The Dollar You Mint: How CDP Stablecoins Work in 2026

When most people picture borrowing in decentralized finance, they picture a pool. You deposit a token someone else supplied, you draw out a different token, and an interest-rate curve sets the price. That is how Aave, Morpho and Compound work, and in 2026 it is a large business: Aave alone crossed $30 billion in deposits over the summer, and its token has rallied hard, trading near $180 in early October after a roughly 19% week, according to CoinGecko.

There is a second kind of on-chain credit that does not borrow a dollar at all. It mints one. Lock up collateral, and the protocol creates a brand-new stablecoin out of nothing, hands it to you as debt, and destroys it when you pay back. No depositor funded the loan. You are both the borrower and, in a sense, the issuer. These are collateralized debt positions, or CDPs, and the dollars they produce (DAI, USDS, GHO, crvUSD and BOLD, among others) sit at the odd intersection of lending and money printing.

The category is small next to the giants. The whole stablecoin market is worth roughly $311 billion, with Tether’s USDT and Circle’s USDC together holding about 83% of supply, per DefiLlama. Crypto-collateralized stablecoins add up to only around $15 billion. Yet this corner is where the most interesting monetary engineering of the cycle is happening, and where a US stablecoin law signed in 2025 is quietly drawing a line that these dollars fall on the wrong side of. This piece is part of our ongoing defi-on-chain series, and it picks up where the broader lending coverage leaves off.

Borrowing a Dollar vs Minting One

Start with the mechanical difference, because everything else follows from it. In a peer-to-pool market, liquidity is scarce. Suppliers deposit USDC, borrowers take it out, and the two sides meet on a utilization curve: the more of the pool that is borrowed, the higher the rate climbs. If nobody supplies, nobody can borrow. The protocol is a matchmaker.

A CDP has no supply side. When you open a position and mint DAI, no other user lent you that DAI; the Sky protocol issued it against your collateral and will burn it when you repay. The protocol behaves less like a matchmaker and more like a central bank with one very strict rule: every dollar in circulation must be backed by more than a dollar of collateral locked in a smart contract. That is why CDP stablecoins are always overcollateralized. To mint $100 you post something like $150 or more of volatile crypto, a cushion that lets the system sell your collateral and still make the dollar whole if prices fall.

The upshot is that a CDP is a loan to yourself, denominated in a dollar the protocol manufactures on demand. The interest you pay is not compensation to a lender; it is a monetary-policy fee the protocol uses to manage how many of its dollars exist. Repay the debt and the minted dollars vanish from supply. Default, and the protocol liquidates your collateral to buy them back. Understanding that single inversion (you are not taking someone’s dollars, you are issuing new ones) explains every design choice that follows.

Why mint at all, rather than just sell? Three reasons drive the demand. The first is leverage: mint a dollar against ETH, buy more ETH, and repeat, a loop that multiplies exposure (and risk) without a centralized broker. The second is liquidity without a sale, the ability to spend dollars while keeping your collateral and deferring a taxable event. The third is more subtle: a decentralized dollar that no single company can freeze or seize is useful as base money for the rest of DeFi, which is why DAI spent years as the collateral and trading pair of choice across the ecosystem. Minting is not a fringe activity; it is how a large slice of on-chain dollars come into existence in the first place.

How a CDP Actually Works

The lifecycle is the same across every protocol, even though the names differ (Sky calls them vaults, Liquity calls them troves). You deposit collateral, usually ETH, a staked-ETH token, a blue-chip asset or even another stablecoin. You mint the protocol’s stablecoin up to a cap set by the minimum collateralization ratio. You pay a stability fee on the outstanding debt for as long as it exists. And you watch one number: the ratio of your collateral’s market value to your debt. Let it fall to the minimum and a liquidator steps in.

A worked example makes the risk concrete. Say you lock $10,000 of ETH into a vault with a 150% minimum collateralization ratio. The table below shows what happens as you mint more aggressively. The more you mint, the thinner your cushion, and the smaller the price drop that triggers liquidation. These figures are illustrative, not live market data, but the arithmetic is exactly what the contracts enforce.

