Table of Contents
The stablecoin market hit $321 billion in May 2026. USDT and USDC control over 80% of that total. Every one of those coins is supposed to hold a one-dollar peg, but the mechanisms behind that promise vary dramatically in structure, risk, and regulatory treatment.
Three distinct designs dominate the landscape: fiat-backed, crypto-collateralized, and algorithmic. Each uses a different method to maintain price stability. Each carries different failure modes.
Understanding how these systems actually work matters because not all stablecoins are equally stable. The May 2022 collapse of Terra/Luna erased $45 billion in market value within days and demonstrated that mechanism design is not a theoretical concern.
Fiat-Backed Stablecoins: The Reserve Model
Fiat-collateralized stablecoins operate on the simplest premise in crypto: hold one dollar of real-world assets for every stablecoin token issued. Redeem on demand. The peg holds because the reserve is there.
USDC circulates around $76.5 to $78 billion as of May 2026. Its reserves consist of roughly one-third U.S. Treasuries, with the remainder held in overnight repurchase agreements and cash. Gemini's GUSD takes a more conservative approach, backing 100% with bank deposits.
Tether reported approximately $141 billion of direct and indirect U.S. Treasury bill exposure as of March 31, 2026. These reserves are held by regulated financial institutions and typically consist of cash, cash equivalents, and short-term U.S. Treasuries.
The mechanism is straightforward. When you buy USDC through an authorized channel, Circle receives your dollar and mints a new USDC token. When you redeem, Circle burns the token and returns your dollar. The supply expands and contracts based on demand, but the reserve backing never drops below 1:1.
This design eliminates reflexivity risk. There is no feedback loop between the token price and the reserve value. A run on redemptions does not destroy the reserve, it just converts it back to dollars.
The failure mode here is not algorithmic. It is institutional. A fiat-backed stablecoin fails if the issuer mismanages reserves, lies about holdings, or becomes insolvent. It fails if the custodian bank collapses or if regulatory action freezes the reserve.
Under the U.S. GENIUS Act, stablecoin reserves must be one-to-one in cash and short-term U.S. Treasuries with monthly audited disclosures. Comparable full-reserve rules now apply in the EU, UK, Hong Kong, Singapore, Japan, and the UAE.
Fiat-backed stablecoins dominate because reserve transparency works. USD-backed coins hold their value because cash and short-term U.S. Treasuries do not experience the volatility of crypto assets. Regulation has shifted from theoretical to operational across most major jurisdictions.
Crypto-Collateralized Stablecoins: Over-Collateralization as Safety
Crypto-collateralized stablecoins replace bank reserves with on-chain collateral. The most prominent example is DAI, which sits at around $5 to $6 billion in market cap as of April 2026. When you include USDS, the combined figure reaches $8 to $10 billion.
Users generate DAI by locking assets like ETH or wrapped Bitcoin into MakerDAO Vaults. The system requires collateral ratios typically above 150%. If you want to mint $1,000 in DAI, you need to lock at least $1,500 worth of ETH.
This over-collateralization accounts for the price volatility of the backing assets. If ETH drops 20% overnight, your $1,500 collateral might fall to $1,200. The system does not wait for you to respond. If the collateral value falls below a safety threshold, the protocol automatically liquidates enough ETH to cover the outstanding DAI and protect the peg.
The liquidation mechanism is what separates crypto-collateralized designs from algorithmic ones. There is no reliance on future user behavior or market sentiment. The collateral is there, locked in a smart contract, and the liquidation triggers are programmatic.
DAI has survived through a decade of iteration and conservative liquidation mechanics. It weathered the March 2020 crash, multiple bear markets, and shifting collateral compositions. The design works because it does not depend on confidence. It depends on math.
The failure modes are different from fiat-backed models. Smart contract bugs can drain collateral or freeze liquidations. Oracle failures can feed incorrect price data, triggering premature liquidations or failing to liquidate when needed. Liquidation cascades can occur if collateral becomes illiquid during extreme market stress.
In March 2026, the stablecoin ResolvUSD was exploited for approximately $80 million and depegged to as low as $0.14 before its market cap fell 55.9%. Protocol vulnerabilities remain a real risk even in otherwise sound designs.
Crypto-collateralized stablecoins occupy a middle ground. They are more capital-efficient than holding raw collateral but less capital-efficient than fiat-backed models. They are more transparent than custodial reserves but more complex to audit. They are permissionless but carry smart contract risk.
