A Little Unsteady
I think it would be fair to say that, over the course of crypto, three classes of coins have served as the backbone of the industry: BTC, ETH and ERC20 tokens, and stablecoins. It wasn’t until recently that stablecoins started to get the amount of attention as the other major sources of liquidity, but when it did the space exploded.
For years there was only one major stablecoin — Tether. And the truth is that, by market cap, it’s still the king, with nearly 90% of the total stablecoin dominance. But there are other projects with major attention in the space. For now we’ll focus on USDT and two others, which together represent the largest classes of stablecoins: USDC and DAI.
Without stablecoins, crypto liquidity venues would look a lot like Coinbase: coins trading against USD with wire transfer deposits and withdrawals.
There are two major reasons that stablecoins exist. The first is that, especially in crypto, banking can be difficult and unstable. It’s hard even to open a bank account as a crypto institution — anyone who’s tried has probably been told a few times that their business just doesn’t quite fit the risk profile of the bank. And even once you have one, they’re liable to be closed at any time; pretty much no major players in the space have never had a bank account shut down on them. And if you were relying on USD fiat support for your BTC/USD pair, losing your bank account can mean losing your exchange. More generally, many people in crypto just don’t trust banks at all, and would rather not be reliant on them.
The second major reason is transferability. If you have USD on Bittrex and want to send it to Coinbase you’re out of luck (unless you want to wait for wire transfers to clear). But if you have USDT on Kraken you can send it to Binance, and if you have USDC on OKEx you can send it to Uniswap.
For most of its history, crypto has been reliant on stablecoins. For years USDT and Bitfinex served as the sole fiat on/offramp for most of the non-US ecosystem. Everyone else piggybacked on their banking, and instead of support USD they just listed all of their markets against USDT. So users could send USD to Tether, create it, send it to any exchange, and then buy there — the Tether treasury really was supporting most o the industry.
But as more attention has come on stablecoins, so has more skepticism. And lots of people are worried about the systematic risk posed to the ecosystem by stablecoin collapse. If you want to know what could bring down markets on Binance, OKEx, Huobi and Bitfinex all at once — the answer is USDT imploding.
Not all stablecoins are made equal, though. And one question that many venues have to answer is exactly how much they trust them. What should an exchange’s primary BTC/USD(?) pair be? What collateral level should be allowed on a DeFi stablecoin loan?
Most people agree on the main strengths and weaknesses of the coins:
Strengths: market liquidity and USD backing. Weaknesses: regulatory status, proof of funds, banking risk. There are no broadly similar stablecoins.
Strengths: USD backing, trust, regulation. Weaknesses: market liquidity, banking risk, blacklisting. TUSD, PAX, HUSD, and BUSD are all close to equivalent, and a number of others (USDS, etc.) are similar.
Strengths: No banking/centralized risk. Weaknesses: market liquidity. A number of DeFi stablecoins are similar, e.g. sUSD.
How do you trade off between those?
Well, first you need to know the context. Are you worried about small dislocations or large ones? Are you worried about being regulated or about touching unregulated funds? Does your userbase have better banking or exchange connectivity?
There isn’t necessarily a one size fits all answer. But some of this is fairly universal. And in the end there are two dominant factors here:
- How much adoption does it have?
- How easy is it to create and redeem?
Liquidity is really important! But it’s mostly a function of those two factors: market liquidity comes from adoption, and peg stability comes from redemptions.
Adoption
No matter what platform you’re using, liquidity matters — in fact it’s possibly the most important factors for a stablecoin. How much size can you do without temporarily depegging it? How much liquidity will you expect in your markets?
USDT has the greatest market liquidity by far. It trades billions of dollars per day, and you can easily trade $100m of it on an exchange without moving the USDT peg close to 1%. No other stablecoin comes close. Sure, USDC is backed 1:1 by USD; try telling that to any of its orderbooks.