Notice there is no fixed repayment date and no lender to renegotiate with. You can hold the position for an hour or a year; the only clock that matters is the price of your collateral against your debt. That freedom is the appeal. A long-term ETH holder can mint dollars to spend without selling the ETH, keeping the upside and deferring the tax event. The danger is that the same position liquidates automatically, without warning, the moment the market turns against it.

The 2026 Market Map

Five CDP stablecoins matter in 2026, each with its own design philosophy. The table pulls them together; the sections that follow unpack each one. Supply figures come from CoinGecko and DefiLlama and move daily, so treat them as a snapshot rather than a fixed truth.

A few patterns jump out. Sky’s USDS and DAI dwarf the rest; together they are roughly $11 billion, most of the crypto-collateralized category on their own. The others are boutique by comparison. They also disagree on nearly everything else: who sets the interest rate, whether governance can touch the contracts, how liquidation is handled, and what counts as acceptable collateral. Those disagreements are the whole story.

Sky: DAI, USDS and the Original CDP

Every CDP stablecoin is a descendant of DAI. MakerDAO launched it in 2017 as the first decentralized dollar minted against crypto collateral, and the mechanics that every rival now copies (vaults, stability fees, liquidation auctions, a savings rate) were invented or popularized there. In 2024 MakerDAO rebranded to Sky, introduced USDS as a parallel upgrade of DAI, convertible one for one, and swapped its MKR governance token for SKY. DAI did not disappear; it still carries around $4.3 billion in supply, but the growth has moved to USDS.

USDS is now the third-largest dollar stablecoin of any kind, behind only USDT and USDC, at roughly $6.7 billion as of late September after peaking near $8.95 billion on 1 April 2026, per Eco. Sky is also a real business. The protocol reported about $338 million in annualized revenue and $158 million in net profit through a recent stretch of volatility, with institutions, in the words of cofounder Rune Christensen, increasingly “obsessed” with stablecoins, per Sandmark.

Two levers keep USDS where Sky wants it. The Peg Stability Module lets anyone swap USDC for USDS one for one at zero fee, which pins the price to its reserve currency and quietly turns part of the backing into plain USDC. The Sky Savings Rate, currently 3.60% on the staked sUSDS token, is the demand lever: raise it and holders park dollars to earn yield, shrinking free-floating supply; cut it and they wander off. Christensen has made the case for yield bluntly, telling reporters that more than $300 billion of stablecoins “earn no yield” at all, a gap Sky built its product around, per The Block.

What actually backs those billions is worth pausing on, because it is the opposite of pure crypto. Alongside ETH and staked-ETH vaults, Sky holds a large reserve of USDC and tokenized US Treasuries, deployed through its Spark lending arm and real-world-asset allocations. That is how the protocol funds a savings rate at all: borrower stability fees plus the yield on short-term government debt. It also means USDS is a hybrid, part crypto-collateralized CDP dollar and part Treasury-backed fund, which is precisely the ambiguity regulators are now trying to resolve.

On the back end, surplus revenue feeds a Smart Burn Engine that buys SKY on the open market, a value-accrual design that echoes Aave’s. SKY itself trades around $0.095, a market value near $2.2 billion, per CoinGecko. The irony Christensen would appreciate: the original decentralized dollar now runs much of its reserve through USDC and short-term Treasuries, which is exactly why the new US stablecoin law treats it as a problem. More on that below.

crvUSD and the Art of the Soft Liquidation

Curve’s crvUSD is the engineer’s CDP. Where Sky and Aave liquidate a troubled position in a discrete event, selling collateral at a penalty, crvUSD tries to avoid a single moment of liquidation at all. Its collateral sits inside a mechanism called LLAMMA (a lending-liquidating automated market maker) that behaves like a specialized pool. As the collateral price falls through a series of bands, LLAMMA gradually converts it into crvUSD; as the price recovers, it converts back. Instead of one brutal sale, you get a continuous, reversible wind-down. A borrower who gets soft liquidated and then sees ETH bounce can end up roughly whole, the opposite of the all-or-nothing experience on older designs. The tradeoff is drag: while you are in the soft-liquidation zone, you are effectively selling low and buying high in small increments.

crvUSD is modest in size, around $228 million in supply, per CoinGecko. The borrow rate recently sat near 2.1%, among the cheapest dollars in DeFi, with PegKeeper contracts holding reserves to defend the peg by minting or burning into balanced pools, per Curve’s own weekly metrics. Savers can hold scrvUSD, a vault that pays a governance-set slice of borrower interest; that yield was running around 1.4%, with the remainder flowing to the Curve DAO. The CRV governance token trades near $0.38 for a market value just under $600 million, per CoinGecko.

