Ethereum Compound v3 USDC - Gauntlet IR Curve Recommendations (5/16/23)

Simple Summary


Gauntlet recommends the following changes to the Ethereum v3 USDC IR curve:

  • Increase Annual Supply Interest Rate Slope Low from 0.0325 to 0.034
  • Increase Supply Kink from 80% to 95%
  • Increase Annual Supply Interest Rate Slope High from 0.4 to 0.76
  • Decrease Annual Borrow Interest Rate Base from 0.015 to 0.01
  • Decrease Annual Borrow Interest Rate Slope Low from 0.035 to 0.0333
  • Increase Borrow Kink from 80% to 90%
  • Decrease Annual Borrow Interest Rate Slope High from 0.25 to 0

Current vs Proposed IR Curves

Utilization vs APRs

Utilization vs Daily USDC Reserve Growth (holding current $185M USDC borrows fixed)

Currently the protocol is facing the following problems:

  • Stagnant protocol growth/migration and poor borrower UX
  • Negative net protocol profit (USDC reserve growth - COMP distributions)

The proposed IR curve attempts to solve these problems as follows:

  • Flatline Borrow APR past the borrow kink
  • Adjust borrow/supply kinks & slopes to incentivize efficient and appealing equilibrium APRs


Problem 1: Stagnant protocol growth/migration and poor borrower UX

As seen in the above chart, USDC borrows in Ethereum v3 USDC stagnated around $220M from 4/27/23 to 5/11/23, and decreased to $185M by 5/15/23.

The current IR curve is shown above. The borrow kink occurs at 80% utilization, which corresponds to 4.3% Borrow APR. The slope of the borrow interest rate to the right of the kink is 0.25, which results in borrow APR rapidly increasing at >80% utilization.

The most recent proposal to increase COMP rewards for USDC borrowers was executed on 4/5/23, when equilibrium was roughly 66% utilization. A large borrower then joined the protocol on 4/14/23 to take partial advantage of the rewards, increasing utilization to 80%. However, as indicated above, borrowers are empirically unwilling to borrow substantially past the 80% utilization kink. This causes stagnation and prevents the protocol from utilizing the remaining 20% (~$50M) of borrowable USDC.

Borrowers are so averse to borrowing past the kink that even appealing Net Borrow APRs do not incentivize them to do so.

During the rare times when a borrower does decide to borrow slightly above the kink, borrow APR increases for all the other borrowers in the protocol, at times over 6%. This results in higher Average Borrow APR and poor borrower UX.

On 5/11/23, a USDC borrower increased their borrow position, thereby increasing utilization to 86% (5.8% Borrow APR). Shortly after, a different borrower with address 0x0cc5a715b06c8ece9f2fa49cdc75750539255779, who previously borrowed $15.2M USDC, completely closed their borrow position, and in fact switched to supplying $3.5M USDC. Therefore, this high variable borrow APR above the kink can result in borrowers closing their positions, despite appealing Net Borrow APR.

Problem 2: Negative net protocol profit (USDC reserve growth - COMP distributions)

The above chart shows Daily Net Protocol Profit at different utilization levels under the current IR curve, holding the current $185M USDC borrows fixed. Even at the current most efficient utilization level of 80%, the protocol accumulates $5,320 of Daily USDC Reserve Growth, loses $16,850 of daily COMP rewards, for a best-case Net Daily Protocol Profit of -$11,530, which essentially results in the protocol swapping COMP for USDC at 68% slippage at best.

Above is an hourly time series since 3/1/23 of realized USDC reserve growth relative to COMP distributions. As we see, the protocol sometimes effectively swaps COMP for USDC at 100%+ slippage.

As seen above, since 3/1/23, the protocol has accumulated ~$200k USDC reserves, lost ~$1.2M COMP rewards, for a Net Protocol Profit of -$1M, essentially swapping COMP for USDC at ~83% slippage on average.

Therefore, Compound could improve their allocation of COMP rewards relative to the TVL and IR curves.


Proposed IR Curve Feature 1: Flatline Borrow APR past the borrow kink

Setting a max Borrow APR even at high utilizations results in borrowers always being incentivized to join the protocol, regardless of utilization. This avoids stagnation and improves efficiency, allowing borrowers to takefull advantage of all the borrowable USDC and the appealing Net Borrow APR.

