Algorithmic Stablecoins and the Death Spiral

You’ve now met two kinds of stablecoin that keep their promise the boring, sensible way. Fiat-backed coins hold actual dollars in a bank. Crypto-backed coins lock up a pile of volatile crypto worth more than the coins they mint. Both answer the question “what’s behind this dollar?” with something you could, in principle, go and touch.

Algorithmic stablecoins answer that same question with a shrug. Behind the dollar is… almost nothing. No vault of cash, no over-collateralized crypto — just code, incentives, and faith that the market will keep playing along. When it works, it’s beautifully capital-efficient: a dollar conjured from a few lines of Solidity and a clever trade. When it breaks, it doesn’t wobble — it breaks all at once, and takes everyone holding it down together. This is the lesson where we watch that happen.

What “algorithmic” means

Picture a currency with no central bank vault, where instead a robot stands in the town square shouting “I will print or shred coins until the price is right.” That’s the algorithmic pitch. There’s little or no external collateral; instead the supply expands and contracts automatically to nudge the price back toward $1.00.

The mechanism rests on one promise: the protocol will always let you trade 1 stablecoin for $1.00 of something it can mint. “Something it can mint” is the load-bearing phrase — the protocol isn’t paying you out of a reserve it already holds, it’s printing the payout on demand. As long as that “something” is itself worth a meaningful amount, arbitrage keeps the peg honest:

• If the coin trades above $1.00, the protocol lets you create new coins for $1.00 of value and sell them at, say, $1.02. That extra supply pushes the price down toward the peg.
• If the coin trades below $1.00, the protocol lets you destroy a coin and collect $1.00 of value back, so buying a $0.98 coin and redeeming it for $1.00 is free money. That buying pressure pushes the price up.

It’s elegant on paper. The catch — the entire rest of this lesson — is what happens to “$1.00 of something it can mint” when the something starts falling.

When/why anyone builds these

The appeal is capital efficiency. A crypto-backed coin might need $150 of locked ETH to issue $100 of stablecoin — capital sitting idle. An algorithmic coin issues $100 of stablecoin against essentially $0 of pre-committed collateral. If it works, you’ve created a dollar out of thin air and good incentives. That efficiency is exactly why builders keep trying it, and exactly why it keeps blowing up.

The seigniorage two-token model

The classic algorithmic design — the one Terra made famous — is the seigniorage two-token model. (“Seigniorage” is just the old word for the profit a mint earns by creating money; here the protocol plays mint.) It uses a pair:

• A stablecoin that targets $1.00. On Terra this was UST.
• A volatile sister token that absorbs the price swings. On Terra this was LUNA.

The protocol hard-wires one rule: you can always swap 1 stablecoin for $1.00 worth of the sister token, and $1.00 worth of the sister token for 1 stablecoin — minting and burning each side to make the trade balance.

• Above peg (UST at $1.02): burn $1.00 of LUNA, mint 1 new UST, sell it for $1.02, pocket two cents. New UST supply drags the price back down.
• Below peg (UST at $0.98): buy the cheap UST, burn it, mint $1.00 of fresh LUNA, sell that for $1.00, pocket two cents. UST demand pulls the price back up.

Worked example. UST slips to $0.99. An arbitrageur buys 1,000,000 UST on the market for $990,000. They burn it, and the protocol mints them $1,000,000 worth of brand-new LUNA. They sell the LUNA for $1,000,000. Profit: $10,000, minus fees — and along the way they removed a million UST from circulation, helping shove the price back to $1.00. Multiply that by thousands of bots and the peg holds itself.

This only works while one assumption is true: LUNA has a high price and a deep, liquid market. The protocol is promising to pay out in LUNA, so LUNA has to be worth enough that minting more of it doesn’t crater its value. Hold that thought.

Two other algorithmic flavors are worth naming so you recognize them in the wild:

• Rebase (Ampleforth / AMPL): one token, no sister. Instead of changing price, the protocol changes everyone’s balance. If AMPL trades above target, every wallet’s balance expands overnight; below target, every balance shrinks. Your number of tokens changes while you sleep.
• Fractional (FRAX, early versions): a hybrid — each coin is part real collateral, part algorithmic, with the ratio adjusting over time. It’s a deliberate hedge against the all-or-nothing fragility of pure algo designs (and tellingly, FRAX kept drifting toward more collateral, not less).

