[ARFC] USDT GSM Bucket and Exposure Cap increase


title: [ARFC] USDT GSM Bucket and Exposure Cap increase
author: @TokenLogic
created: 2025-14-01


Summary

This publication proposes increasing the USDT GSM Bucket and Exposure Caps to 24M and 16M respectively.

Motivation

Over the past 24 days, the USDT GSM has rapidly filled from 0% to its full capacity. Notably, in the last 24 hours alone, $3.4M USDT was deposited, bringing the module’s total holdings to the maximum exposure cap of $8M.

image
Ref: Aave Analytics | TokenLogic

The rapid shift in market conditions underscores that the current $8M exposure cap is insufficient to meet user demand. Increasing the cap would enable the DAO to better accommodate market participants while further reinforcing GHO’s peg.

Liquidity Dynamics

The recent inflows can be largely attributed to volatility in the USDT price evident on Coingecko.

The volatility resulted in the Balancer 3pool rebalancing in favour of USDT and GHO. The zero fees from USDT to GHO via the GSM, supported USDT to GHO to USDC routing via aggregators.

image
Ref: Balancer at 11:28 GMT 13th January 2025.

Screenshot 2025-01-14 at 15.17.59
Ref: https://www.coingecko.com/en/coins/tether

The price volatility was limited to USDT and as a result, no new USDC inflows into the GSM. However, with USDT still exhibiting slight volatility, the 3pool has experienced a minor imbalance, shifting liquidity from GHO to USDT, which now constitutes 48.2% of the pool. This indicates the potential for significant USDT inflows into the GSM.

Peg Stability and Market Resilience

The USDT GSM allows users to exchange GHO for USDT for a 0.20% fee, reinforcing GHO’s peg stability and providing a robust defense against market volatility.

Increasing the Bucket Cap and Exposure Cap enables the GSM to Mint GHO upon receiving addition USDT deposits, strengthening the stability mechanism and further safeguarding GHO’s value during turbulent market conditions.

Additionally, with GHO’s Borrow Rate reduced to 9.00% and the current GSM fully utilized, GHO is well-positioned to support a significant expansion in supply.

StataUSDT GSM Migration

Using the ERC4626 GSM created by Aave Labs, a new USDT GSM is to be proposed that enables the Aave DAO to earn the aToken yield from USDT deposits into the Core instance of Aave Protocol.

This transition will enhance capital efficiency, diversify GHO’s collateral base, and unlock additional revenue opportunities.

Revenue Projections: At the current $8M cap, the USDT GSM would generate approximately $565k in annualized revenue. Doubling the cap to $16M could elevate revenue to an estimated $1.13M.

Specification

The GHO Stewards are to perform the following parameter adjustments:

Description Current Proposed Change
GHO Bucket Capacity 12M 24M 12M Up
USDT Exposure Capacity 8M 16M 8M Up

This proposal is to be implemented by the GHO Stewards.

Disclosure

TokenLogic does not receive any payment for this proposal.

Next Steps

The GHO Stewards will implement this proposal.

Copyright

Copyright and related rights waived via CC0.

2 Likes

The transaction for this proposal has now been executed, by the GHO Stewards.

The ARFC was executed quite quickly since the inception of the post, we should consider if these Risk Steward actions are actually ARFCs, given it leaves very little time to comment or discuss (understandably as the delegated facility of Risk Steward actions need to act timely) and in most cases wouldn’t probably need to be subject to a community discussion given the delegated mandate.

On another note and discussion with @tokenlogic, worth considering whether the 0.2% fee is worth removing. My personal view is that GHO should have as minimal barriers as possible for arbitrageurs. If arbitrageurs know that they can mint and burn GHO without fees, they would be more open to hold GHO since they can convert it back without fees.

The downside is that less of the GHO is routed trough aggregators from dexes (creating less fees for GHO pools). In the future this can be improved by introducing a fee manager that allows to whitelist addresses (such as dex aggregators) to route swaps with custom or zero fees and creating a bd opportunity for the Aave DAO. However, for now, I’d see that arbitrageurs would most likely anyways use these pools to swap into other stables as well part of wider strategies, or when there is not enough USDT in GSM, they would naturally route trough GHO pools as which is what is most likely happening when the caps are reached as in this case.