This change also gives borrowers long-term peace of mind knowing their max rate is capped regardless of utilization fluctuations, thereby improving UX, reducing existing borrower outflow, and further incentivizing borrowers to join.

As seen in the above chart, USDC suppliers quickly join the protocol when supply APRs spike above 5%. Therefore, under the proposed IR curve, even though the protocol would incur short-term losses at high utilizations due to high supply APR and low borrow APR, equilibrium should reestablish quickly, and the greater long-term reserve growth should outweigh the short-term losses.

Also, the long-term reserve growth as a result of greater borrows would mitigate and possibly entirely overtake the relative COMP distribution losses, thereby resulting in positive Net Protocol Profit.

Proposed IR Curve Feature 2: Adjust borrow/supply kinks & slopes to incentivize efficient and appealing equilibrium APRs

With borrowers now incentivized to borrow past the borrow kink, the protocol can increase its intended equilibrium utilization range to offer more appealing supply and borrow APRs, thereby further incentivizing both suppliers and borrowers to join the protocol.

The largest single-user v3 USDC borrower borrows $27M USDC. Given the current USDC supply of $266M, 90% utilization would still result in ~$27M borrowable USDC. And borrowable USDC will only increase as TVL increases. So an intended 85-95% utilization range is reasonable.

Therefore we should ensure:

  • USDC borrowers are more incentivized than suppliers to join the protocol at ≤ 85% utilization
  • Both USDC suppliers and borrowers are incentivized to join the protocol at 90% utilization
  • USDC suppliers are more incentivized than borrowers to join the protocol at ≥ 95% utilization
  • The IR curve yields highest positive reserve growth between 85%-95% utilization

To determine which APRs to set at these benchmark utilizations, we should first analyze empirical behavior of suppliers and borrowers at various historical utilizations and APRs.

Note that Borrow APR is not equivalent to Net Borrow APR due to the varying COMP borrow distributions, and borrowers have empirically been unwilling to borrow past the kink. On the other hand, the protocol has never allocated COMP distributions to USDC suppliers, and a supply kink still remains in the proposed IR curve.

Therefore it’s difficult to speculate on exactly how much more/less borrowers will be incentivized vs suppliers under the new IR curve, and it will most likely require refinement in future proposals after analyzing how users respond to this change.

Supply APR Analysis

USDC supply in the protocol has consistently increased since the USDC depeg on 3/11/23, even when USDC supply APRs were as low as 2.1%. Interestingly enough, the rate of USDC supply has decreased of late despite higher supply APRs. The Average Supply APR since 3/1/23 is 2.55%. As mentioned before, suppliers are quick to join the protocol when supply APR spikes above 5%.

Borrow APR Analysis

In the few weeks prior to the most recent COMP distribution migration on 4/5/23, utilization was roughly around 66%, corresponding to a Borrow APR around 3.8%. After the migration, utilization increased to 80% and has since fluctuated slightly above and below the 80% kink (corresponding to 4.3% borrow APR), with borrows stagnating and regressing due to reluctance of increasing borrows above the kink. The Average Borrow APR since 3/1/23 is 4.13%, and the Average Net Borrow APR since 3/1/23 is 0.34%.

Proposed IR Curve

The proposed IR curve is shown above.

Utilization Supply APR Borrow APR USDC Reserve Growth
80% 2.72% 3.66% 7.21%
85% 2.89% 3.83% 11.24%
90% (borrow kink) 3.06% 4.00% 14.94%
95% (supply kink) 3.23% 4.00% 14.94%
100% 7.03% 4.00% -75.88%

As shown in the above table, the proposed IR curve ensures the following:

  • USDC reserve growth is positive within the desired 85%-95% equilibrium range
  • USDC borrowers are incentivized to take full advantage of all borrowable USDC
  • USDC suppliers are greatly incentivized at ≥ 95% utilization to join the protocol and decrease utilization back down to a level of positive USDC reserve growth

Additionally, compared to the realized 2.55% Average Supply APR and 4.13% Average Borrow APR since 3/1/23, the proposed IR curve offers more appealing APRs for both suppliers and borrowers at the intended utilizations, thereby incentivizing protocol growth.