The death spiral

Here’s the fatal flaw, and it lives right inside that elegant mint-and-burn loop. Run the simulation, then read what each step is doing.

Walk the loop in prose:

1. A big holder dumps a lot of stablecoin (UST), pushing it below $1.00.
2. Arbitrageurs do exactly what the protocol invites: burn UST to mint $1.00 of LUNA, then sell that LUNA to lock in the spread.
3. All that fresh LUNA hitting the market explodes LUNA’s supply and crashes LUNA’s price.
4. But LUNA is the thing “backing” UST. Now the backing is worth far less than it was a minute ago.
5. Holders notice and panic: if the protocol can only pay them in collapsing LUNA, they’d rather flee UST now. So more people sell UST, pushing it further below peg.
6. Defending that deeper depeg mints even more LUNA — back to step 3, but worse.

Each turn of the wheel makes the next turn nastier. This is reflexivity: price and confidence feed each other downward, with no floor. The mechanism designed to defend the peg becomes the engine that destroys it. Both tokens race toward zero, and they get there fast.

Case study: Terra/UST, May 2022

This isn’t a hypothetical. In May 2022 it happened to the biggest algorithmic stablecoin ever built, in real time, in front of everyone.

The setup. UST was the largest algo-stable in existence, with a market value around $18 billion at its peak. Demand for it was supercharged by Anchor Protocol, a Terra lending app that dangled a roughly 20% “savings” yield on UST deposits. Twenty percent, on a “stable” dollar, in a world where banks paid one. That yield wasn’t earned from sustainable lending — it was largely subsidized from a reserve that was visibly draining. It pulled in enormous deposits, which meant enormous demand for UST resting on an incentive that couldn’t last.

The break. In early-to-mid May 2022, a series of very large UST withdrawals and sales knocked it off peg. The mint-and-burn defense kicked in exactly as designed — and so did the death spiral. Watch UST’s price over the following days:

The damage. As UST defenders burned UST and minted LUNA without pause, LUNA’s supply went from hundreds of millions of tokens to trillions in a matter of days — textbook hyperinflation. LUNA’s price went from roughly $80 to fractions of a cent. UST never recovered its peg. All told, somewhere north of $40 billion of value evaporated, wiping out retail holders worldwide and triggering a wave of failures across the crypto lending sector that leaned on Terra.

The pitfall it teaches. The warning sign was hiding in plain sight: a double-digit “stable” yield with no sustainable source funding it. The yield wasn’t a feature; it was the kindling.

Why algorithmic designs are fragile

Strip Terra away and the structural lesson is general. Pure algorithmic stablecoins rely on two things that both vanish in a crisis:

1. Perpetual confidence. The peg holds only while enough people believe it will hold and keep arbitraging it. There’s no reserve to enforce it if belief evaporates — faith is the collateral.
2. A liquid, valuable sister asset. The payout token has to stay worth enough that minting more doesn’t crash it. But minting more is exactly what a depeg forces the protocol to do.

The deep problem is that the “collateral” is endogenous — created by the same system it’s supposed to rescue. Contrast that with fiat-backed and crypto-backed coins, whose backing is external: dollars in a bank, or ETH locked in a vault, don’t evaporate just because the stablecoin is having a bad day. Endogenous backing does the opposite — a crisis destroys the backing precisely when you need it to pay people out.

The graveyard is well populated: UST (the $40B+ catastrophe), Basis Cash (an earlier seigniorage design that never held peg), and Iron Finance (a partial-algorithmic coin, TITAN/IRON, that bank-ran to zero in 2021 — Mark Cuban famously got caught). The survivors in the stablecoin world overwhelmingly lean on real, external collateral; the further a design drifts from “we actually hold the value,” the shorter its life expectancy.

Key Takeaways

Next up: Depegs, risks and regulation — how stablecoins of every kind break, the failure modes beyond the death spiral, and the rules now being written to keep the dollar in “digital dollar.”