Founder Michael Egorov has spent 2026 pushing a careful-expansion line. He has warned that scaling crvUSD too fast amplifies peg risk when Bitcoin swings, arguing the system’s capacity has to grow in measured steps, per Bitget News. His instinct on failure is just as telling. When crvUSD carried roughly $700,000 of bad debt in April 2026, Egorov pitched a market-based fix, letting the shortfall be auctioned and absorbed rather than papered over, an explicit contrast with how larger lenders socialize losses, per CoinDesk. That split (soak losses up by rule, or backstop them by treasury) runs right through the CDP world.

GHO: Aave’s Native Dollar and the Facilitators

GHO is the newest of the big CDP dollars and the most tightly woven into an existing lending market. You mint GHO against collateral you have already supplied to Aave, so a user running an Aave position can borrow the protocol’s own stablecoin without unwinding anything. Unlike a normal Aave borrow, the GHO you receive is freshly minted rather than drawn from a pool, which makes it a true CDP, and the interest you pay flows straight to the Aave DAO rather than to suppliers.

For Aave the appeal is margin. When a user borrows USDC on Aave, most of the interest goes to the suppliers who funded it; when the same user mints GHO, there is no supplier to pay, so the entire stability fee accrues to the protocol. GHO turns Aave from a pure marketplace into an issuer that earns the full spread on its own dollar, the same logic that makes a bank want to issue deposits rather than broker someone else’s. That is why a lending protocol with tens of billions in deposits bothered to launch a stablecoin at all.

The clever part is the facilitator model. Any contract allowed to create GHO must be approved by Aave governance and given a minting bucket with a hard cap. The current facilitators include the Aave V3 market itself, a cross-chain module that locks GHO on Ethereum and mints it on other networks through Chainlink’s CCIP, a flash-mint facilitator, and a GHO Stability Module (GSM) that swaps GHO against USDC and USDT to defend the peg. That architecture lets Aave scale GHO across chains while keeping a governance throttle on every source of supply, though it also inherits cross-chain messaging risk, the same surface that has broken other cross-chain systems this year.

GHO supply sits around $698 million, close to its all-time high near $699 million set in August, per DefiLlama. It has a persistent quirk: it tends to trade a hair below a dollar. In mid-September it sat about 12 basis points under $1.00, roughly ten below USDC, which is why Aave’s Gho Stewards spend so much time tuning parameters. Their September 2026 update raised the Ethereum base borrow rate from 3.00% to 3.25% and adjusted GSM redemption fees to nudge the peg back up, per the Aave governance forum. Holders who stake into GHO through Aave’s Umbrella safety module earn around 8.4% in stkGHO, but they also take first-loss slashing risk if the protocol runs a deficit, per Aave.

Liquity: The Immutable Dollar Meets User-Set Rates

If crvUSD is the engineer’s CDP, Liquity is the decentralization maximalist’s. Its first version, launched in 2021, mints LUSD against ETH only, charges 0% ongoing interest (just a one-time borrowing fee), and allows a minimum collateralization ratio of 110%, the thinnest in the category. The radical part is governance: there is none. The contracts are immutable, with no admin keys, no DAO that can change parameters, and no multisig that can pause the system. Nobody can touch LUSD, which is the entire point. LUSD is a legacy product now, around $29 million, but it remains the reference design for anyone who thinks a stablecoin should be a piece of infrastructure, not a company.

Liquity V2 and its BOLD stablecoin, which went live in early 2025, keep the immutability and the 110% floor but add one genuinely new idea: user-set interest rates. Instead of governance or a curve deciding your borrow cost, you choose it yourself when you open a trove, and you can change it later, per Liquity. The rate you pick determines your place in the redemption queue, the clever twist explained in the next section. Three quarters of BOLD revenue is routed to a Stability Pool where depositors earn yield and backstop liquidations, per Liquity.