The proposed Borrow APRs are also set to levels which should still be appealing to users even if COMP borrow distributions are diluted or reduced, which allows the protocol optionality in the future to reduce COMP emissions/losses without greatly affecting borrower UX, and eventually achieve positive Net Protocol Profit.

The specific changes required to create this proposed IR curve are:

  • Increase Annual Supply Interest Rate Slope Low from 0.0325 to 0.034
  • Increase Supply Kink from 80% to 95%
  • Increase Annual Supply Interest Rate Slope High from 0.4 to 0.76
  • Decrease Annual Borrow Interest Rate Base from 0.015 to 0.01
  • Decrease Annual Borrow Interest Rate Slope Low from 0.035 to 0.0333
  • Increase Borrow Kink from 80% to 90%
  • Decrease Annual Borrow Interest Rate Slope High from 0.25 to 0

Potential risks of the proposed IR curve

Greater USDC reserve losses at ≤ 72% utilization and ≥ 98% utilization

Above is a comparison of Utilization vs Daily Reserve Growth between the current and proposed IR curves, holding the current $185M USDC borrows fixed. The proposed IR curve has greater USDC reserve losses at ≤ 72% and ≥ 98% utilizations. However, as seen below, these losses are still minimal relative to the daily losses in COMP distributions.

Limited USDC supplier withdrawals at high utilizations

The proposed IR curve incentivizes higher equilibrium utilization and can incentivize large borrowers to max out utilization at 100%. Existing USDC suppliers should theoretically be incentivized to remain in the protocol at higher utilizations and corresponding higher supply APRs. Existing and new USDC suppliers should be also be incentivized to increase supply to the protocol to free up USDC liquidity. However, USDC suppliers for whatever reason may still want to withdraw their supply at high utilizations, but would be limited or restricted from doing so, which would result in poor supplier UX.

Additional Recommendations (if possible)

  • Add a USDC supply cap and dynamically set it to an amount which results in 72% utilization, to avoid excess USDC supply and negative reserve growth.
  • Transfer $10M v2 USDC reserves to Ethereum v3 USDC. If the protocol experiences sustained utilization of 100% for 24 hours with no new suppliers entering the protocol (despite high supply APR enticing new suppliers to join), allocate $5M reserves to USDC supply to decrease utilization, which will decrease reserve losses and increase borrowable USDC. Adjusting desired utilization dynamically via reserves is much quicker, direct, and stable than adjusting COMP distributions, which are sensitive to TVL & COMP price fluctuations.

Next Steps

  • Targeting on-chain vote on 5/22/23

By approving this proposal, you agree that any services provided by Gauntlet shall be governed by the terms of service available at


Thanks for the proposal, @nlord. The charts and your explanations of them are excellent!

To my knowledge, Compound has never had a USDC liquidity crunch, so I think we have some room to play with. Currently, 28% of USDC supply isn’t being utilized. This unutilized capital is a buffer for those wishing to withdraw (or borrow) USDC quickly. Having a too-large buffer is capital inefficient, but on the other hand, having one that’s too small is problematic. While shrinking the buffer will improve efficiency, avoiding a liquidity crunch is essential.

I agree that capping the borrow APR at 4% will improve the UX of borrowers, but I worry about what would happen during edge-case scenarios where utilization remains at 100%. We don’t want depositors to panic if they think they’ll have trouble withdrawing their USDC because that could lead to a mass exodus. While governance could change the parameters if such an event occurs, the seven-day governance period adds some challenges.


Really appreciate the feedback @TylerEther . We agree that the 100% utilization liquidity crunch is one of greater risks in this proposal. Some things we’d note are:

Even at 100% utilization, suppliers are allowed to withdraw their USDC if there are ample reserves in the protocol

At 100% utilization, USDC suppliers are still able to exit their positions by essentially “swapping” their supplied USDC with USDC reserves. However, there are currently only $1.26M USDC reserves in Ethereum v3 USDC, which is minimal compared to the largest USDC supplier who supplies $40M USDC. Therefore transferring $10M USDC reserves from v2 to v3 would allow for the protocol to still incentivize high utilization, while also allowing suppliers to withdraw a considerable amount of USDC even at 100% utilization.

A case could be made to swap some of the $21.88M v2 DAI reserves (and USDT/ETH/WBTC reserves) to USDC and transfer an even greater quantity of USDC reserves to v3 for this use-case, and make it clear to suppliers that the protocol offers reserves as backstop liquidity.

