[Signal Request] High Leverage G-UNI Vault to Replace PSM-USDC Exposure

The Gelato Network has submitted their MIP6 application for the G-UNI token. I would like to make an argument for fast-tracking this collateral to replace a good chunk of the PSM-USDC exposure. First, I’ll explain how G-UNI works and its relation to Maker.

G-UNI is an ERC20 tokenization of a Uniswap V3 position. It works for arbitrary pairs with arbitrary ranges, but for the purposes of Maker we are going to focus on a fixed range for the USDC-DAI pair. There are many projects building ERC20 tokenizations of Uniswap V3 positions, but G-UNI is the most straight-forward implementation I have found so far. G-UNI does not have farming rewards of random tokens, admin upgradability (it technically does, but they will turn this off) or large/complex code. It makes our job much easier for doing the analysis to protect vault user’s funds.

G-UNI takes the standard approach of having an operator (Gelato Network nodes) re-invest the profit back into the pool. Node operators get 1% of the profit re-invested, so they are incentivized to do a good job. Node operators do not have access to funds beyond a portion of the immediate profit. Since re-investment will occur fairly frequently the amount of funds at risk from a malicious node operator at any given time is minimal.

My idea is to provide a standard vault with very high leverage on this DAI-USDC G-UNI pair at a fixed spread of 0.9994 - 1.0014. This range should cover most of the trades where the peg is breaking high. The specific numbers are a feature of Uniswap in that the price ranges are bucketed to certain numbers. This is actually to our advantage as the only prices available are 0.9994, 1.0004 and 1.0014 which means actively managed positions can’t really outcompete us by much.

Looking at recent Uniswap numbers, APRs on this pool are around 4% based on the 140M TVL. If we offer 100x leverage (101% CR) then this APR raises to 400% which should attract a lot more capital. This should be enough to attract a couple billion TVL into the pool (depending on the SF) which will indirectly unwind the PSM as minted DAI is sold for USDC. Furthermore with such a massive amount of liquidity, the Uniswap pool will become a great place to do trades between USDC and DAI of virtually any size. This will help to increase fees and attract even more capital.

Why not use the PSM USDC directly in the pool?

I see two advantages of not just using the USDC in the PSM:

  1. Complexity. The fee re-investment in Uniswap V3 is not trivial. It requires front-run protection mechanisms to ensure that rebalancing the two tokens doesn’t result in excessive loss of fees. So far the common strategy I’ve seen is to have a trusted operator take care of this. Building this internally at Maker will take a lot of time. By using a third party solution with external token we can minimize the amount of work required on our end.

  2. Blacklist risk. It’s not clear how much of a concern this is, but at the very least it is causing hesitation with some folks in the crypto space. By adding the G-UNI token as just another vault we no longer have USDC on our books. USDC does not touch any of the Maker smart contracts, so Circle would not be able to blacklist the Maker protocol directly. They could still blacklist the G-UNI token, but that becomes more like blacklisting a Uniswap pool which seems like a very bad PR move as there is collateral damage. As of right now the PSM USDC is entirely owned by the protocol - this change would make the vault users owners of the USDC again.

Pros

  • Unwind the PSM
  • Start earning fees on the USDC that would otherwise be sitting in the PSM
  • Blacklist risk mitigation
  • Uniswap USDC-DAI swaps can take very large trades
  • Minimal work to onboard (Standard GemJoin and Oracles are straight forward)

Cons

  • High leverage is much more risky to the protocol
  • Giving up some fees to users in exchange for blacklist mitigation
  • Takes some time that is otherwise used for other more productive collateral

Vote

Do we want to prioritize adding high-leverage G-UNI for the USDC-DAI pair?
  • Yes
  • No
  • Abstain

0 voters

Next Steps

This poll will run until Thursday July 8th, 2021. If accepted (and the MIP6 vote is greenlit) then the relevant Core Units will begin work onboarding G-UNI as a high priority item.

16 Likes

Just curious, the system already has Dai-USDC as collateral and it’s been under used and stay at 5 million. If we want a leveraged version, why can’t we lower CR for existing Dai-USDC LP instead of having Gelato one?

1 Like

This was definitely the idea with my initial proposal to lower the CR on the Uniswap V2 DAI-USDC pool. The problem is the fees for V2 are drying up. You can view the G-UNI token as analogous to the Uniswap V2 pool token, but with way more fees.

Here is a screenshot of V2 fees:

As you can see it’s way less than the 15k / day on V3.

2 Likes

Tentative yes here, with the caveat that this is an unreleased and somewhat technically complex token that will need a deep examination by the PE team.
This token should be allowed to cure on mainnet for some amount of time so that any vulnerabilities are discovered before we rush to add it to the system.

13 Likes

Ah makes sense. So would there be also a plan to vote for “vanilla” USDC Dai Uniswap v3 ?

I would love to be an enthusiastic “yes” but I’m forced to ask… who has audited Gelato’s smart contracts and where can we read these audits? Has the PE team personally audited the code and what is your confidence level in its security?

3 Likes

There is no such thing since V3 does not issue a fungible token. Specific project tokens are the best we can do for now. There is no reason we cannot onboard additional projects in the future that have different approaches, but Gelato has built what I consider to be the closest thing to what we need now.

3 Likes

What are scenarios that would lead to liquidations?