BOLD is still small, about $38 million in supply trading just under a dollar, with Liquity V2 total value locked around $91 million and growing, per DefiLlama. The model has also spawned more than a dozen friendly forks, protocols that copy the open-source V2 code and share liquidity incentives. Whether user-set rates beat a governed stability fee is still an open question, but Liquity has at least proven that a CDP can run with no one at the wheel.

The Stability Fee Is Monetary Policy, Not a Utilization Curve

Here is the mechanical idea that separates CDPs from peer-to-pool lenders most sharply. On Aave or Morpho, the borrow rate is discovered: it rises and falls automatically with utilization, climbing steeply past a kink point when the pool runs low on liquidity. The market sets it block by block. On a CDP, there is no pool to run dry, so utilization means nothing. The stability fee is instead a deliberate policy choice, a dial the issuer turns to control how much of its dollar exists.

The logic is the one a central bank uses. If the stablecoin trades above a dollar, demand is outrunning supply, so the issuer cuts the stability fee to make minting cheaper and coax more dollars into circulation. If it trades below a dollar, supply is too high, so the issuer raises the fee to make carrying debt painful and encourage repayment, which burns dollars. Sky’s governance votes on this directly. Aave’s Gho Stewards do it through parameter updates. crvUSD leans on an automated monetary-policy contract plus its PegKeepers. Liquity V2 outsources the whole decision to borrowers and lets the redemption mechanism sort out the rest.

The savings rate is the mirror image. Where the stability fee manages the supply of dollars, the savings rate (Sky’s 3.60%, Curve’s scrvUSD, Aave’s stkGHO) manages demand by paying people to hold the dollar rather than spend it. Together the two form a crude but real interest-rate policy. It is why reading a CDP stablecoin feels less like reading a loan and more like reading a tiny economy, complete with a target, a lever on each side, and a governance body arguing about where to set them.

How a Minted Dollar Holds Its Peg

A fiat-backed stablecoin holds its peg the boring way: the issuer promises to redeem one token for one real dollar, and arbitrage does the rest. A CDP dollar has no such promise, because there is no central issuer sitting on a pile of cash. It holds its peg through a stack of mechanisms, and understanding them is the difference between trusting one of these dollars and gambling on it.

The floor under the price is overcollateralization plus liquidation. Because every dollar is backed by more than a dollar of collateral that the protocol will sell if needed, the market believes the dollar can always be made whole, which is most of why it trades near par. On top of that floor sit protocol-specific tools. Sky runs the Peg Stability Module, a USDC swap that hard-pins USDS to its reserve currency. crvUSD runs PegKeepers that algorithmically expand or contract supply. Liquity runs redemptions: if BOLD or LUSD trades below a dollar, anyone can buy it cheap and redeem it with the protocol for exactly a dollar of collateral, an arbitrage that mechanically pushes the price back up. GHO leans on its Stability Module and governance tuning, which is why it needs constant attention to shave off that stubborn discount.

Redemptions are where Liquity V2’s user-set rates pay off. Redeemers are routed to the troves paying the lowest interest rate first, not the riskiest ones. So a borrower who sets a rock-bottom rate to save money accepts a higher chance of being redeemed against (having debt repaid and collateral taken at par), while a borrower who pays up buys protection. The interest rate becomes a bid for safety. It is an elegant way to let a market, rather than a committee, decide who gets redeemed when the dollar slips.

When CDPs Break: Oracles, Liquidations and Bad Debt

Every one of these mechanisms depends on two things working: an accurate price feed and a willing liquidator. When either fails, a CDP can leave dollars in circulation that are no longer fully backed, and the system takes on bad debt. The canonical cautionary tale is MakerDAO’s Black Thursday in March 2020, when a violent ETH crash and network congestion left some liquidation auctions won with bids of almost zero. The system ended up undercollateralized and had to mint and sell MKR to recapitalize. Every CDP design since has been, in part, an answer to that day.

The oracle is the sharpest edge. A CDP liquidation fires off a reported price, so if that price can be manipulated or lags the real market, liquidations trigger wrongly or not at all. 2026 has been a banner year for this: researchers counted a record 32 price-manipulation exploits across DeFi lending, per The Cryptonomist, and oracle manipulation has gone multi-chain. CDPs that use robust time-weighted feeds and conservative collateral tend to survive; those that trust a thin spot price do not.

