Wednesday, June 05, 2024

Short Perps to Create a Stablecoin: Ethena

 In March 2023 Arthur Hayes wrote a blog post about how to create a better stablecoin. Stablecoins are essential for people who want to transact in fiat, a way for people to deposit into and out of crypto exchanges, and as the numeraire for various trading pairs. On-chain stablecoin provider MakerDao requires considerable over-collateralization (150%) and is not genuinely decentralized or immutable (e.g., they helped Jump Capital retrieve their stolen ETH on-chain via an upgradable contract in an allied protocol). Off-chain stablecoins are subject to regulatory interference and can and have confiscated funds. Many large, centralized exchanges (CEXes) rely on stablecoins, putting the largest CEXes in a precarious state where regulators could use Circle (USDC) or Tether (UST) to attack these exchanges.

Hayes promoted the idea that a stablecoin backed by a short crypto position on a perp exchange. As perp short positions tend to make 11% annually, so if one took $3700 of ETH, deposited it as collateral, then shorted ETH, this would be like a $3700 bank account earning 11%. One needs no over-collateralization because there is no risk in the position, as the net ETH delta is zero. For example, consider where someone deposited 1 ETH at $3700 and instantly shorted 1 ETH on the perp exchange. While the notional of the short position is $3700, the account’s net market value is the sum of their collateral, 1 ETH, and the mark-to-market on their short position.

No matter where the ETH price goes, the position is solvent, as the net value of this position is $3700 (excluding funding revenue). The deposit mustn’t be a long perp position, but instead, real ETH; otherwise, it would simply pay what their short position receives. One can issue the owner of this position $3700 worth of a stablecoin as a liability with the same security as if one deposited $3700 in a bank and can spend $3700 using a check.

While this could work with decentralized perps like GMX and DyDx, those markets are not trustworthy for major players, so they are relatively small compared to the trustless AMMs. CEXes have 25 times more perp open interest, and the CEXes that are most popular are in a regulatory grey zone and desperately need a more secure stablecoin alternative. Thus, the CEXes should be eager to help create specialized custody arrangements where the stablecoin issuer would deposit their ETH or BTC. These CEXes would then know these perp short positions are fully collateralized, and the stablecoin issue would have a more secure position at these CEXes. This special treatment would serve both parties well.

Another thoughtful Hayes insight is that by using several CEXes, such as Binance, OKX, etc (all non-US), regulators would have no singular attack point (see this podcast interview). This is essential, and Hayes mentioned that a single exchange issuing a stablecoin would be a problem because they are all vulnerable to regulatory pressure when isolated. With several exchanges in various regulatory domains, all a regulator would do is move more crypto activity out of their sphere of influence.

So, Guy Young read the post, set up a team, and created Ethena (stable ticker USDe, traded most on Curve). Young added a couple of wrinkles, such as using staked ETH as collateral, but basically, it’s based on harvesting the short perp financing rate. Hayes is one of the principal backers, and surely Hayes introduced him to key people at the various CEXes essential for the model to work. This approach is not decentralized, but as a practical matter, most dapps have centralization points. Usually, these centralization points are for emergencies, like recovering Jump Capital’s stolen ETH, but we know emergency powers can be abused (e.g., Chancellors Hitler or Palpatine). While it helps to be as decentralized as possible, as a practical matter, when there are billions of dollars at risk, giving someone with the right incentives the ability to stop all transactions on a protocol and potentially retrieve assets is a preferred solution most people in crypto are comfortable with.

As Kyle Samani noted when promoting Solana, most people care more about speed than decentralization. However, there is a critical limit here. Low latency blockchains and dapps are nice for inconsequential sums, and so naturally lend themselves towards trader bros willing to spend $1000 trying to get in early on the next Dogwifhat meme coin. Bitcoin and Ethereum are the big players precisely because they are the most effectively decentralized and, thus, the best place to store wealth in case of a storm (the Nakamoto number is a lousy metric). Any chain or dapp promoting their near sub-second speed is ephemeral, as they must be centralized to work.

Often these are just fleeting regulatory arbs, like EtherDelta or DyDx, and the regulators have or will shut them down by offering their CEO an offer he can’t refuse (apply KYC to everyone or go to jail). More commonly, the lack of trust in these shuts out the big money needed for an Ethereum-sized chain, leading to a market cap peak around one to ten billion dollars. Without token appreciation, the token rewards program and staking programs become unattractive, and then everyone cashes out and moves on to the next pump and dump (the Axie Infinity life cycle).

