Blockchain
Stablecoins: Algorithmic and Collateralized
In May 2022, $40 billion vanished in 5 days. The stablecoin UST, which was supposed to always be worth exactly $1, fell to two cents, dragging LUNA from $80 to zero with it. Meanwhile, another stablecoin, DAI, operating on a completely different model, calmly survived the collapse without even trembling. Why did two systems solving the same problem - pegging to the dollar - produce diametrically opposite results? The answer lies in three fundamentally different approaches to creating stability: fiat reserves, crypto over-collateralization, and pure algorithmic mechanics.
- **USDT (Tether)** - the largest stablecoin ($83B), with more trading volume than Bitcoin itself. Yet Tether avoided a proper audit for years and paid a $41M fine to the CFTC for false statements about reserves
- **The UST/LUNA collapse** (May 2022) destroyed $40B and became the largest financial catastrophe in cryptocurrency history. Consequences: founder Do Kwon arrested, SEC tightened regulation, BUSD ($23B) forcibly shut down
- **MakerDAO** created DAI - a stablecoin that has operated without a central issuer for 7 years, survived every market crisis, and accepts not only crypto but also $2B+ in US Treasuries as collateral
Предварительные знания
Fiat-backed: USDT, USDC, and centralized reserves
Cryptocurrencies are volatile: ETH can drop 20% in a day. For everyday payments, locking in profits, or use in DeFi, a stable asset is needed. A **stablecoin** is a token pegged to the value of an external asset (usually the US dollar). The stablecoin market exceeds $130B - it is the circulatory system of DeFi, through which the majority of trading volume flows.
The simplest model is **fiat-backed**. An issuing company holds dollars (or equivalents) in bank accounts. The user sends $1 to the company's account and receives 1 stablecoin. To get the dollar back - they burn the stablecoin and receive a dollar in return. Every token must be backed by a real dollar in reserves.
**USDC depeg in March 2023.** Circle held $3.3B of its reserves in Silicon Valley Bank (SVB). When SVB failed on March 10, USDC fell to $0.87 - panic: what if the $3.3B was lost? DeFi protocols liquidated positions with USDC collateral. Over the weekend, the Fed guaranteed SVB deposits, and on Monday USDC returned to $1.00. Lesson: even the "most reliable" stablecoin depends on the traditional banking system.
**Key risks of the fiat-backed model:** 1. **Custodial risk** - you must trust the issuer that reserves are real. Tether avoided a full audit for years and paid a $41M CFTC fine in 2021 for false statements about reserves. 2. **Regulatory risk** - the SEC can halt issuance (as with BUSD). 3. **Bank run** - if everyone tries to redeem at once, the issuer may not be able to sell illiquid assets fast enough. 4. **Censorship** - Tether and Circle can freeze addresses at law enforcement request (over $1B frozen).
USDC fell to $0.87 in March 2023 due to SVB's bankruptcy. What exactly happened to the reserves?
Crypto-backed: collateral in cryptocurrency and over-collateralization
Fiat-backed stablecoins solve the stability problem but create a new one: **centralization**. Tether can freeze your address; Circle depends on banks. Crypto-backed stablecoins attempt to create a stable dollar **without trusting a central issuer** - only smart contracts and cryptocurrency collateral.
The main challenge: if the collateral is ETH, and ETH can drop 40% in a week, how do you ensure stability? The answer: **over-collateralization**. To receive $100 in stablecoins, you must deposit $150–200 in ETH. The 50–100% buffer covers collateral volatility. If the collateral price falls below a threshold - the position is **liquidated**: collateral is sold on the market and the debt is repaid.
**Why does Liquity (LUSD) require only 110% collateral?** The key innovation is the **Stability Pool**. LUSD holders deposit tokens into the pool in advance. When a liquidation occurs, the debt is instantly repaid from the pool (not through auctions, as in MakerDAO). In return, Stability Pool participants receive the liquidated ETH collateral at a discount. Fast liquidation = smaller buffer. Downside: LUSD only accepts ETH as collateral and has no governance - contracts cannot be updated.