Algorithmic Stablecoins: The Reflexive Failure Mode
An algorithmic stablecoin holds no hard reserves. Instead, it relies on a two-token arbitrage loop to maintain the peg. When the stablecoin trades below $1, the protocol allows users to burn the stablecoin and mint the sister token at a profitable rate. When it trades above $1, users can burn the sister token and mint the stablecoin.
In theory, arbitrageurs keep the peg stable by exploiting these pricing inefficiencies. In practice, this design is vulnerable to reflexive death spirals.
Terra/Luna is the worked example of what happens when the mechanism becomes reflexive. UST was backed by LUNA, not fiat. When UST depegged slightly in May 2022, redemption mechanics required LUNA to absorb losses.
As LUNA's price fell, its redemption capacity weakened, further undermining UST confidence. Anchor Protocol, the primary lending platform supporting UST, offered 18% yields and saw its total value locked fall from $18 billion to under $1 billion. Users only stayed if they believed the peg would hold. Once confidence broke, the arbitrage loop inverted.
At the moment when LUNA's market cap fell below UST's market cap, the value of LUNA was mathematically insufficient to redeem all the UST in circulation. UST was no longer a stablecoin. The price of UST collapsed to $0.10. LUNA fell from $119 to nearly zero.
This was not a liquidity crisis or a bank run. It was a structural failure. The mechanism that was supposed to defend the peg became the mechanism that accelerated its collapse. Selling pressure triggered more LUNA issuance, which triggered more selling pressure. The feedback loop was self-fulfilling.
Algorithmic stablecoins currently account for less than 2% of total stablecoin capitalization. After peaking at over $22 billion in early 2022, their value collapsed to under $3 billion following the failure of TerraUSD and LUNA.
The EU's MiCA framework prohibits purely algorithmic stablecoins outright. The U.S. regulatory environment has moved toward similar restrictions. Hybrid models like FRAX v2 and Ethena's delta-neutral USDe exist but remain marginal.
Pure algorithmic stablecoins are now a regulatory dead letter in most major markets. The theoretical elegance of a reserve-free peg could not survive empirical contact with reflexive market dynamics.
Comparison of Mechanisms
| Type | Example | Collateral | Reflexivity Risk | Regulatory Status | Primary Failure Mode |
|---|---|---|---|---|---|
| Fiat-backed | USDT, USDC | USD cash, T-bills | None | Compliant (US/EU) | Issuer insolvency or reserve mismanagement |
| Crypto-collateralized | DAI | ETH, WBTC, USDC | Low (liquidations contain risk) | Permitted | Smart contract bug or liquidation cascade |
| Algorithmic | UST (defunct) | LUNA (failed) | Extreme (self-fulfilling collapse) | Banned or restricted | Death spiral when confidence breaks |
What This Means for Users
Most retail users do not pay direct fees to hold stablecoins. Issuers profit from the float, the spread between what they earn on reserves and what they pay holders. Network gas fees, foreign exchange spreads on fiat off-ramps, payout fees, KYC tier restrictions, and chain support each move the all-in cost by 30 to 200 basis points.
The choice of stablecoin matters less for holding costs than for tail risk. A fiat-backed stablecoin can lose its peg if the issuer becomes insolvent. A crypto-collateralized stablecoin can depeg if the smart contract is exploited or if collateral becomes illiquid during a crash. An algorithmic stablecoin can enter a death spiral if confidence falters.
Fiat-backed stablecoins are now the default for payments, liquidity provisioning, and yield-bearing instruments. They are the only design that has regulatory clarity across major jurisdictions. They are the only design that survived the 2022 collapse cycle without systemic failure.
Crypto-collateralized stablecoins serve a different function. They are permissionless, auditable on-chain, and do not depend on custodial trust. DAI remains relevant because it does not require KYC, does not rely on a single issuer, and can be generated by anyone with sufficient collateral.
Algorithmic stablecoins remain a niche experiment. Hybrid designs with partial collateralization exist, but they have not achieved meaningful scale. The regulatory environment has moved decisively against pure algorithmic models, and the market has followed.
Edge Cases and Ongoing Risks
Even fiat-backed stablecoins carry risks that are not immediately obvious. Reserve illiquidity is one. Most issuers hold the majority of reserves in short-term U.S. Treasuries. During periods of stress, rapid redemption runs could force fire sales of those Treasuries, creating feedback loops between the stablecoin market and traditional financial markets.