Satoshi’s cypherpunk principles—immutability, permissionless, censorship-resistant, transparency, decentralization, pseudonymity, and cryptographic security—are essential properties of the blockchain but do not need to be applied at all levels. The blockchain must be completely decentralized, but that does not mean everything on it has to be. As long as there is open entry and exit by dapps and bridges on the blockchain there is decentralization via competing centralized agents. This is how decentralized markets work. It is a category error to insist the characteristics of the whole must apply to the parts.

Ethena’s endgame is limited to being a player among several, anyway. This is about $24B and $13B in Bitcoin and ETH perp open interest, respectively. As only half of that can be short, and Ethena would not want to be more than half the market, that’s about $10B total in capacity. Ethena reached $2B in stablecoins outstanding faster than any other stablecoin. While this is great, I hope they have the humility to be satisfied with a $10B stablecoin instead of trying to capture a majority of this $150B market. The short-perp funding angle is good, but there are probably many others, and they should be left to other specialists to make them easier to monitor.

An interesting aspect of this strategy is how the funding rate will respond to the new short demand. Traditionally, customers are net long, so exchanges pay off their reluctant shorts by fixing the funding rate via the perp premium mechanism (see earlier posts here and here on why the official perp-premium financing rate mechanism is a fairy tale). I reached out to retired econ professor Adam Gehr, who wrote the seminal academic paper on perpetual futures markets in 1988, the one referenced by Robert Shiller. Shiller’s paper is more commonly referenced, but that’s merely because Shiller has a Nobel Prize, which many find compelling. Gehr told me Shiller contacted him a few years ago to say that he thought the crypto perp premium funding mechanism was more like the model in Gehr’s paper, as clearly it was unlike his own mechanism. Other than that, Shiller had nothing to say. Gehr just thought the perp premium mechanism was odd and uninteresting. Gehr agreed with my take that this mechanism is profoundly different than his approach, which involved a separate session after the market close to equilibrate the funding rate, nothing like a continuous average of a perp/spot price ratio (traders looking for 0.1% price discrepancies, are on average flat when they go home, so they funding rate doesn’t matter as it all washes out).

The funding rate is necessary because of the supply and demand imbalance for leverage. The funding rate equilibrates this well as there are naturally ten times as many wanting to be short as to be long. However, the mechanism used is an anachronism from the days before stablecoins. With only Bitcoin going into and out of BitMex, how could a trustless market give users trust that perp prices were trustless? The perp price premium story was perfect because it was plausible, and as a practical matter, a market based on Shelling points works. As perp exchanges like GMX and Synthetix use the customer net long demand and work fine, it is clear this mechanism is not required.

For years, Binance et al. have set their funding rates via this indirect mechanism. While it makes them money, it is probably a hassle. Eventually, outsiders may want to look at their trading tapes, which would expose an insider conspiracy. Thus, it is essential that if Ethena takes over this role as the major short provider, the price setters work for them. Arthur Hayes is the OG of perps, and he should be aware of this issue and handle it appropriately. It will be tricky, however, as it is like the Deep State, something that doesn’t officially exist. Now would be a good time for everyone to transition to something that doesn’t require institutional lying (lying on little things leads to lying on more significant things).

Hayes seems sharp, but who knows how much he knows? I can’t tell if he really believes in his perp-premium funding rate mechanism or if he knows it was just good cover to get BitMex’s perp markets off the ground. Often, executives are unaware of the shenanigans required to implement their vision. An excellent example of a disaster here was when created their market, a bunch of outsiders acquired a net long position; that group set the perp price premium so the longs would receive the funding rate, as the guys simply assumed the perp premium worked as advertised, without any contrivance. Naivete is costly on the blockchain.

Ethena can succeed, and I hope they do, but there are significant risks as they have only a $10MM insurance fund with $3B in stablecoins outstanding. A negative funding rate should be easy to manage, but the ‘operational risk’ could easily sink them. A bad bridge or a couple of rogue employees costing them a mere 1% of gross assets would make them insolvent (risks are nicely outlined here). They would benefit from slowing down and increasing their insurance fund. Currently, their assets yield 20% via the short financing and staking yield, but as most users do not stake their USDe stablecoin, that gets multiplied into a return of 40% for USDe stakers (see Ethena’s dashboard here). Returns above 20% are sufficient for growth, so they can afford to pay out less, especially at this early point in their life cycle. Stakers are residual claimants, so they would ultimately own this money indirectly anyway. The insurance fund is the book value of a company, useful as a cushion for hard times.

No comments: