Algorithmic stablecoins are designed to maintain their peg without actually holding fiat instruments (currencies like USD, EUR, JPY, or securities like U.S. Treasury bonds) or crypto assets (BTC, ETH, or SOL, among others). They use code or ‘algorithmic’ mechanisms instead: supply rebases, dual-token mint-and-burn cycles, or short positions that hedge a long. 

The category nearly died in May 2022 when Terra’s UST collapsed, taking LUNA down with it. But it didn’t, and in 2026, modern variants like USDe, frxUSD (formerly, FRAX), GHO, and crvUSD are still active and available. Their architectures, however, are significantly larger than UST’s. 

This guide covers why Terra’s model failed, how new-age algorithmic stablecoins work, where and how you can access them across chains, and what the real risks look like in 2026. 

Key Takeaways

• Algorithmic stablecoins use programmatic mechanisms (supply rebases, mint-and-burn loops, or short positions) to maintain their peg without holding fiat or crypto reserves.

• The category traces back to two papers: Robert Sams’ Seigniorage Shares and Ferdinando Ametrano’s Hayek Money, both published in 2014. Vitalik Buterin then based his essay, The Search for a Stable Cryptocurrency, on these papers, laying the groundwork for algorithm-driven stablecoins. 

• Terra’s UST collapse in May 2022 nearly demolished this category, draining roughly $40B from its market cap. The lesson was harsh but specific: an endogenous backstop token cannot be a reliable safety net for the very system on which its value depends. 

• The 2026 lineup thus looks very different. USDe (delta-neutral), frxUSD (now fully collateralized), GHO (overcollateralized Aave borrow), and crvUSD (soft-liquidation) have replaced what came before. 

• Pure algorithmic stablecoins are essentially extinct in 2026. Collateral-backed designs that use algorithmic levers at the margins are the survivors. 

What is an algorithmic stablecoin?

Algorithmic stablecoins try to maintain a price peg — typically $1 — through smart-contract logic and economic incentives, rather than a collateral vault held one-for-one against the supply. The protocol expands supply or makes minting (a.k.a. issuing) more profitable when the price drifts above $1. Whereas when the price drops below $1, it makes burning more profitable, besides contracting supply. 

While this asset class entered the limelight around May 2022, its origins stem from research published in 2014: 

• Robert Sam introduced a dual-token volatility-absorbing model in his paper, A Note on Cryptocurrency Stabilization: Seniorage Shares. 

• Ferdinando Ametrano proposed a single-token rebasing system in Hayek Money.

• Vitalik expanded on their theses and published The Search for a Stable Cryptocurrency.

Together, they laid out the blueprint for most algorithmic stablecoins that followed. Ampleforth inherited Ametrano’s model, for instance, while Basis and Terra leaned on Sam’s work. And this is where the algorithmic model(s) stood out from the other, more dominant stablecoin categories: fiat-backed and crypto-collateralized. 

Fiat-backed stablecoins maintain their peg by holding fiat currencies like USD or short-dated treasuries off-chain. This is by far the largest category, with the two leading coins, USDT and USDC, comprising nearly 90% of the global stablecoin supply as of May 2026. 

Crypto-collateralized stablecoins (DAI, for example), on the other hand, rely on over-collateralized baskets of other cryptocurrencies and on-chain assets such as ETH or BTC to maintain their peg. 

Algorithmic stablecoins, however, are either partially collateralized or hold no collateral at all, letting the underlying mechanism do the work instead. This design also determines their failure mode vis-à-vis their traditional counterparts. Fiat- and crypto-backed coins fail when the issuer mismanages reserves or when their collateral declines faster than liquidations can keep pace. The purely algorithmic ones collapse when the incentive system breaks down, which is exactly what happened with Terra/LUNA in May 2022. 

For a parallel discussion of how stablecoin issuance is increasingly tied to dedicated chains, see stablechains aren't new.

What went wrong: Lessons from Terra/LUNA

Before digging deeper into how algorithmic stablecoins maintain their peg and what works for them in 2026, it’s worth noting what didn’t work in the past, using Terra’s example. 

Terra used a seigniorage model (see below), in which LUNA absorbed UST’s volatility. The mint-and-burn loop was simple: burn $1 of UST, mint $1 worth of LUNA, and vice versa. Arbitrageurs kept the peg tight as long as LUNA had a bigger market cap than the UST it was supposed to backstop. 

In May 2022, massive UST redemptions ran the loop in reverse for hours. UST burned, LUNA minted at scale, hyperinflating its supply and collapsing its price. The backstop didn’t merely shrink; it disappeared. Both tokens went to zero within days, wiping out over $40 billion. 

This failure has often been framed as decisive proof that algorithmic stablecoins don't work. That, however, is sloppy. The real problem was that LUNA, the volatility absorber, was endogenous: its value was a function of the same system it was supposed to insure. So under stress, the insurance was worth nothing, precisely when it was needed most. 

