GHO vs crvUSD: Peg stability and borrow rates compared

GHO vs crvUSD: Peg Stability and Borrow Rate Mechanisms

Curve’s crvUSD uses LLAMMA soft liquidations, PegKeeper contracts, and an automated borrowing-rate policy that responds to market conditions.

That distinction determines how each decentralized stablecoin reacts when demand changes, collateral prices fall, or the market trades the token away from its reference value. GHO places more of the control loop in governance and fixed conversion infrastructure. crvUSD moves more of the control loop into contracts that adjust collateral exposure, pool liquidity, and borrowing costs as the deviation develops.

The relevant comparison is therefore not which token has the stronger label. It is how each protocol translates a peg deviation into an on-chain action.

Governance-Led Stability: How GHO Uses the GSM and Manual Rate Control

GHO is issued through Aave’s lending markets. A user supplies accepted collateral, borrows GHO, and creates a debt position that must remain within the protocol’s collateralization limits. The basic mint/burn mechanics resemble other collateralized stablecoins: new GHO enters circulation when borrowers open or increase debt, while repayment removes GHO from the borrower’s obligation and reduces the outstanding supply.

The peg mechanism adds a second layer. The GHO Stability Module, or GSM, allows 1:1 swaps between GHO and selected governance-approved stablecoins, including USDC and USDT. In a simplified scenario, the system operates as follows:

1. If GHO trades above one dollar, a trader can acquire GHO through the GSM at the fixed conversion rate and sell it in the market at the higher price.

2. The additional GHO supply increases available market liquidity and creates an arbitrage loop that pushes the market price downward.

3. If GHO trades below one dollar, a trader can buy discounted GHO in the market and exchange it through the GSM for an accepted stablecoin at the 1:1 rate.

4. The resulting reduction in GHO supply removes some of the excess selling pressure and allows the market price to move toward the reference value.

This mechanism depends on the assets accepted by the GSM, the available capacity of each module, and the confidence that the reserve asset can be redeemed or transferred without interruption. A 1:1 conversion path is useful only when it remains accessible at the moment market participants need it.

The GSM is not an algorithm that independently decides how much capacity to provide or which stablecoins to accept. Aave Governance controls those parameters. Governance can add or remove assets, change exposure limits, and adjust the structure of the module as the risk profile of the underlying reserves changes.

Borrow rates are a policy variable

GHO’s borrowing cost is also governance-led. Unlike a typical utilization-based lending asset, GHO does not rely on a continuously automated interest-rate curve that moves mechanically with the utilization ratio. Aave Governance sets the borrow rate and can revise it through the protocol’s governance process.

The rate influences the peg through borrower demand:

• A lower borrow rate reduces the cost of creating GHO and can expand supply when users find the debt attractive.
• A higher borrow rate increases the cost of maintaining GHO debt and can discourage additional minting.
• If GHO trades persistently below one dollar, a higher debt cost may reduce new borrowing and encourage repayment.
• If GHO trades above one dollar, a lower borrowing cost can make new issuance more attractive, increasing supply.

This is a slower control path than an automated market response. The protocol does not necessarily reprice debt each time the market moves by a small amount. Governance must identify the condition, assess its persistence, and execute a parameter change.

Aave has also used incentives tied to AAVE staking. The initial discount for users staking AAVE in the Safety Module was 30%, reducing the effective cost of GHO borrowing for eligible accounts. The discount creates an additional relationship between GHO demand and AAVE staking, but it also complicates the system. The effective borrowing cost is no longer determined only by the base GHO rate. It depends on user eligibility, the discount policy, and the continued operation of the Safety Module.

GHO’s stability loop is partly a market mechanism and partly a governance process: the GSM can create a conversion path, but the protocol still decides where that path starts and how much capacity it has.

The key limitation: control latency

Manual rate control is not inherently weak. It can be deliberate, transparent, and aligned with broader risk management. The limitation is response time.

Suppose GHO begins trading below its target because secondary-market liquidity contracts while borrowers continue to repay slowly. The GSM can support arbitrage, but only within its configured capacity. If the deviation persists, the borrowing rate may need to change. That change requires governance coordination rather than an instantaneous contract-level response.

The system therefore has a control latency:

• market price moves;
• arbitrageurs test the available GSM capacity;
• governance observes the deviation and its cause;
• a proposal changes the borrow rate or module parameters;
• borrowers and liquidity providers respond.

If the deviation is temporary, governance may not need to act. If the deviation reflects a structural imbalance, the delay becomes part of the peg risk. The protocol must distinguish noise from a change in equilibrium before it changes the policy variable.