Governance is the other quiet risk, and it cuts both ways. A protocol that can vote to change collateral types, stability fees or liquidation parameters can also be captured: Sky’s biggest decisions have at times hinged on a handful of large token holders, and a governed system is only as safe as the people holding the keys. Liquity’s answer is to remove the keys entirely, which makes it impossible to capture but also impossible to rescue if a parameter turns out to be wrong. There is no free lunch here, only a choice about who you would rather trust: a changeable committee or unchangeable code.

When bad debt does appear, the CDP world splits along the philosophical line Egorov drew. One camp socializes the loss through a backstop: Sky has its surplus buffer and can mint SKY in extremis, and Aave has run full bailouts, most dramatically the “DeFi United” effort led by Aave founder Stani Kulechov that raised about 132,650 ETH (roughly $303 million) to cover bad debt after a bridge exploit poured borrowed losses onto the protocol, per Bankless. The other camp, Curve and Liquity, prefers to let the shortfall be absorbed by rule, through auctions, stability pools and redemptions, with no treasury check written. Neither approach is obviously right. A backstop protects users but concentrates discretion in a few hands; a market-based cleanup is credibly neutral but can leave someone holding the loss.

Yield on the Mint: Savings Rates Against a Hiking Fed

The savings rates attached to CDP dollars do not float free. They are measured against the one number every dollar yield now answers to: the risk-free rate set by the Federal Reserve. And that number just moved the wrong way for DeFi. On 16 September 2026 the Fed raised its target range by a quarter point to 3.75% to 4.00%, its first hike since 2023, per the Federal Reserve. The effective funds rate now trades near 3.88% and SOFR near 3.85%, per the Fed’s H.15 release, with futures pricing further increases into 2027 and the next decision due 28 October.

That reset matters because on-chain savings rates have spent two years compressing toward the base rate. Sky’s savings rate was above 8% in 2024; it is 3.60% now. scrvUSD pays around 1.4%. Only stkGHO, at roughly 8.4%, clears a Treasury bill, and it does so by taking slashing risk, not by magic. The table lines the key rates up against the Fed.

The strategic point is unforgiving. When a short-term Treasury pays close to 4% with no smart-contract, oracle or governance risk, a CDP savings rate below that is asking you to take real risk for a negative spread. It is the same math that melted the restaking premium this year. It also explains the structural reason the yield sits on a wrapper (sUSDS, scrvUSD, stkGHO) rather than on the base coin, a design point that turns out to be a legal necessity, not just a convenience.

The GENIUS Act and the SEC: Is a Minted Dollar Even Legal?

Here is where the story turns, for US readers especially. In July 2025 President Trump signed the GENIUS Act, the first federal framework for stablecoins, per Congress. It creates a clean category called a “payment stablecoin”: a token backed one for one by high-quality liquid assets (cash and short-term Treasuries), redeemable at par, with reserves disclosed monthly and examined by an accounting firm. Compliant issuers like Circle fit neatly. The catch for everything in this article is in the definition of what qualifies.

A dollar backed by volatile crypto, by overcollateralization, or by an algorithm has no path to becoming a payment stablecoin under the law, per analysis from the law firm Paul Hastings. DAI, USDS, crvUSD, GHO and BOLD are, by construction, exactly the kind of dollar the statute does not bless. They are not illegal; they simply sit outside the payment-stablecoin perimeter, in a category with reduced formal access to US users and institutions. The law also bars payment-stablecoin issuers from paying interest to holders, the US echo of the same rule in Europe, and another reason yield lives on sUSDS rather than USDS.

So what are they, legally? Probably not securities. SEC Chair Paul Atkins has spent 2026 pulling the agency back from the enforcement posture of the prior era, framing self-custody as a “core American value” and pushing an innovation exemption for on-chain products, per the SEC. A minted, overcollateralized dollar with no issuer promising a profit is a poor fit for the securities laws. But the GENIUS Act and the SEC together leave CDP dollars in a gap: not payment stablecoins, not obviously securities, and governed in practice by code and DAO votes rather than a regulator. Whether that gap is a loophole or a dead end is now a political question, tangled up with the midterm fight over crypto policy. Sky, for its part, argues its decentralized architecture simply does not map onto the centralized model the law was written for.