They have two auditing firms looking at the code to be done by mid-July. I have personally done a thorough audit of the code, and it is the safest implementation I’ve seen so far. Other PE members are currently reviewing as well. We will obviously not proceed without due diligence from all Core Units.

6 Likes

Impermanent loss is the only driver of losses with Uniswap. This is very relevant for pairs which do not stay at the same price (such as ETH-DAI), but not so much for stable-stable pairs and in particular DAI-USDC because we have the PSM providing hard bounds.

A loss event would occur if the price of DAI-USDC drifts too far apart, but we are already very vulnerable to that anyways.

2 Likes

Would it be possible to create a PSM-G-UNI instead of a regular vault (accounting for $1 per DAI and USDC in the underlying)?

In such a way, we would get most of the fees (actually more than the fees with the tin and tout fees), be a truly risk-free arbitrage that will be used. I understand it will be harder, but the token itself is new so there are a lot of risks and it will scale slowly and need audits.

The excess value of the collateral can be sent to the vow in a permissionless way.

1 Like

Thank you. I personally think it’s too early to rush and fastrack it when audits are not yet done. Voting no for now (won’t make much difference though haha)

1 Like

Seems possible, but I was looking for a fast path to get something online in a reasonable amount of time without occupying too much PE time that can otherwise be spent on high quality collateral such as SushiSwap LP tokens, institutional vaults, sETH, etc.

The audits should be done right around when this poll goes on chain. I’m not suggesting we put it up tomorrow.

1 Like

Hey @hexonaut, some rather basic questions from me - pardon the lack of knowledge on this topic.

When you say loss event, what exactly does that mean? If we were putting the NFT of USDC-DAI LP into the vault and the peg broke upwards, this would simply mean that we’re left with 100% USDC but that’s not exactly a “loss” for the protocol. What is different with G-UNI when the peg breaks?

If the pool gets this large, doesn’t the APR go way lower making the SF of 1% unsustainable?

And lastly, I’ve had this question even about the v2 USDC-DAI LP tokens so I may as well ask here. Doesn’t leveraged use of these vaults effectively make DAI unbacked? If someone did utilize the 100x leverage on these vaults and DAI’s value crashed, the LP token collateral would lose value at the same time. How does it make sense under an emergency shutdown?

Yes sorry, I’m talking about a case where you earn a negative ROI. This may or may not translate protocol losses depending on the CR. 101% is my example, but it could be 102% or higher if we need more protection from IL.

Ah yes you are correct. I messed up the calculation. Something like 0.1-0.25% would be more appropriate for billions of TVL at the current APRs. It is still achievable.

I’ll think on this some more, but I don’t believe it does as it would similar to adding a vault for DAI at high leverage. Obviously this makes no sense, but it doesn’t make the newly minted DAI unbacked. It’s backed by the original collateral deposit that minted the “first DAI”. The DAI used as collateral is not accessible to anyone. This is not identical, but similar to the DAI-USDC LP backing.

2 Likes

You are describing something similar to a cash secured loan. It’s sitting somewhere earning something, but can be yanked if the collateral is necessary. So it’s not a unusual as it sounds if I understand it correctly (which I may not)

1 Like

Perhaps I should think about this more too but these do not seem to be the same thing to me. DAI vaults are not a problem as you correctly explained but USDC-DAI LP token vaults are (even without leverage).

For instance, I put in an LP token worth (1USDC + 1DAI) as collateral and mint 2DAI with it. If DAI goes below peg, the collateral transforms from (1USDC + 1DAI) to 2DAI. If we now perform an emergency shutdown, the 2DAI in the vault is worth less than 2USD (because DAI is below peg) but it backs 2DAI. In the extreme case where DAI value goes to zero due to external factors, the vault collateral also goes to zero simultaneously leaving DAI completely unbacked. Leveraged use of such vaults makes this problem even worse.

Now that I think about it, maybe all xxx-DAI LP tokens used as collateral have the same issue?

Yes we are making assumptions that USDC and DAI equal $1 USD which is done elsewhere already. I’ll admit I’m not an expert on all the edge cases of this assumption, but it’s already being done so I don’t think there is any danger beyond what risk we already have.

1 Like
  1. At 140M TVL earning 15K/day this gives a return of 3.8% yearly ROI (non-compounded).
    10x this liquidity and this return drops to .38%. This gets worse if you think many billion (3B) would be required. I don’t see this commanding a high SF on TVL. Literally the SF has to be significantly less than the return per capital cycling which is going to eat into the returns. In effect the SF will have to be lower than the return otherwise people will be losing money and want to unwind. Given SFs are manually controlled there in effect is a rate risk here that multiplies with leverage.

  2. Untested protocol.

  3. Implications for the protocol if the markets for some reason go out of this band one way or another.?! (i.e. the two basic edge cases).

  4. liquidation risk? Exactly how are the oracles to be structured and where is the penalty via fees?

On face value we could get more liquidity, though I don’t see how we hand off blacklisting risk here (seems like it is just transferred into another entity)

How does this unwind the PSM? Why buy DAI at 1.001 USDC when one can just borrow some DAI using some other collateral and lever up with USDC using the vault?

You suggest this will unwind the PSM but not replace it. Can you explain how this works to unwind the PSM and what then will replace the PSM?

1 Like