Algorithmic Resilience: How crvUSD Uses LLAMMA

crvUSD uses a more automated architecture. Its central mechanism is LLAMMA, the Lending Liquidating AMM Algorithm. The system does not wait for a collateral position to reach a single liquidation threshold and then sell the entire position in one event. Instead, LLAMMA is designed to move a position progressively between collateral and crvUSD exposure as the collateral price changes.

This is commonly described as soft liquidation. The term refers to the gradual transformation of the position rather than a single binary liquidation transaction.

Consider a user who borrows crvUSD against a volatile asset. While the collateral price remains within the relevant range, the position retains a greater share of the original asset. As the price declines, LLAMMA sells portions of that collateral into crvUSD through its automated market structure. If the market later recovers, the process can work in the opposite direction, converting part of the position back into the collateral asset.

The mechanism changes the liquidation threshold from a single cliff into a range of prices. That can reduce the probability of a sudden full liquidation, especially during orderly market declines. It does not eliminate losses. A borrower can still lose collateral through the liquidation process, and the AMM’s execution depends on market liquidity and price behavior.

A simplified sequence looks like this:

1. The user deposits collateral and borrows crvUSD.

2. The collateral price declines toward the LLAMMA band.

3. The algorithm gradually exchanges collateral exposure for crvUSD exposure.

4. The user’s debt remains denominated in crvUSD while the collateral composition changes.

5. If the collateral price recovers, LLAMMA can reverse part of the conversion.

6. If the decline continues beyond the relevant range, the position may hold substantially more crvUSD and less of the original collateral.

This structure directly connects collateral management with peg stability. When the system converts collateral into crvUSD, it creates demand for the stablecoin. At the same time, it changes the risk carried by the borrower’s position. The mechanism is therefore both a liquidation engine and a source of market activity.

PegKeepers add an elastic supply layer

crvUSD also uses PegKeeper contracts. These contracts can mint or burn crvUSD in selected Curve stableswap pools depending on the token’s market price.

When crvUSD trades above one dollar, a PegKeeper can mint crvUSD and place it into the relevant pool. Additional supply tends to reduce the price premium. When crvUSD trades below one dollar, the system can reduce the outstanding PegKeeper debt by burning crvUSD as the pool balance moves in the opposite direction.

The mechanism is not equivalent to the GHO Stability Module. The GSM offers a direct 1:1 swap between GHO and approved stablecoins. PegKeepers operate through designated Curve pools and adjust the pool’s composition. One is a conversion facility; the other is an algorithmic liquidity-balancing instrument.

The distinction matters during stress. A direct swap facility depends on its configured reserve capacity and counterparty assets. A PegKeeper depends on pool depth, the behavior of arbitrageurs, and the quality of the stablecoin pair used in the pool. Both mechanisms can support a peg, but they expose the system to different failure modes.

Dynamic borrow rates and the January 2026 smoothing update

crvUSD borrowing rates are dynamic. They respond to the peg deviation and to PegKeeper debt rather than remaining fixed by a governance-set schedule.

The intended control logic is straightforward:

• if crvUSD trades below its target, the borrowing rate can increase;
• a higher rate discourages new borrowing and can encourage debt reduction;
• if crvUSD trades above its target, the rate can decrease;
• a lower rate may support additional borrowing and new crvUSD supply.

The system also considers PegKeeper debt. If PegKeepers have issued a significant amount of crvUSD to stabilize a pool, that debt indicates that the market requires additional intervention. The borrowing-rate policy can incorporate this condition rather than relying only on a single spot-price observation.

The difficulty is that a control system reacting too quickly can become unstable. Small price deviations, temporary pool imbalance, or thin liquidity can produce abrupt rate changes. Borrowers then receive a rapidly changing cost signal, which can cause them to open, close, or refinance positions at precisely the wrong time.

Curve DAO approved an EMA smoothing update in January 2026. The exponential moving average reduces the influence of isolated observations and makes the borrowing-rate response less sensitive to short-term volatility. In engineering terms, the EMA acts as a filter. It does not remove the feedback loop; it changes the speed and smoothness of the response.

That creates a trade-off:

GHO and crvUSD Under the Same Market Shock

A comparison becomes clearer when both systems are exposed to identical conditions. Consider a broad decline in demand for decentralized stablecoins. Users reduce leverage, liquidity providers withdraw capital, and secondary-market spreads widen.

Scenario one: the token trades below one dollar

For GHO, the first response comes from arbitrage around the GSM. If the module offers sufficient capacity, traders can buy discounted GHO and convert it into USDC, USDT, or another accepted stablecoin. This creates a direct floor mechanism, although the strength of that floor depends on the remaining module capacity and the market’s ability to execute the transaction.