GSMs do not need of course to run without any revenue as there are few proposals already highlighting to allocate surplus assets into BUIDL, Aave and other similar initiatives as GSMs were originally built to have capital allocation features in mind.

For comparison USDS PSM on Sky does not have fees at the moment.

As GHO keeps growing, over time these fees can be introduced back.

5 Likes

We support reducing the buy fee. However, we oppose reducing it to below 0.1% due to concerns that this could significantly decrease the volume of GHO Pools on DEX.

Thank you, @stani, for raising this important point regarding the GSM’s fee structure and its broader implications on liquidity and arbitrage dynamics. The 0.2% GHO burn fee provides a buffer of peg protection at a 0.2% depeg level, attracting arbitrageurs to buy back GHO from the secondary market and burn it once this buffer is reached. GHO supply benefits when it trades at a premium or when USDT or USDC trades below GHO. In the current instance, the depeg of USDT caused the USDT GSM to be fully refilled.


USDC GSM Holdings vs GHO Price. Source: Dune, January 17th, 2025.

The USDC GSM chart shows that its last two emptying events, in July and November 2024, coincided with GHO’s downward depegs. If exposure caps reach 100% for USDC (8M) and USDT (16M), the GSM will represent 14.7% of the 163.2M circulating GHO, consistent with the previous 10-15% supply ratio. However, there is a risk of depeg with just 3M GHO available within a 1% price impact.

Removing the 0.2% burn fee would allow arbitrageurs to sell freely and encourage minting or redeeming GHO on small secondary market price fluctuations of USDT, USDC, or GHO, weakening the GSM’s “peg stability buffer” property. This may reduce the time USDC or USDT stays in the stability module, leading to future revenue loss from the proposed stataToken yield. Retaining these funds without fee sharing from the GSM module is unlikely, as all revenue currently flows to the Aave Treasury.

While the primary utility of the GSM is to serve as a GHO peg stability buffer, the impact of lost revenue cannot be ignored. To strike a balance, the burn fee could be reduced by 5 bps to assess whether the increased GHO-stable volume in the GSM could offset the loss in future yield revenue while properly monitoring how the liquidity in existing LPs reacts to this change.

Notably, the Balancer GHO/USDT/USDC pool has a 0.05% fee. Reduced fees below 0.1% could divert volume from some high-fee GHO LPs, potentially reducing DEX liquidity due to lower APRs. In the future, this liquidity crunch might aggravate a GHO depeg.

Disclaimer

This review was independently prepared by LlamaRisk, a community-led non-profit decentralized organization funded in part by the Aave DAO. LlamaRisk is not directly affiliated with the protocol(s) reviewed in this assessment and did not receive any compensation from the protocol(s) or their affiliated entities for this work.

The information provided should not be construed as legal, financial, tax, or professional advice.

4 Likes

Thanks for the analysis @llamarisk, was much needed.

You hit the nail right about the core question on whether 1) GSM is kept as a reserve of last resort for liquidity, which means the more there are costs/friction in between arbitrageurs the more this purpose is fulfilled or 2) GSM is kept as a price insurance for arbitrageours to encourage holding GHO, knowing there is 1:1 redeemability when GSMs have float.

The reason latter is interesting is because it means that users would be able to utilize GHO as a onchain savings accounts while they are not utilizing funds for strategies, overnight or longer period of time (less than 45 days).

Lets take an example, if there is a USDC or USDT float in the GSMs, a trader would mint GHO via GSM (depositing USDC into the GSM) and supplying the GHO into a yield opportunity (in this case GHO staking or Prime/Emergence market in the future) over period of time. This means they are willing to convert existing stables into GHO, knowing they would have a guaranteed price to convert back to USDC without fees. In case there are fees, that would limit any short time opportunities to use GHO as a “piggy bank” since there is 0.2% fee to convert out of GHO.

Hence, the shorter the time period and more frequent usage, the less attractive it would be to use GHO in such a use-case, meaning that Sky’s model captures more of this type of opportunity where the fees in and out are 0 and there is also an attractive sUSDS rate users can dip in and out.

In terms of the yield opportunity with the floats, I think the pecking order should be to grow the scale of GHO as primary objective, given a scaled GHO brings more scaled revenue opportunities.

