‘Built to Fail’? Why TerraUSD’s Growth Is Giving Finance Experts Nightmares

Late on Monday, April 18, the stablecoin terraUSD (UST) edged out Binance’s BUSD to become the third-largest stablecoin by market cap. There are now nearly $18 billion UST in circulation. That’s well below the nearly $50 billion total for Circle’s USDC, or the $82 billion worth of Tether’s USDT roaming the Earth.

But UST is also much different from those competitors, in ways that could make it incredibly risky.

You’re reading Money Reimagined, a weekly look at the technological, economic and social events and trends that are redefining our relationship with money and transforming the global financial system. Subscribe to get the full newsletter here. Michael Casey is away.

Stablecoins are tokens tracked by a blockchain, but in contrast to assets like bitcoin (BTC), they’re intended to consistently match the buying power of a fiat currency, most often the U.S. dollar. Stablecoins were first created to give active crypto traders a tool for moving quickly between more volatile positions, though as we’ll see, the potential for big interest rates on loans has also helped attract capital.

USDT and USDC are so-called “backed” or collateralized stablecoins. They keep their 1:1 dollar “peg” because they are (ostensibly) backed by bank accounts holding dollars, or by other dollar equivalent assets, for which tokens can be redeemed – although Tether has been notoriously reticent to specify the nature of its reserves.

UST, by contrast, began life as what’s known as an “algorithmic” stablecoin. These could also be referred to as “decentralized” stablecoins because decentralization is their primary reason for existing. A collateralized stablecoin like USDT or USDC is reliant on banks and traditional markets. That makes them in turn subject to regulation, enforcement and ultimately, censorship of transactions. Circle and Tether are run by centralized corporate entities with the ability to blacklist users and even seize their funds. Both systems have done this, sometimes at government behest.

In principle, algorithmic stablecoins like UST don’t have this censorship risk because they are not run by centralized corporate structures and do not hold backing in traditional institutions like banks. Of course, in reality “decentralization” is relative, and most such systems today still have key men, such as Do Kwon at Terraform Labs, or affiliated organizations that provide labor and funding. Whatever a system’s “decentralized” branding, regulators can still go after such public targets, a risk that’s worth keeping in mind.

Instead of being backed by assets like dollars or bonds, algorithmic stablecoins are intended to maintain their dollar peg by what looks to some like financial alchemy. But experts are highly skeptical that these systems’ designers have actually figured out how to turn lead into gold.

Broadly, algorithmic stablecoins rely on a pair of tokens, both created ex nihilo at the token’s launch. One token is the stablecoin itself – terraUSD, frax, neutrinoUSD – and the other is the corresponding “balancer token” ­– LUNA, FXS or WAVES, respectively. Generally, the balancer token can be “burned,” or destroyed, to create more of the corresponding stablecoin.

The defining algorithm in an “algorithmic stablecoin” is a function that changes the amount of the balancer token required to create the stablecoin. This changing burn value is intended to create arbitrage opportunities for traders: gaps between the price of the balancer token and the stablecoin that they can profit from, but that also drive the price of the stablecoin toward its peg. In principle, this would attract profit-motivated traders to drive the stablecoin back to $1 whenever the peg wavered.

The nuances of these systems vary, but it doesn’t take much financial expertise to see reason for skepticism. Let’s hear it from an expert anyway.

“You have a two-token system and one token has no intrinsic value, it’s only derived from secondary trading value, and the other token is supposed to be a stable token, and you have an arbitrage,” wonders Ryan Clements, an assistant professor of Business Law at the University of Calgary. “At some point you look at the one with no intrinsic value and you think, why is this worth $5?”

Clements’ past research has focused on arbitrage risks and instability in exchange-traded funds, where there can be a similar disconnect between the price of a synthetic asset and that of its underlying components. Last year, Clements published a paper about algorithmic stablecoins called “Built to Fail,” arguing that the systems are inherently fragile and can never be truly “stable.” (Note that MakerDAO’s DAI is a bit of an outlier with some unique design features, and these critiques don’t necessarily apply to it.)

Clements argues the theory behind algorithmic stablecoins relies on three specific flawed assumptions. An algorithmic stablecoin, he writes, can only function when three conditions are met at all times:

But Clements argues these assumptions break down frequently, particularly during times of instability.

One clear sign of this, according to Kevin Zhou of hedge fund Galois Capital, is that the teams behind algorithmic stablecoins sometimes have to essentially step in and do the balancing arbitrage themselves.

“Some of these algo [systems] are market-making,” says Zhou. “I don’t know if it’s them directly, or they give a loan to a market maker to maintain the peg.”

There’s even clearer evidence from algos that have already collapsed, most notably Iron Finance, which unwound in a “death spiral” in June 2021. As Novum Alpha CEO Patrick Tan explained the Iron Finance unwind at the time, “The speed at which [balancer token] TITAN started to fall caused [supposed stablecoin] IRON to lose its peg, allowing traders to redeem IRON, which was priced at US$0.90 as the peg started to slip, for US$0.75 in stablecoin and US$0.25 in TITAN.” The system’s stablecoin and balancer tokens have both since dropped to near zero.

Similar opportunities for profit can emerge whenever an algorithmic stablecoin slips too far from its peg. Once the arbitrage emerges, it becomes self-perpetuating and self-accelerating. Without outside intervention, it is an unavoidable death sentence.

The team behind LUNA/UST seem to agree with the idea that such systems can’t work. They don’t say so out loud, but they’ve made a variety of efforts to add more stability into the system, and those efforts have accelerated alongside UST’s growth.