It’s worth emphasizing how significant this adoption dominance is. USDT vs USD represents a significant fraction — maybe the majority — of all OTC volume in crypto. The top 3 spot exchanges by volume are almost entirely USDT based. USDT has ten times the market cap of its nearest competitor and nearly 100 times the volume.
It’s worth noting, too, that USDT is particularly dominant outside of the west. Partially this is because banking is harder, and so the value of USDC’s on-ramp is lower; partially this is because regulation is seen much more skeptically. Many people in the west see a stablecoin and think “oh boy I sure hope this is backed by an FDIC insured bank account!” Many in Asia see one and think “I sure hope there’s no government affiliated institution that could freeze the majority of these funds”. Many people have wondered why bad Tether legal news didn’t crash USDT’s price more; the answer, probably, is that those who cared most about the news weren’t those who used Tether. Instead, the need for a digital form of wealth separate from any large government or institution, and the need to transfer it easily, tends to drive USDT demand.
Sure, USDC has some adoption in American and institutional crypto, and DAI in DeFi. But overall USDT always has been, and still remains, the king.
So if your platform wants to be able to service the majority of crypto clients, and is expecting hundreds of millions of dollars of volume, USDT is the easiest choice by far. It will allow the vast majority of crypto clients and volume to access your venue in a way that neither DAI, nor a US bank account, will. And it will be able to draw on the combined liquidity of the world’s top spot crypto exchanges.
Redeemability
Sure, USDT trades a lot — but is it a stablecoin? Is it $1?
In the end, there’s only one thing that makes something a real stablecoin: the ability to redeem it for $1.
All of the market liquidity is, in the end, dependent on that. If you can’t redeem a stablecoin no liquidity providers will be willing to buy large amounts at $0.98 — and so nothing will stop it from falling to $0.90, or $0.50, or $0.10.
On this front, USDC wins. You can turn $1 into 1 USDC with no fees, for infinite size, and instant deliver using its creation portal, Coinbase’s, FTX’s USD basket, and lots of OTC desks. If you’re asking for an offer on $2b of a stablecoin, someone will be willing to sell you USDC below $1.01. And if you’re asking for a bid on 50% of the market cap, someone will buy it above $0.99: because they’re confident they can create/redeem, so it’s a pure arbitrage. Try getting that with any other coin.
So there’s a very real sense in which 1 USDC = $1. How about USDT?
The answer here is…. it’s complicated.
There are fees on USDT creations and redemptions; no one will bid $1 for 5b USDT because at best they can redeem for $0.999. And unlike USDC, whose creations and redemptions take as long as an ETH block, USDT can take days. It’s not on Silvergate’s SEN, the gold standard for instant crypto USD transfers; it’s not even banked in the US. And the process is way more manual, introducing human delays.
But beyond that, questions have persisted around the funds backing USDT. Rumors circled for years that it was unbacked, and a year ago it was revealed that hundreds of millions of dollars backing USDT were were frozen by various governments in connection with a third party payment processor they used to use.
So what happens when you try to create or redeem USDT for USD?
Well, first of all you’ll pay 0.10% in fees. And second you’ll likely have to wait a day for wires to clear, and for the operation to be processed.
But does it work at all? Can you actually get USD back for your USDT?
The answer, generally, is yes. Billions of USDT have been created and redeemed over the last few years. It’s awkward and slow and expensive; and sometimes you even get pushback (which is terrifying!). But in the end you can do it, and in fact some people do, sometimes for large size.
And how about it’s banking situation? Well for the last year or two it’s been relatively stable, although prior to that Tether hopped from bank to bank. There is risk of that happening again, which could cause serious delays.
USDT isn’t perfect. But there are dollars backing it, lots of them — probably somewhere between 95% and 100%, depending on how you treat frozen funds. And the truth is, even if they’re short $500m — which we’re not saying they are! — Bitfinex generates enough revenue to cover that if it has to. LEO likely raised enough to close that gap.
So in the end you can create and redeem USDT, but it can be slow and expensive and sometimes disrupted.
And DAI? how are creations and redemptions?
