What is an algorithmic stablecoin, really?
An algorithmic stablecoin is a token designed to track a target price, usually $1, without holding matching reserves of dollars or short-dated debt. Instead, the protocol uses a second, volatile token and a set of smart-contract rules to expand and contract supply. When the price trades above the peg, the system mints more stablecoins and absorbs the new supply, pushing the price back down. When the price trades below the peg, the system is supposed to remove supply, typically by letting holders swap stablecoins for the volatile companion token at a discount, then burning the stablecoins.
This design is often called a 'seigniorage' model, a name borrowed from the traditional right of a government to profit from issuing currency. The promise is that you get the convenience and decentralization of a dollar-pegged token without needing a bank account, a custodian, or even a real dollar. The companion token, sometimes called a 'share' or 'bond' token, is supposed to absorb the volatility that the stablecoin refuses to hold.
That promise sounds elegant on a whiteboard. In practice, it asks a volatile, often thinly traded asset to behave like a piggy bank during a bank run. The mechanism only works while confidence holds, and confidence is exactly what disappears when the peg starts to slip.
The risks of algorithmic stablecoins: what actually breaks
The defining risk of a pure algorithmic stablecoin is the so-called 'death spiral.' It is not a bug, it is a mathematical feature of the design. The loop works in both directions: a falling peg causes holders to flee the stablecoin into the companion token, which crashes the companion token's price, which weakens the system's ability to defend the peg, which causes more holders to flee. Once the loop is running in reverse, it is extremely hard to stop without external intervention.
Real failure modes are well documented. TerraUST collapsed in May 2022 after several days of large withdrawals from Anchor Protocol, a lending market that had promised around 20% yield on UST deposits. The native LUNA token fell from roughly $80 to a fraction of a cent, UST lost its dollar peg, and an estimated $40 billion of value evaporated within about a week. Earlier designs had already failed in similar ways. Basis Cash, an empty-shell clone of a never-launched project called Basis, saw its 'share' token collapse to near zero in 2021 because there was no real demand absorbing emissions. Iron Finance's 'TITAN' token went through a textbook bank run in June 2021, taking the partially collateralized IRON stablecoin with it.
Other risks sit one layer below the death-spiral math. Governance attacks can change the parameters that supposedly hold the peg. Smart-contract bugs in the minting and burning logic can be exploited for free stablecoins. Liquidity in the off-ramp pools can be thinner than the headline market cap suggests, so a 'paper' $1 peg evaporates the moment someone tries to redeem in size. And because algorithmic stablecoins are mostly used inside crypto, rather than for actual payments, they tend to trade in correlated conditions, which is the worst possible environment for a design that needs uncorrelated demand to defend the peg.
For users, the practical lesson is brutal. With a centralized stablecoin like USDT or USDC, a de-peg event is painful but usually recoverable, and there is a legal entity that can be sued, audited, and eventually made whole. With a pure algorithmic stablecoin, the protocol either holds the peg or it does not, and if it does not, the recovery mechanism is the same volatile token that is currently collapsing.
Why regulators closed the door on pure algos
Regulators did not need a PhD in mechanism design to react to Terra. They watched $40 billion disappear in a week, they watched retail investors, many of them new to crypto, lose their savings, and they watched the contagion spread briefly to broader crypto lending markets. The political lesson was simple: a financial instrument that can vaporize in 72 hours, with no balance sheet, no insurance fund, and no named defendant, is not something a modern central bank wants operating inside its jurisdiction.
The clearest policy response is the US GENIUS Act, the Guiding and Establishing National Innovation for U.S. Stablecoins, which moved through Congress in 2025. The act defines a permitted payment stablecoin as one backed by specific high-quality liquid assets, cash, short-dated US Treasuries, and similar instruments, held by a regulated entity, with regular public disclosures. Algorithmic stablecoins are explicitly excluded. The text does not mince words: a token whose peg is maintained 'primarily' by an algorithm, rather than by reserves, is not eligible for the regulatory safe harbor the act creates.
Outside the US, the same logic has shown up in different forms. The European Union's MiCA framework, the UK's proposed rules for fiat-backed tokens, and the Monetary Authority of Singapore's stablecoin guidance all gravitate toward reserve-backed, audited, redeemable designs. None of them carve out space for a token that relies on a companion coin to defend its peg. Algorithmic stablecoins are not banned outright in most jurisdictions, but they are stripped of the regulatory cover that lets a product call itself a regulated payment instrument.
The practical effect is that any token marketing itself as an 'algorithmic stablecoin' inside a major regulated market in 2026 is, by definition, a hybrid. It either has collateral or it is operating in a legal gray area. Neither label is neutral, and neither is a fix for the underlying death-spiral math.
The hybrid era: FRAX, USDe, and the move to 'partial algo'
After Terra, the most ambitious projects did not try to rebuild the pure algorithmic model. They tried to graft real collateral onto it, and then argue that the resulting hybrid is a new kind of animal. The two most cited examples are Frax and Ethena.
Frax started life in 2020 as a fractional algorithmic stablecoin, partly backed by USDC, partly by a seigniorage share token called FXS. After Terra, the team progressively raised the collateral ratio. By 2023, FRAX was fully backed by reserves, and the algorithmic component was effectively dormant. The project's own documents now describe it as a 'crypto-collateralized' stablecoin. The shift is an admission that the original model could not survive a serious stress event, and that the market is unwilling to trust a partial-peg mechanism with real money.