It is important to note the terraUSD stablecoin is the central function of the entire Luna layer 1 blockchain, which also has general smart-contract functionality. Luna also has a variety of other tools and functions, including the Anchor lending platform, non-fungible tokens (NFT), initial coin offerings (ICO) and the like. The strategy here, as laid out by Luna bull José Maria Macedo of Delphi Digital on a recent episode of the “Bankless” podcast, is that demand for these tools will provide another source of support for the price of LUNA and, in turn, improve the stability of UST.

Clements acknowledged this ecosystem play makes UST the algorithmic stablecoin that “seems to have the most logical potential for stability.” But, he said, “I don’t think an ecosystem play is nearly as stable as something like USDC.”

The biggest question mark in the ecosystem, according to Clements and others, is Anchor. Anchor has attracted nearly $16 billion of deposits of LUNA, wrapped ETH (backed by actual ether that’s tied up on the Ethereum chain) and, above all, UST usually “bonded” (or locked) for long periods of time. This provides stability by preventing the sale of a large chunk of UST.

The problem is the deposit demand is essentially an illusion. Anchor provides 20% APY on locked deposits, but that rate isn’t profitable for the system: The current APR being offered to borrowers is just 11.81%. A lender is only sustainable if it collects more interest on its loans than it pays out to deposits.

That gap between lender and borrower APY is covered by a reserve fund that has been draining rapidly as the amount locked in Anchor grows. Moreover, even at 11.81%, lending demand seems light: With $15.5 billion locked in Anchor, only $3.2 billion of loans are outstanding, according to data provider DeFiLlama.

“All the UST in Anchor is just rotator capital, farming for the subsidized yield,” said Zhou. “Rotator capital” means money that is moved rapidly from one protocol to another in search of high yield (what Wall Streeters used to call “hot money”). If Anchor’s yield drops below the competition, there could be a significant rush to the exits, during which “rotators” in the staking pool would look to unload their LUNA and UST.

That seems practically inevitable because there isn’t an infinite amount of money to pour into the reserve fund. As soon as that financial nitroglycerin runs out, it could create the conditions for a depegging of terraUSD.

But the people behind Luna have come up with an answer to that, too. Starting in February, the affiliated Luna Foundation Guard (LFG) began purchasing bitcoin to create a further backstop for TerraUSD. The LFG reserve, including bitcoin and other assets, now totals about $2.3 billion.

It’s still unclear how this reserve will actually wind up working. The BTC is still held by LFG, not by the luna/terraUSD protocol. Debate is currently ongoing in the Luna community over proposals for integrating a second, different kind of reserve asset into the previously purely algorithmic system. The broad contours of the proposals, however, all involve an automated and arbitrage-incentivized mechanism that would open swaps from UST to BTC when the UST peg wavers.

There are two reasons to wonder whether this is a real solution. First, it doesn’t change the basic math of UST’s exit liquidity in a crisis. According to Zhou, the total of all non-LUNA reserves and all UST being used in applications on the Luna chain only amounts to about $3 billion, against $18 billion in stablecoins outstanding.

“So there’s about $14 billion that needs to exit through luna in a compression or unwind,” said Zhou. That kind of selling pressure would likely drive down the price of luna itself dramatically, he argued, essentially blocking all the exits at the exact moment the building is on fire. “On a full unwind, some people are going to be left holding the bag, either in terraUSD that is depegged, or in hyperinflated luna that has no bid.”

But it might be even worse than that. What if building a BTC reserve actually increases the likelihood of a bank run?

“This could create an incentive (effectively an economic moral hazard) to utilize strategies to acquire discounted BTC,” Clements says of the plan. Such a coordinated, intentional attack could look similar to George Soros’s legendary attack on the British pound. The Bank of England spent more than 3.3 billion pounds countertrading to defend the pound’s peg against other European currencies, but ultimately failed.

Attacking a pegged asset could be described as an attempt to acquire the backing assets at a discount, by forcing a fire sale of the collateral. Especially given that BTC is a harder asset and generally superior money to LUNA/UST (and if it isn’t, it’s not much of a reserve asset), it is nearly a law of nature that LUNA or UST holders will seek to transform those tokens into BTC.

The sharks may already be circling.

“There’s a lot of incentive to attack the peg, especially for people outside the [crypto] space,” said Zhou. “There’s a lot of these Chicago [traders], they don’t give a [damn]. If they see an opportunity, they’re going to Soros-attack it. They’re just waiting for the right moment … I think [UST] will collapse on its own, but if someone attacks it, it will collapse even faster.”

There are other concerns, both with the BTC reserve plan and terraUSD’s basic structure. For instance, one aspect of a leading proposal for the reserve would rely on offering traders arbitrage incentives to replenish the BTC reserves once a crisis has passed. That’s perfectly sensible from a decentralization perspective, since in the long term you don’t want to rely on outside groups like LFG.

“But what happens if there isn't enough trading activity to replenish the [BTC] reserves outside of crises?” asks Clements. “This is another ongoing vulnerability, in my opinion.”

An even more fundamental question may be exactly how much leash regulators will give Do Kwon and his fellow travelers. The various capital injections and reserve restructurings make it clear there’s not much that’s “decentralized” about luna/terraUSD. At $18 billion, terraUSD is far from a systemic risk for mainstream finance, but continued growth could certainly invite the proverbial Eye of Sauron.

And while it’s not a big worry for Wall Street, luna/terraUSD could be a systemic risk to the cryptosphere. Do Kwon has said he would like to grow the system’s BTC holdings to $10 billion. But if that reserve is integrated in the wrong way, it could create the conditions for a future sudden flood of BTC for sale if the peg is attacked or breaks under market strain.

“Could the utilization of the bitcoin reserve transmit volatility across the crypto ecosystem?” Clements asked on CoinDesk TV’s “First Mover.” “It would seem obvious that it could.”

So while Iron Finance sank alone, terraUSD could drag others down with it.