Well, creations are bad. You can do it — anyone can — but it’s not cheap: you have to lock up way more than $1 for every DAI you create. DAI isn’t backed by USD, it’s backed by ETH; and because of ETH market volatility, DAI requires a lot of extra collateral.
Redemptions, though, barely exist. You can redeem your DAI back for your ETH — but only if you created it. So, sure, theoretically everyone who created and sold could buy back their DAI and unwind their ETH loans. But if Alice creates 1m DAI and sells it to Bob, Bob can’t redeem it — he doesn’t have any ETH loans to Maker! And so as DAI changes hands in exchanges and DEXes and lending pools, it loses a lot of its redeemable. It’s not just slow and expensive: for most people, redeeming DAI is actually impossible. And no one can redeem arbitrary amounts of it.
Worse, if Alice creates 1m DAI and sells, and then forgets about crypto — that’s 1m DAI that can never be redeemed, because the total remaining redemption power will fall $1m short of the circulating DAI.
There is one other way to redeem — you can start an Emergency Shutdown of Maker. But doing so requires having and potentially losing $25m of MKR, waiting for a cooldown period, shutting down the entire Maker protocol, and incurring price and liquidation risk on the assets. The only people in the world with enough MKR to initiate a shutdown are Maker itself and a16z, and someone attempting to accumulate that amount would have to buy 2.5 ADVs of MKR on short notice. Practically speaking, it is not a reasonable option.
So when it comes to creations and redemptions, there’s a clear hierarchy: USDC > USDT > DAI.
So, how does this all shake out?
Well, if you’re trying to figure out what to list, lots of things matter, with liquidity and adoption being key. Thus many exchanges choose USDT over USDC: they’ll likely get more liquidity and volume in their markets, and be able to service more customers.
But that’s not the topic of this article. This article is about risk: and particularly about DeFi risk.
DeFi, like much of CeFi, runs on stablecoins. And with the rise in the scale of the sector — particularly looking at the success Compound and others have had recently — it becomes increasingly important to understand how resilient the coins are.
There are lots of questions one could ask, but for now we’ll focus on one example: if you have a DeFi borrow/lending platform (like Compound, or Synthetix, or Maker for that matter), how scared should you be of stablecoin risk? And generally how scared should DeFi be of stablecoins blowouts, and which stablecoins are the most worrying?
Market liquidity matters here — but less so. Sure, you’ll have less market impact on $10m of USDT than USDC or DAI, but we’re not really worried about 0.50% here; all of the relevant protocols require closer to 20% margin. So which stablecoins are most likely to blow out below $0.80 or above $1.20?
Well, say that someone tries to (or is required to!) buy or sell $200m of ETH / X. What impact does that have on X? The major worry here is that a major source of demand — possibly from a liquidation on some protocol — could blow out the stablecoin, in turn dislodging half of all markets in DeFi and creating cascading liquidations.
From this perspective, USDC is pretty good. You can create/redeem essentially infinite amounts for no cost quickly. So if someone really does try to sell $20m USDC, a liquidity provider can buy it for $0.99 and redeem it all quickly — and cycle through that if they need to, buying up more and more. USDC doesn’t have the most market liquidity, but it’s resilient, because its creations and redemptions work well.
USDT is… ok. Like USDC, liquidity providers can buy up USDT if it’s crashing and redeem it. But because it can take a day (or more, sometimes) to redeem, and the total balance sheet ready to do an arbitrage is limited, it’s hard for the world to do more than $200m or so per cycle (day). And so very large amounts of USDT buying or selling can create temporary dislocations which can take weeks to fully revert. And we’ve seen this before — there have been weeks where USDT got to $0.90, or $1.10. But it’s hard for USDT to blow out too much more than that, because eventually people realize they can make 30% per day buying and redeeming, or creating and selling; and eventually the billions of dollars of balance sheet sitting on the sidelines in crypto (think exchange revenue, borrow/lending books, and VC firms) jump in. And even if Tether never recovers their frozen funds, that’s just a few percent decrease in price given their $10b market cap.