**Key advantage** of crypto-backed over fiat-backed: transparency and censorship resistance. All collateral is visible on-chain, liquidations are automatic, code is open source. Nobody can freeze DAI at your address (unlike USDC/USDT). **Key downside**: capital inefficiency. To receive $100 in stablecoins, you must lock up $150–200 in cryptocurrency. This is "dead" capital that is not working.
A user deposited 10 ETH at a price of $3,000 (total $30,000) and borrowed 15,000 DAI. The minimum collateral ratio is 150%. At what ETH price will liquidation begin?
Algorithmic stablecoins: UST, FRAX, and the death spiral
Fiat-backed stablecoins are centralized. Crypto-backed stablecoins are capital-inefficient (you need $150 of collateral for $100 of debt). Algorithmic stablecoins promised to solve both problems: stability **without collateral or with minimal collateral**, using only math, economic incentives, and arbitrage. This turned out to be the most dangerous model in crypto.
The **seigniorage model** (UST/LUNA, TerraUSD): two tokens work in tandem. UST - the stablecoin ($1). LUNA - the governance token with a floating price. Mechanism: to create 1 UST, you must burn $1 of LUNA. To redeem 1 UST - you burn UST and receive $1 of LUNA. The peg is maintained by arbitrage: if UST < $1 - it's profitable to buy cheap UST and exchange it for $1 of LUNA (arbitrage). If UST > $1 - it's profitable to mint new UST by selling LUNA.
Why did Terra grow to $18B? **Anchor Protocol** offered 20% APY on UST deposits - an unsustainable rate subsidized from Terra reserves. Users minted billions in UST for the 20% yield. When Anchor's reserves were depleted and the rate was cut, a bank run on UST began. The Terra Foundation tried to defend the peg by selling $3B in Bitcoin from its reserves, but this proved insufficient.
**Reflexivity is the fundamental problem of algorithmic stablecoins.** When stability depends on confidence in stability, the system becomes fragile. In "good times" arbitrage works, demand grows, everything is stable. But when confidence is lost, a feedback loop kicks in: selling causes more selling. This is not a bug in a specific implementation - it is a fundamental property of systems without hard collateral. The Terra collapse was not an accident but an inevitable result of the design.
Why did the UST/LUNA death spiral become irreversible?
MakerDAO: CDP, governance, and real-world assets
**MakerDAO** is the oldest and largest DeFi protocol ($8B+ TVL), creator of DAI - a decentralized stablecoin with $5.3B in circulation. Unlike algorithmic stablecoins, DAI has survived every market crisis: the 2020 crash (ETH -60% in a day), the Terra collapse, the FTX collapse. The secret to its resilience is the well-designed Vault system (formerly CDP - Collateralized Debt Position), liquidations, and governance.
**Liquidation in MakerDAO** is a multi-step process. When a Vault's ratio drops below the Liquidation Ratio, **keepers** (bots) call the `bark()` function, launching a **Dutch auction**: the collateral price starts high and decreases every block until someone buys. The buyer pays DAI, which is burned (reducing the debt). The system penalizes the vault owner (13% liquidation penalty), incentivizing them to maintain a healthy ratio.
**DSR (Dai Savings Rate)** - yield for DAI holders. A user deposits DAI into the DSR contract and earns interest (in 2024: 5–8% APY). Where does the yield come from? From the Stability Fee paid by borrowers. If borrowers pay 5.25% and DSR = 5%, the difference goes to the protocol treasury (Surplus Buffer). MKR governance controls both rates, balancing DAI supply and demand.
**The Endgame Plan** - MakerDAO's strategic transformation (2023–2025). Key changes: rebranding to **Sky Protocol** (DAI → USDS, MKR → SKY), creation of **SubDAOs** - semi-autonomous divisions (Spark for lending, Sakura for RWA). Goal - scaling to $100B DAI and gradual decentralization. Critics argue that the rebranding and governance complexity create more problems than they solve.