I would be curious if we could have some small analysis if there are bigger addresses on Sky utilizing the PSMs the same way, let say with sUSDS. Basically using Sky as onchain piggy bank for shorter time periods (less than 45 days and shorter).

3 Likes

Noting for the community that recently Ethena moved $400M into sUSDS as a backstop for the sUSDe yield. This type of a short term “piggy bank” strategy wouldn’t be feasible if there is a 20 bps for round turn from one stablecoin to GHO and back.

1 Like

The concept of using GHO for on-chain savings aligns with expanding its utility. However, it’s essential to consider the GSM’s current limitations. Minting GHO via the GSM does not ensure 1:1 redemption at any point in the future, as it depends on the sufficient float, which can deplete quickly during market stress or GHO depegs.

We can explore similarities between GSM and PSM mechanisms and consider the factors that would help ensure longer-term 1:1 swap guarantees for GSMs.

Sky PSM Flow

LitePSM USDC A is utilized by Sky protocol to maintain a 400M DAI buffer, with its USDC reserves held in the PSM Pocket. Currently, the PSM holds 3.27B USDC, with a theoretical ceiling of 10B. These reserves are collateral for minting USDS and sUSDS (equivalently DAI/sDAI). While PSM can be interacted with directly to perform 1:1 swaps, the illustration below details the underlying process of minting sUSDS by depositing USDC via PSM.


sUSDS Mint Using USDC. Source: Etherscan, January 27th, 2025.

PSM Usage

Unlike GSM, which is primarily used for price arbitrage when discrepancies arise between USDT/USDC/GHO secondary market prices, the PSM does not exhibit these flows. This difference in usage stems from the fact that the PSM is also utilized in the sUSDS minting process. As explained previously, in this process, USDC is deposited into the PSM to mint DAI, which is then converted into the yield-bearing sUSDS.


Reserves vs DAI & USDC Market Prices. Source: Dune, February 6th, 2025

The usage of the PSM can be linked to the Sky Savings Rate (SSR). When the SSR exceeds broader market yields, the USDC balance in the PSM module increases, signaling the attractiveness of holding sUSDS. This pattern was evident in September and October 2024, when the SSR ranged between 6.25% and 6.5%, while market sentiment remained neutral. In November 2024, favorable market conditions led to a sharp rise in yields offered by other platforms, reducing the appeal of holding sUSDS. This was reflected in a decrease in the USDC balance within the PSM. However, as the SSR was raised to 12.5% and market positivity dissipated, USDC holdings in the PSM began to increase sharply again. This supports the hypothesis that PSM is used for holdings.


Sky Savings Rate. Source: Sky Ecosystem Dashboard, February 6th, 2025

Nonetheless, similarly to the GSM, the swap aggregators, MEV Bots, and other automated entities mainly execute interactions with PSM. Around 10% of interactions are made by the USDS PSM Wrapper, which, as explained above, performs the sUSDS minting procedure. Therefore, this PSM Wrapper part can be attributed to the longer-term holdings in the PSM, which remain in the PSM according to the attractiveness of SSR yield opportunity.


User Distribution by PSM Interaction Volume. Source: Dune, February 6th, 2025

To analyze the shorter-term “piggy bank” operations where PSM is used as a deposit vault ensuring 1:1 swaps at a later time, we have inspected the wallets with smaller amounts of interactions (<500) but executing both USDC deposit and withdraw actions. The findings indicate that most deposits were short-term (<10 days). Nonetheless, there were still ~230 instances where PSM was used as a longer-term 1:1 swap guarantor.


PSM Deposit Duration. Source: Dune, February 6th, 2025

Considerations

Based on the above findings, the PSM can guarantee longer-term deposits due to the direct integration with sUSDS minting. Corresponding USDC amount deposited in the module serves as a stable liquidity source. Therefore, even large volumes in aggregator and MEV swaps do not cause the PSM to become imbalanced. If GSM is integrated to serve as a deposit vault for direct yield source, it will reach properties similar to PSM’s. In that case, we could also expect increased revenues for Aave’s DAO as GHO supply and usage would increase. However, at the current setup, we believe that the activity of automated actors would only become more apparent if the swap fees are set to 0, which in turn could cause instability in the GSM deposits and nullify expected revenue of the upcoming stataUSDT/stataUSDC GSM integration.