2024 rebrand of FRAX, now backed by BlackRock's BUIDL fund tokenised by Securitize, blending TradFi reserves with on-chain rails.
- LUSD (Liquity) — the immutable purist's pick, backed only by ETH, no governance, no fees beyond a one-time issuance charge.
- Resolv USR — a delta-neutral design similar to USDe. On 22 March 2026 an exploit minted approximately 80M unbacked USR and extracted ~$25M, depegging the token. The episode is a useful reminder: novel design surface equals novel attack surface.
Why They Still Matter Post-GENIUS
The GENIUS Act, signed in July 2025, gave the US a clear federal framework for fiat-backed payment stablecoins: 1:1 reserves in cash and short Treasuries, monthly attestations, OCC or state oversight, mandatory law-enforcement compliance. USDC and USAT fit into this regime. DAI, USDS, USDe, LUSD — none of them do, by design.
That regulatory gap is a feature, not a bug, for a specific set of users:
- Censorship resistance. USDC has frozen addresses on OFAC orders multiple times. DAI and LUSD cannot. USDS technically can but has never used the function. If your threat model includes a state actor, this matters.
- Permissionless composability. DeFi protocols can integrate crypto-collateralised dollars without KYC pipes or issuer relationships. The whole point of permissionless finance is that nobody gets a veto.
- Yield without intermediaries. SSR and sUSDe pass yield directly to holders without a custodian skimming the spread. USDC's yield, by contrast, accrues entirely to Circle.
- Geographic neutrality. A user in a sanctioned jurisdiction or one without easy USD banking access can mint DAI against ETH. They cannot mint USDC.
The trade-off is that you take on smart-contract risk, oracle risk, and (for some designs) funding-rate or custody risk in exchange for shedding issuer risk. That's the deal. It's been the deal since 2017. The mental model from our stablecoin types overview — fiat-backed, crypto-backed, algorithmic — still holds; the GENIUS Act just sharpened the regulatory contrast between buckets.
A Mental Comparison Table
| Stablecoin | Backing | Yield (Apr 2026) | Approx Supply | Censorship-resistant? |
|---|---|---|---|---|
| USDS | Crypto + RWA, over-collateralised | 3.75–4.5% (SSR) | $7–11B | Partial (freeze function exists) |
| DAI | Crypto + RWA, over-collateralised | DSR (lower) | $3–4B | Yes |
| USDe | Delta-neutral perp hedge | 3.7–5% (sUSDe) | $5.9B | No (CEX custody) |
| LUSD | ETH only, 110% min ratio | None native | $0.28B | Yes (immutable) |
| crvUSD | Crypto, soft-liquidation | Variable | $0.29–0.37B | Yes |
| USDC (for context) | Cash + Treasuries | 0% to holders | $77B | No |
No single design wins on every axis. USDS is the most liquid and most integrated. DAI is the conservative legacy choice. LUSD is the immutable purist's pick. USDe pays the highest yield and carries the highest non-obvious risk. Resolv reminds us that clever new mechanisms break in clever new ways.
In Part 5 we'll close the series by looking at how to actually use these things — picking a chain, sizing exposure, and the practical red-flag checklist for parking value on-chain in 2026.