How to Read a CDP Before You Mint

Because ETH and Bitcoin are the dominant collateral, the health of every CDP position ultimately rides on crypto prices, and the mood there has firmed as Bitcoin pushed into Uptober on a friendlier macro backdrop. A rising market widens everyone’s cushion; a sharp drop tests every mechanism in this article at once. Before you lock collateral and mint, a short checklist separates the durable dollars from the fragile ones.

• Collateral quality. Is the backing blue-chip (ETH, staked ETH, Bitcoin) or long-tail tokens that can gap to zero? Thin collateral is where soft liquidation and auctions both fail.
• Your buffer, not the minimum. A 110% or 150% minimum is the liquidation line, not a target. Mint well below it so a routine 20% drop does not wipe you out.
• Liquidation style. Discrete auctions (Sky, Aave) punish you at once; LLAMMA soft liquidation (crvUSD) bleeds you slowly but reversibly; stability pools and redemptions (Liquity) are different again. Know which one you have signed up for.
• Oracle type. Robust, time-weighted feeds survive manipulation; thin spot prices do not. The oracle is the single most common point of failure.
• Who sets the stability fee. A DAO vote, an automated contract, or you? Governed rates can change under you; immutable ones cannot be rescued if the design is wrong.
• Admin keys and immutability. Can anyone pause, upgrade or drain the system? Liquity cannot; most others can. That is a tradeoff between safety and agility, not a free lunch.
• Peg history and bad-debt record. Has the dollar held par through a real crash? Has it taken bad debt, and if so, who paid?
• Savings-rate sustainability. Is the yield funded by real borrower demand or by emissions that will fade? Against a 4% Treasury, a subsidized rate is a countdown, not an edge.

None of this requires trusting a brand. It requires reading the contracts and the governance, because in a CDP there is no bank, no deposit insurance and no one to call. The dollar is only as sound as the collateral behind it and the code that liquidates it. That is the deal you accept the moment you mint one.

Frequently Asked Questions

What is a CDP stablecoin?

A collateralized debt position stablecoin is a dollar token you mint by locking up more than a dollar of crypto collateral in a smart contract. Instead of borrowing an existing coin from a pool, you create a new one as debt, then repay to burn it and unlock your collateral. DAI, USDS, GHO, crvUSD and BOLD are the main examples in 2026.

How is minting DAI or GHO different from borrowing on Aave?

On Aave or Morpho you borrow tokens that other users supplied, and a utilization curve sets the rate. With a CDP there is no supplier; the protocol mints a brand-new stablecoin against your collateral and sets the stability fee as a policy lever. You are both the borrower and, in effect, the issuer of the dollar.

What keeps a CDP stablecoin pegged to one dollar?

Overcollateralization and liquidation put a floor under the price, so the market trusts the dollar can be made whole. On top of that, protocols run peg tools: Sky’s USDC swap module, Curve’s PegKeepers, Liquity’s redemptions and Aave’s GHO Stability Module. If the price slips, these mechanisms expand or contract supply to push it back toward a dollar.

Are CDP stablecoins legal under the GENIUS Act?

They are not banned, but they do not qualify as payment stablecoins under the 2025 GENIUS Act, which requires one-to-one backing by cash and Treasuries. Because DAI, USDS, GHO, crvUSD and BOLD are backed by crypto and overcollateralization, they sit outside that category, with reduced formal access to US users and no path to the payment-stablecoin label.

Can you earn yield on a CDP stablecoin?

Yes, but the yield sits on a wrapped version, not the base coin. Sky pays about 3.60% on sUSDS, Curve pays a share of borrower interest on scrvUSD, and Aave offers around 8.4% on stkGHO in exchange for slashing risk. With the Fed’s risk-free rate near 4%, any CDP savings rate below that means taking real risk for a thin or negative spread.

By Yuki Tanaka, senior DeFi correspondent at HOGE Wire. Figures verified as of 5 October 2026; this is analysis, not financial advice.