Ethena's USDe is the more interesting case. USDe is a 'synthetic dollar' that maintains its peg not by holding dollars, but by holding a delta-neutral position in crypto. A user deposits a token, say ETH, and Ethena opens a long spot position and an equivalent short perpetual futures position. The short cancels out the price exposure, so the value of the collateral stays roughly stable in dollar terms. The yield that USDe pays comes from funding rates on the perpetual swaps, which are often positive when the market is bullish. There is no companion token, no mint-and-burn seigniorage share, and no algorithmic expansion of supply.
This is not an algorithmic stablecoin in the Terra sense. It is a collateralized product whose collateral is a perpetual futures hedge. Ethena is explicit about this, and USDe's market behavior during the October 2024 and early 2025 drawdowns showed the difference: USDe briefly traded below peg, and the response was a rebalancing of hedges, not a hyperinflation of a partner token. The system is fragile in different ways, particularly around exchange counterparty risk and the cost of carry when funding rates flip negative, but it is not running the same death-spiral math as Terra.
The honest summary is that hybrids are a real engineering step forward, and they are also a marketing shell game. Calling a hybrid 'algorithmic' is a way to suggest the upside of seigniorage without owning the downside of a death spiral. Regulators have noticed, which is why the GENIUS Act's definition is so careful. The line between a crypto-collateralized stablecoin and an algorithmic one is drawn by what happens when confidence fails, not by what happens when confidence holds.
The difference between 'algorithmic' and 'crypto-collateralized'
The terminology has been deliberately muddied, so it is worth being precise. An algorithmic stablecoin defends its peg by minting and burning a volatile companion token. The peg is a function of the protocol's incentives and the market's belief that the incentives will work. There are no dollars, no Treasuries, no custodied assets, and no legal claim against a reserve fund.
A crypto-collateralized stablecoin holds other crypto assets as backing, usually over-collateralized to absorb volatility, and lets users mint the stablecoin against that collateral. MakerDAO's DAI is the canonical example. The peg is a function of liquidation mechanics: if the value of the collateral falls, the system sells it to keep the stablecoin whole. This is also fragile, and the March 2020 'Black Thursday' event showed how fast oracle failures can cascade, but the failure mode is liquidation, not hyperinflation of a partner token.
The two models share a word, 'crypto,' but they do not share a death spiral. A crypto-collateralized design can fail if collateral collapses, oracles misreport, or governance is captured. An algorithmic design can fail for those reasons, plus the additional reason that its core mechanism is mathematically hostile to confidence. Conflating them is not a small error. It is the error that lets a project market a Terra-style risk as a Maker-style safety profile.
Why CeFi reserves and audits don't fix the structural problem
A common response to algorithmic stablecoin criticism is to point at centralized stablecoins like USDT and USDC and say, 'at least the new designs are audited.' This conflates two different questions. The first is whether the issuer is honest. The second is whether the design can survive a loss of confidence. Centralized reserves answer the first question, and they do not answer the second.
Even a fully reserved, fully audited centralized stablecoin can de-peg. USDC did exactly that for roughly 60 hours in March 2023 when Silicon Valley Bank failed and a portion of its reserves was temporarily inaccessible. The peg held because the reserves were sound, the bank reopened under new ownership, and the issuer had a credible redemption pipeline. A pure algorithmic stablecoin, or a heavily algorithmic hybrid, has no such fallback. There is no reserves manager who can wire dollars to a market maker to defend the peg, because there are no dollars, only a companion token whose price is falling in lockstep with confidence.
This is the deepest structural lesson of Terra. The death-spiral math is not a function of the operator's honesty. It is a function of the design. As long as a token's peg depends on the willingness of holders to absorb a volatile, possibly illiquid asset during a crisis, the design is fragile. Adding audits, governance, or a custodial wrapper does not change the math. It only changes which entity takes the reputational hit when the math runs out.
What this means for users, builders, and regulators
For users, the practical takeaway is that 'algo' is now a marketing word, not a technical description. If a product describes itself as an algorithmic stablecoin in 2026, the relevant questions are: what is the actual collateral, who custodies it, what is the off-ramp during a crisis, and what specifically is the algorithmic part doing. If the answers are vague, the risk profile is closer to Terra than to USDC, regardless of what the whitepaper says.
For builders, the honest path forward is one of three: build a transparent, reserve-backed stablecoin and accept the regulatory overhead; build a crypto-collateralized over-collateralized design and accept the capital inefficiency; or build a synthetic dollar like USDe and accept the exchange and funding-rate risk. None of these are algorithmic in the Terra sense, and pretending otherwise is a way to get de-platformed by exchanges and regulators at the same time.
For regulators, the GENIUS Act and similar frameworks have already drawn the line. Pure algorithmic stablecoins are excluded. The remaining ambiguity is around hybrids, especially synthetic dollars, and around tokens that claim to be collateralized while running a significant algorithmic component. The next two years of policy will be about closing those loopholes, and the projects that survive that process will be the ones that do not need the loophole in the first place.
How to track algorithmic and hybrid stablecoins the smart way
Algorithmic and hybrid stablecoins are among the most volatile corners of crypto, and the news cycle around them is loud, fast, and often misleading. A new 'algo' launch, a yield bump, a governance vote, or an exchange listing can move a peg by hundreds of basis points in a single session. Tracking that manually is a losing game. Zippfeed surfaces stablecoin and regulation headlines with sentiment scoring, bullish, neutral, or bearish, and an importance rating, so you can separate structural risk from short-term noise and react before the peg, not after.