DAI is just as risky as UST because both are "not real dollars." All stablecoins are equally dangerous and any can go to zero overnight.
DAI and UST are fundamentally different systems. DAI is backed by collateral at 150%+ (every DAI is backed by $1.50+ of assets in smart contracts). UST had no collateral at all - its peg relied solely on arbitrage with LUNA. DAI survived a 60% ETH drop in a day (March 2020, Black Thursday) because over-collateralization + liquidations + MKR backstop create three layers of protection. Crypto-backed stablecoins with over-collateralization are fundamentally more resilient than algorithmic ones.
After the UST collapse, media and newcomers started treating all "non-dollar" stablecoins with equal skepticism. But this is like comparing a gold-reserve bank with a Ponzi scheme - both can fail, but the mechanisms and probabilities are fundamentally different. The key distinction: is there real collateral exceeding the debt, and does a liquidation mechanism exist to seize it if necessary?
What happens to the MKR token if the MakerDAO system incurs losses and there is uncovered debt (bad debt)?
Key takeaways
- **Fiat-backed stablecoins** (USDT, USDC) are the simplest model: $1 in account = 1 token. Reliable with an honest issuer, but centralized, subject to censorship, and dependent on the banking system (USDC depeg from SVB)
- **Crypto-backed stablecoins** (DAI, LUSD) solve centralization through over-collateralization: $150+ of crypto for every $100 of debt. Transparent and censorship-resistant, but capital-inefficient
- **Algorithmic stablecoins** (UST) promised stability without collateral, but reflexivity makes them fundamentally fragile: loss of confidence triggers a death spiral. No purely algorithmic stablecoin has survived long-term
- **MakerDAO Vaults** - a system with three layers of protection: over-collateralization (150%+), automatic liquidations (Dutch auctions), and MKR backstop (MKR issuance to cover bad debt). DSR and Stability Fee balance DAI supply and demand
- Those $40B that vanished in 5 days are a direct consequence of a system with no collateral. DAI survived that same crisis because every token is backed by real assets with a safety margin. The backing model determines the stablecoin's fate
Related topics
Stablecoins are a key DeFi primitive connecting trading, lending, oracles, and regulation:
- AMM: Uniswap and constant product — Stablecoin pools (USDC/USDT, DAI/USDC) are the most popular DEX pools. Curve specializes in swaps between stablecoins with minimal slippage. The UST depeg began with an outflow from a Curve pool
- Lending protocols: Aave, Compound — Stablecoins are the primary asset for borrowing in lending protocols. DAI is created through a lending-like mechanism (Vaults). GHO - Aave's stablecoin - is minted directly from lending positions
- Oracles: Chainlink and the data problem — MakerDAO uses Chainlink oracles to determine collateral prices and trigger liquidations. Oracle manipulation can cause cascading liquidations or allow minting stablecoins at an inflated collateral price
- Regulation and compliance — The SEC shut down BUSD ($23B), classifying it as a security. The EU (MiCA) and the US are developing separate rules for stablecoins. Regulatory pressure is the main threat to the fiat-backed model
Вопросы для размышления
- Fiat-backed stablecoins are centralized but occupy 90%+ of the market. Crypto-backed ones are decentralized but capital-inefficient. Is a stablecoin that combines both advantages possible, or is this an unresolvable trade-off?
- MakerDAO accepts US Treasuries and other real-world assets (RWA) as collateral. Does this not turn a "decentralized" stablecoin into a disguised fiat-backed one? Where is the boundary between decentralization and pragmatism?
- After the UST collapse, many say algorithmic stablecoins are impossible in principle. But what if the problem was not the algorithm, but the lack of collateral and the unsustainability of Anchor (20% APY)? Could a hybrid model (partial collateral + algorithm) work?