Modern stablecoin providers learned from Terra’s collapse and innovated architectures that are resilient to such death spirals. 

How algorithmic stablecoins maintain their peg

Four pegging mechanisms currently dominate the stablecoin ecosystem, with distinct risk profiles: 

• Rebasing: The total supply expands or contracts across all wallets to nudge the per-token price toward $1. For instance, if your 10 tokens are worth $20, the rebase mechanism halves your balance to 5. Your dollar value stays the same, but the supply contracts. Launched in 2019, Ampleforth adopted this model. But very few projects use pure rebasing today, because constantly fluctuating wallet balances are an accounting and composability nightmare. 

• Seigniorage or dual-token: A second token (a.k.a. bond token) absorbs volatility. When the stablecoin trades above the peg, the protocol mints more of it to capture the value, and vice versa. Basis Cash (BAC) and Empty Set Dollar (ESD) were among the first to experiment with this model, though neither survived. Terra’s UST/LUNA pair worked for a while, eventually becoming the most prominent failure of the dual-token mechanism.

• Fractional-algorithmic: Each token is partially collateralized (often with USDC or other stables), while the rest is backed algorithmically. frxUSD (formerly, FRAX) was the best-known example of this model, with a demand-adjusted collateral ratio. After deep stress-testing during the 2022-23 cycle, however, Frax migrated to a fully-collateralized iteration (v3), eventually dropping the algorithmic component entirely.

• Delta-neutral: The protocol longs spot crypto and shorts perpetual futures (“perps”) of equal size. This produces a synthetic dollar, as the long and short positions cancel out in dollar terms. Ethena’s USDe popularized this in 2024 and reached a peak circulation of $14B in 2025, before contracting in sync with market-wide trends since the October 2025 liquidation cascades.

The 2026 lineup: USDe, frxUSD, GHO, crvUSD

Given the upsides of the above models, and particularly their downsides, new-age stablecoin issuers have mostly adopted hybrid mechanisms rather than purely algorithmic ones.

With its delta-neutral framework, Ethena’s USDe currently leads the non-collateralized stablecoin sector. It balances a long position in ETH/BTC with an equivalent perps short to maintain a nearly stable dollar value. The funding rate from the short side is paid out as yield to USDe stakers, with variable APYs in the 4–15% range depending on broader market dynamics. At the time of writing, USDe is the fifth-largest stablecoin by outstanding supply. 

USDe is widely available across Ethereum, Arbitrum, Optimism, Solana, Mantle, and several other L2s. However, the stablecoin is not accessible to users in the EU and EEA because Ethena exited this region due to structural non-compliance with MiCA’s 1:1 fiat-backing rule.

Ethena is a LI.FI partner; for context on the integration, see Ethena x LI.FI: a direct path to USDe and ENA.

frxUSD (formerly, FRAX) by Frax Finance is another example of how erstwhile algorithmic stablecoins are faring today. The project migrated to a fiat-redeemable (i.e., fully collateralized) stablecoin backed 1:1 by tokenized U.S. Treasuries. Its reserves are now held by governance-approved ‘enshrined custodians’, including BlackRock’s BUIDL, Superstate’s USTB, and WisdomTree’s WTGXX. Given such changes, frxUSD has moved so far away from its 2022 era that its ‘algorithmic’ label is now strictly historical.

That said, Aave’s GHO and Curve’s crvUSD represent two innovative, semi-algorithmic models to have emerged in recent years: 

• GHO is a borrow-side stablecoin issued against collateral on Aave. Users mint GHO by supplying collateral and paying a protocol-defined borrow rate, which acts as an interest-rate lever to control supply. Mechanically, it’s closer to DAI than to UST, and can be called ‘algorithmic’ only in a loose sense. GHO crossed $500M in circulating supply by Q1 2026 and is live on Ethereum, Base, and Avalanche, with Arbitrum on the roadmap. After Aave V4’s launch in March 2026, GHO became the native settlement asset across the new hub-and-spoke architecture. 

• crvUSD (Curve) uses Curve’s LLAMMA system that gradually liquidates collateral using a continuous AMM. Each loan has a range of bands, rather than a single, hard liquidation threshold/price. As collateral falls through these bands, the AMM steadily converts it to crvUSD and reverses the process if prices recover. This avoids the cliff-edge cascades that hit stablecoins like DAI (MakerDAO) during black-swan moves. crvUSD has mostly held around $250M in supply since 2025, primarily on Ethereum.

The objective bottom line, therefore, is that pure algorithmic stablecoins — no collateral, no hedge — are effectively extinct. What survived are fully or partially collateralized designs that use algorithmic levers around edges. 

Where to access algorithmic stablecoins across chains

There are several articles on algorithmic stablecoins, but almost none of them address the elephant in the room: where do they actually live and, more importantly, how to access them? 

USDe and frxUSD are deployed across multiple chains. GHO is mostly Ethereum-native. crvUSD is multi-chain, but its liquidity is concentrated on Ethereum. So if you want to actually use one of these tokens, you have to figure out how to get them to the chain where you have your funds. 

