Post mortem — USDC depegging effect on Gearbox lending
Overview
Over the weekend 10–12 March 2023, the DeFi ecosystem faced an unprecedented stress test as centralized major stablecoin USDC faced extreme pressure: Circle, the issuer of USDC, had a whopping $3.3 billion deposited with SVB, which represents around 7.5% of the total USDC market cap. As the market scrambled to process this development, USDC started to drift away from its crucial $1 peg. The fallout extended to other stablecoins like DAI and FRAX, all of which are partially backed by USDC.
As traders flocked to Curve to swap USDC & DAI for USDT, the pool balance got meaningfully distorted with USDT’s share falling to a fraction.
As traders rushed to dump USDC and DAI, USDT shot up to $1.15 on Curve.
USDC & DAI
At the same time, DAI’s 3pool liquidity was also severely depleted. DAI, being an overcollateralized stablecoin, is backed by various collateral assets, with USDC comprising approximately 50% of it.
In response, MakerDAO proposed an executive vote to pause the Peg Stability Mechanism (PSM), which mints DAI against centralized stablecoins such as USDC, GUSD, and USDP. The goal was to insulate DAI from the price fluctuations of these centralized stablecoins during times of instability.
USDC’s dominance in the 3pool was further highlighted by the USDC/DAI price, which dipped to 0.96 before bouncing back to $1.
Liquidations and Slippage
The 4% slippage for USDC/DAI began at $166 million and fell to a low of $15 million, while the 4% slippage for USDC/USDT started at $219 million and plummeted to a mere $1.5 million.
Gearbox
Gearbox markets include a variety of stable coins, such as USDC, DAI, USDT, FRAX, GUSD and sUSD, and their derivatives, curve LP tokens and yearn vaults.
Currently, the biggest leveraged asset is the sUSD curve LP token, with over $25m that is used as collateral.
Over the past weekend, together with the rest of the Gearbox community, we monitored the situation as it was unfolded.
sUSD LP token is composed of sUSD, USDT, DAI, and USDC. And its price oracle simply returns the minimum price of these underlying assets, multiplied by a virtual price, which for the purpose of this analysis can be assumed to be constant.
Hence, with the price drop of USDC and DAI, the price oracle value of the sUSD LP token also went down.
The only stable borrow markets on Gearbox are USDC and DAI (and in particular not USDT). Naturally, the USDC price drop could not trigger any liquidations in the USDC debt market, as the price of the debt asset (USDC) and the collateral asset (sUSD LP token) moved together.
Moverover, when USDC price was at its bottom, the sUSDC LP pool was composed from over 80% of USDC. And thus a liquidation of the asset could have been easily handled (as both debt and collateral are USDC).
Hence, our main focus for the DAI vs USDC price movement, and in particular the available DEX liquidity of DAI vs USDC.
Because of MakerDAO’s PSM, the price of USDC and DAI were 1:1 for the most of the weekend, however, the PSM has a daily minting limit of $950m, and when it was fully utilized, the price of DAI spiked all the way to 1.04 USDC.
The exposure of the sUSD market to the DAI debt pool is around $11m, and with Gearbox’s 4% liquidation bonus, there was always more than that amount available for purchase for sale.
So is everything ok?
Gearbox’s liquidation threshold ratio and liquidation bonus are relatively conservative for the stable vs stable lending market (with comparison to Aave’s and Compoud’s v3). In retrospect, these parameters provided a sufficient buffer to facilitate the price drop without any major issues. But of course, had things continue to deteriorate, then things would eventually break.
Liquidations in other platforms could have further drain the available curve liquidity, and further price drop would also drain curve liquidity, as the stable swap invariant offers low liquidity when the price is off peg.
Of course Curve is not the only source of liquidity, and especially as the price is off-peg, we would expect other venues and active market makers to take a more prominent position.
Moving forward
There is always a tradeoff between the security buffer that the liquidation threshold provides and the usability of the platform.
The current thresholds were set based on data that does not include the last weekend events.
In retrospect they were proved to be sufficient, and thus, there was no need for any emergency operation (like freezing the markets), however, there is no guarantee that they will also be sufficient in the future, and the possibility to take speedy actions in a real time manner is limited.
Hence, giving the new uncertainty in the market, we are updating our models, and will soon present to the community updated recommendations.