So Tether bends, but it is fairly unlikely to fully break. It’s not perfect and in a large crash — like March 12th — it might move 5–10%, adding a bit more pressure to loans. But it probably won’t blow out a sector on its own.
So what about DAI? Well, to put it frankly, DAI is scary. Really scary.
It’s a really cool idea and protocol — a decentralized dollar with no bank account necessary!
But creations are slow and expensive and require permanently locking up extra capital. On a market crash lots of people may need to sell ETH/DAI to meet margin calls — creating buying pressure on DAI. And locking up all of the necessary collateral can put a lot of pressure on crypto liquidity providers’ balance sheets, making it difficult to do too much size. And so DAI can blow out, and in fact it did on March 12th, reaching as high as $1.20 briefly. This is really scary! Lots of otherwise safe positions can blow out just because DAI does.
But even that isn’t as bad as the other direction. What happens if a set of customers are collectively lending 500m of DAI to borrow $400m of ETH. Then ETH rallies 18%. What happens?
Well, those customers are now close to bankrupt — and have to be liquidated. That means that someone has to take on their position — someone has to buy their 500m DAI in return for $472m ETH. Even if they could gather $472m ETH with no impact, that means they’re buying 500m DAI for $0.98. What can they do with that DAI?
Well, in a real sense basically no liquidity providers can reedem very much DAI. Remember you can only redeem what you created! And the largest market makers probably weren’t the ones who created it. And so the liquidity providers can’t buy at $0.98 and redeem to make money. They can’t even buy at $0.90, spend a week redeeming, and make money; they just can’t redeem. And so there’s fundamentally nothing that stops it from dropping. A lot.
How much balance sheet would the world be willing to spend on a non-redeemable stablecoin with $500m of sell-pressure and no prospects to be able to sell out? What price would they buy at? I don’t know, maybe $0.50? Maybe then someone would step in and day: “sure, I’ll pay $0.50 each for 500m DAI; that sounds like a good trade, and oh yeah I happen to have $250m of cash lying around ready to instantly deploy to a decentralized lending platform during a huge market move, I’m sure my risk managers won’t mind.”
DAI is brittle. When it bends, it breaks — and it breaks hard.
USDT, DAI, and USDC each have different types of strengths and risks, and comparing different profiles isn’t always easy. But in this case, when it comes to existential risk posed to DeFi, the ordering is pretty clear: USDC is the safest; USDT is in the middle; and DAI is the scariest.
We know that it won’t be popular with a lot of people to say that DAI is scarier and has greater crash risk than USDT, but we think it’s true.
That’s not to cast judgement on which is a better overall project, necessarily. DAI is innovative and cool and decentralized, and that’s all important! And USDT has the greatest market liquidity, which is important for everyday utility.
But at the end of the day, if a lending book falls over, it doesn’t really matter if it was because of an ETH/DAI liquidation or a bank freeze: either way it’s just bad. And when it comes to liquidity — both market depth and crash risk — DAI is a house of cards.
This is why we think DAI is one of the largest existential risks in DeFi. So much of the space is built on it — similar to how CeFi is built on USDT. And unlike ETH — DeFi’s other major building block — systems are built assuming that DAI doesn’t move.
So what implications does this have for DeFi?
- Protocols should be built such that DAI dropping to $0.50 wouldn’t cause undue systematic failure.
- DEXes shouldn’t use DAI as the sole stablecoin.
- The industry should be careful to not hold too much of their funds in DAI.
- Borrow/lending books should treat DAI to have similar downside to ETH and (w)BTC, but with significantly less market liquidity. This means requiring similar levels of collateral for small positions, and potentially significantly more for larger positions.
- DeFi likely has to accept that, where stablecoins are necessary, they are likely to going to need to be backed by a bank account. The whole point of a stablecoin is to represent $1, and nothing crypto-native will be able to fully accomplish that without being backed by fiat.
- Someone should work to build a version of DAI which is more freely redeemable.