Historically, that meant bridging the underlying asset, swapping on the destination DEX, and hoping the route (and liquidity) exists. Three steps. Three transactions. Three points where things can go wrong.

LI.FI is the universal liquidity layer that collapses these flows. It aggregates and orchestrates 27+ bridges and 31+ DEXs across 60+ chains and finds the best path for any given transfer, whether that’s a same-chain swap, a bridge, or a bridge-and-swap combined. 

• For institutional users moving stablecoins at scale, the Stablecoin API tightens default slippage to 0.1% and caps price impact at 2%, with mint-and-burn-preferred routing through bridges like Glacis, Mayan, and Celer. 

• For end users, Jumper is the consumer app powered by LI.FI’s infrastructure. It supports USDe, frxUSD, GHO, and crvUSD swaps across the chains where each is deployed, including the cross-chain case where you want to start on one chain and end on another.

Figuring out which algorithmic stablecoin to use is a research question. But with LI.FI, moving in and out of any of them, once you’ve decided, doesn’t have to be. 

Risks to watch

Hybrid, new-age stablecoins are considerably more robust than their purely algorithmic predecessors. But they still carry risks that are worth considering, such as:

• De-pegging: Failing to maintain its peg is the most brutal, almost existential risk for any stablecoin. And modern stablecoins aren’t free from it. USDe’s stability, for instance, depends on funding rates staying positive. In a sustained bear market with negative funding, the model can be stressed. Likewise, frxUSD, though now backed by tokenized Treasuries, carries the same risks as any RWA-backed stablecoins, including issuer risk, custody risk, and redemption-throughput risk. 

• Smart contract risk: More moving parts than a fiat-backed stable means more code, and more places where a bug can drain funds. 

• Liquidity fragmentation: The same stablecoin on different chains can have very different liquidity depth and slippage. A 100K-worth USDe swap on Ethereum may play out very differently from the same trade on Solana.

• Governance risk: Most protocols discussed so far are governed by token holders who can adjust parameters. That’s a feature for adapting to market conditions, but also a risk if governance is captured or makes a poor call. 

• Counterparty risk in hedges: For delta-neutral designs, the short position is held on a perps venue. If that venue fails, the hedge fails. 

• Regulatory risk: MiCA’s 1:1 fiat-backing rule, for instance, already pushed Ethena out of the EU. Other jurisdictions are watching, and synthetic-dollar structures don’t fit cleanly into stablecoin frameworks designed around fiat-redeemable issuers.

FAQs

Are algorithmic stablecoins safe after the Terra/Luna collapse?

Terra’s specific design, with an endogenous backstop token whose value depended on the same system it was meant to insure, has been abandoned by every serious project. Modern algorithmic stablecoins use exogenous collateral (frxUSD, GHO) or external hedges (USDe). The death-spiral failure mode that killed UST is structurally different from how today’s algorithms work. And while they are not riskless, they don’t carry the specific failure mode that took down Terra. 

What is USDe and how does Ethena's delta-neutral model work?

USDe is Ethena’s synthetic dollar. The protocol holds a spot ETH or BTC long and an equal-sized short position in perpetual futures. The long and short cancel out in dollar terms, producing a stable $1 unit. The funding rate from the short side is paid out as yield to users who stake USDe (sUSDe). It works as long as funding rates stay positive enough to cover protocol costs.

Can algorithmic stablecoins completely lose their peg?

Yes. Terra’s UST went to near-zero in May 2022. Modern designs reduce this risk, but they don’t (can’t) eliminate it. Anyone using these tokens must size their positions accordingly. 

What is the difference between algorithmic and fiat-backed stablecoins?

Fiat-backed stablecoins hold fiat currencies like USD (or short-dated treasuries) in a bank to back each issued token. Algorithmic stablecoins don’t hold any such reserve and rely on code, incentives, and sometimes partial on-chain collateral to maintain the peg. The trade-off is decentralization and capital efficiency on the algorithmic side, versus straightforward issuer-backed redemption on the fiat side. 

Which algorithmic stablecoins are available on multiple blockchains?

USDe is on Ethereum, Arbitrum, Optimism, Solana, Mantle, and several L2s. frxUSD is native to Fraxtal and Ethereum, with bridges across several other networks. GHO is available on Ethereum, Base, and Avalanche, and might be deployed on Arbitrum in the future. crvUSD is primarily Ethereum-native. Liquidity depth varies significantly by chain.

Try it!

Moving USDe, frxUSD, GHO, or crvUSD across chains has one of DeFi’s most route-fragmented operations. But not anymore. 

Jumper enables cross-chain swaps for all of them, across 27+ bridges and 31+ DEXs, using LI.FI’s universal liquidity infra. Meanwhile, for protocols and treasury teams moving stablecoins at scale, the Stablecoin API gives tighter defaults and direct integration. Try them out.