Also along these lines, if possible we recommend the protocol dynamically supplying its own reserves if the protocol experiences sustained utilization of 100% for 24 hours. This would also be a great way for the protocol to gain yield on its reserves and free up liquidity for suppliers and borrowers.

A mass exodus can’t occur at 100% utilization

If USDC suppliers exit the protocol at 100% utilization by draining the USDC reserves, then the remaining suppliers would be unable to leave the protocol at 100% utilization. This would be poor supplier UX, but wouldn’t cause an immediate mass exodus.

USDC is not a collateral asset in v3, suppliers only gain to earn yield

Since USDC is not a collateral asset in v3, the protocol does not have to worry about USDC suppliers unable to be liquidated at 100% utilization, so the v3 USDC comets can incentivize higher utilizations than other protocols which offer USDC as collateral.

Additionally, since USDC can’t be used as collateral, USDC suppliers only deposit USDC in order to earn yield (on average 2.55% APR since 3/1/23). Under the proposed IR curve, during edge cases when utilization reaches 100% utilization, suppliers will be earning a very high 7.03% APR, and should only be incentivized to withdraw their supply in the case of a depeg or for their own liquidity needs.

Given that large USDC suppliers empirically supply their USDC in Compound for extended periods of time without any withdrawals, we expect the vast majority of users would continue to be happy continuing to supply their USDC in the protocol at high utilization while earning high yield. Additionally, in the past it usually takes no longer than a couple days for new suppliers to enter the protocol at such high supply APR, thereby quickly increasing liquidity.


Hey @nlord, thanks for the detailed response. Transferring some USDC reserves from v2 to III will help prevent a liquidity crunch.

If USDC suppliers exit the protocol at 100% utilization by draining the USDC reserves, then the remaining suppliers would be unable to leave the protocol at 100% utilization. This would be poor supplier UX, but wouldn’t cause an immediate mass exodus.

I’ve noticed scenarios at other lending protocols where utilization was 100%, and people could not withdraw their tokens. In these scenarios, people worry, or even panic, and wait around to withdraw their tokens as soon as possible. This is what I mean by a mass exodus.

While having a high APR for depositors at 100% interest may help attract lenders, we often see the market slow to respond. Having borrowers pay more in interest at 100% utilization is much stronger than paying depositors more, as it’s a difference between the real cost to borrowers vs. the opportunity cost to lenders.

It would be nice to have an additional kink at ~98% as an additional mechanism to prevent 100% utilization. But since we don’t have this feature, maybe there’s a configuration between the current one and the one proposed that could be an improvement. Although I have to admit, the idea of having stable interest rates around the optimal utilization rate sounds appealing.

1 Like

Great points again @TylerEther

We similarly considered setting a higher Borrow Kink with a high Annual Borrow Interest Rate Slope High as a hybrid approach to the current and proposed IR curves. Longer term we plan on recommending this type of hybrid IR curve when the protocol has greater TVL.

The reason we decided to move forward with the more aggressive approach for now is because we view protocol growth and especially migration as highest priorities in the short term. As we mention here, despite two separate v2 → v3 COMP reward migrations, none of the top v2 non-recursive stablecoin borrowers have migrated over. There’s currently ~$72M borrowable USDC in v3, and the highest non-recursive v2 borrower borrows $74M. We want to incentivize the large borrowers from migrating over as much as possible, so flatlining at the kink removes any chance of dissuading borrowers due to variable borrow APR and high max borrow APR.

In the case that the utilization is maxed out at 100% and suppliers don’t quickly join the protocol, then we will certainly look into more quickly proposing the hybrid approach and/or allocating rewards to suppliers. However, for this to happen, it means the IR curve change would have incentivized borrows to increase from $185M to $257M (+39%) without increasing rewards, which would be a great sign. This also assumes no increase in supply, which seems unlikely given that supply APRs will be even more appealing at the current 72% utilization level, much less 100%.

If we look at the Polygon USDC comet, USDC suppliers flocked to the protocol very quickly after rewards were recently added to the supply side, indicating USDC suppliers are quick to join when incentivized.

1 Like

On-chain proposal has been published below:

1 Like