There is a risk of feedback loops, such as a stablecoin issuer moving billions in and out of Treasuries or banks during periods of stress. This could destabilize both markets simultaneously.
Oracle failure remains a risk for crypto-collateralized designs. These systems rely on external price feeds to determine when collateral should be liquidated. Manipulation or outages can disrupt liquidation triggers, leaving the system under-collateralized or liquidating positions prematurely.
Regulatory shock is another potential failure mode. Sudden restrictions on reserve composition, such as limits on Treasury bill holdings or requirements for higher cash ratios, could force issuers to redesign their models midstream. This has not happened yet, but the regulatory framework is still evolving.
Smart contract risk persists across all on-chain stablecoin designs. The ResolvUSD exploit in March 2026 demonstrated that even relatively mature protocols can suffer critical vulnerabilities. No amount of over-collateralization protects against a bug that drains the vault.
The Takeaway
Stablecoins maintain their peg through three distinct mechanisms. Fiat-backed models rely on custodial reserves and regulatory compliance. Crypto-collateralized models use over-collateralization and automatic liquidation. Algorithmic models attempt to maintain stability through arbitrage loops but are vulnerable to reflexive collapse.
Terra/Luna validated the theoretical concerns about algorithmic designs. When confidence broke, the mechanism designed to defend the peg accelerated its destruction. Pure algorithmic stablecoins are now banned or restricted in most major markets.
Fiat-backed stablecoins dominate because reserve transparency works and regulation has caught up. Crypto-collateralized stablecoins survive because they are permissionless and do not depend on custodial trust. Algorithmic stablecoins remain marginal because the reflexivity risk is structural, not fixable through iteration.
The choice of stablecoin is a choice of failure mode. Each design trades one set of risks for another. None are risk-free, but the risks are now well-documented and empirically tested. For more on the differences between the two largest fiat-backed stablecoins, see the peer-reviewed analysis of Terra-Luna failure mechanics, which provides detailed network dependency modeling of reflexive collapse dynamics.
Frequently Asked Questions
What keeps a stablecoin pegged to one dollar?
Three mechanisms maintain the peg: fiat-backed stablecoins hold dollar reserves in banks and Treasuries with 1:1 redemption rights; crypto-collateralized stablecoins require over 150% collateral and automatically liquidate if values drop; algorithmic stablecoins use arbitrage loops between two tokens but carry extreme reflexivity risk, as demonstrated by Terra/Luna's $45 billion collapse in May 2022.
Why did Terra/Luna collapse if it was algorithmic?
Terra/Luna collapsed because its two-token arbitrage mechanism became reflexive. When UST depegged slightly, LUNA had to absorb losses. As LUNA's price fell, its ability to back UST weakened, which further undermined confidence. Once LUNA's market cap fell below UST's, the system could not mathematically redeem all outstanding tokens. The selling pressure became self-fulfilling, and both tokens collapsed to near-zero within days.
Are fiat-backed stablecoins safer than crypto-collateralized ones?
Fiat-backed stablecoins eliminate reflexivity risk because reserves do not fluctuate with token price, but they require trust in the issuer and custodian. Crypto-collateralized stablecoins like DAI are permissionless and auditable on-chain but carry smart contract risk and require over-collateralization. The ResolvUSD exploit in March 2026 showed that protocol vulnerabilities remain a real threat. Each design trades one set of risks for another.
How much collateral does DAI require to stay stable?
DAI typically requires collateral ratios above 150%. To mint $1,000 in DAI, you must lock at least $1,500 worth of assets like ETH or wrapped Bitcoin. This over-collateralization absorbs price volatility. If your collateral value falls below the safety threshold, the MakerDAO protocol automatically liquidates enough collateral to protect the peg. This mechanism has allowed DAI to survive multiple bear markets and the March 2020 crash.
What regulations apply to stablecoins now?
The U.S. GENIUS Act requires stablecoin reserves to be one-to-one in cash and short-term U.S. Treasuries with monthly audited disclosures. The EU, UK, Hong Kong, Singapore, Japan, and UAE have comparable full-reserve rules. The EU's MiCA framework prohibits purely algorithmic stablecoins outright. Fiat-backed stablecoins now operate under clear regulatory frameworks in most major markets, while algorithmic designs face bans or heavy restrictions.