Aave Governance may then raise the GHO borrow rate. The higher rate makes new GHO debt less attractive and may encourage borrowers to repay. The supply response is real, but it is not necessarily immediate. Existing positions do not disappear when the rate changes, and borrowers with sufficient collateral may continue to hold the debt.

For crvUSD, the automated system can respond through several channels. PegKeepers can adjust the pool supply, while the borrow rate can rise as the deviation and PegKeeper debt increase. A higher rate changes incentives without waiting for a governance vote. LLAMMA may also alter borrower collateral exposure if falling asset prices push positions into their soft-liquidation bands.

This response is faster but more coupled. The peg, the borrowing rate, the collateral composition, and the pool’s state can all affect one another. A rate increase may reduce borrowing, but it may also pressure existing borrowers to unwind positions. If those positions sell collateral into a declining market, the resulting price impact can increase stress elsewhere in the system.

Scenario two: the token trades above one dollar

GHO’s GSM provides the clearest supply-expansion route. If users can mint or obtain GHO through the relevant Aave markets at a cost below the market price, they can sell the token at the premium. The additional supply should compress the deviation, subject to collateral availability, borrowing limits, and the borrower’s cost.

Governance may reduce the borrow rate if the premium reflects insufficient GHO supply rather than a temporary liquidity distortion. That decision requires a diagnosis. A higher market price does not automatically mean that the base borrow rate should change; it may result from a temporary imbalance in a single venue.

crvUSD can use PegKeepers to add supply to designated pools. The dynamic rate may also fall, encouraging new borrowing. Here again, the response is algorithmic. The system can act quickly, but the resulting supply is not free. New crvUSD is associated with new collateral positions, and the quality of that collateral determines the system’s ability to absorb a later decline.

Scenario three: collateral prices fall sharply

This is where the architectures diverge most clearly.

GHO’s peg facility does not itself provide a soft-liquidation path. The system relies on Aave’s collateral risk parameters, liquidation thresholds, liquidation incentives, and market liquidators. Once an account crosses the relevant threshold, liquidators can repay part of the debt and claim collateral under the protocol’s rules. The position is managed through discrete liquidation events.

crvUSD uses LLAMMA to distribute that transition over a range. A borrower may lose exposure to the collateral asset progressively rather than at a single liquidation point. This can reduce the need for a large, immediate sale, but it introduces impermanent-loss-like behavior and path dependency. The outcome depends not only on the final collateral price, but also on the route taken to reach it.

A rapid V-shaped decline and recovery can produce a different result from a slow decline followed by a recovery. In both cases, LLAMMA may move collateral into crvUSD and later attempt to move it back. Execution prices, pool liquidity, and the timing of the price path influence the borrower’s final position.

Soft liquidation changes the shape of the loss distribution. It replaces a single liquidation cliff with a wider execution range, but it does not make collateral volatility disappear.

Cross-Protocol Integration: The GHO/crvUSD PegKeeper Partnership

In February 2026, Curve DAO approved adding the GHO/crvUSD pool to the crvUSD PegKeeper set with an initial debt ceiling of 3,000,000 crvUSD. The decision gives crvUSD another PegKeeper counterparty and connects the two stablecoins through a shared Curve liquidity venue.

The integration has two consequences.

First, it diversifies the set of pools through which crvUSD can manage its peg. A PegKeeper system that relies on several pools is less dependent on the condition of any single market. If one pool loses depth or experiences a temporary imbalance, other venues may continue to absorb some of the adjustment.

Second, it creates a direct feedback channel between GHO and crvUSD. A pool containing both tokens is not simply a passive exchange venue. Its balances can influence PegKeeper debt, arbitrage routes, and the effective liquidity available to both assets.

The initial debt ceiling limits the amount of crvUSD that the PegKeeper can create through this relationship. That ceiling is a risk-control parameter. A larger limit would give the mechanism more capacity to respond to a deviation, but it would also allow a larger synthetic exposure to accumulate if the pool remained imbalanced for an extended period.

The partnership does not merge the two peg systems. GHO remains dependent on the GSM and Aave Governance’s rate decisions. crvUSD continues to use LLAMMA, PegKeepers, and its dynamic borrow-rate policy. The shared pool is an integration point, not a common monetary controller.

It also introduces a new risk surface. If GHO experiences a problem with GSM capacity or reserve confidence, the GHO/crvUSD pool can transmit part of that imbalance into Curve’s PegKeeper system. Conversely, if crvUSD’s algorithmic response causes a large pool imbalance, GHO liquidity providers may absorb some of the resulting inventory risk.

The correct interpretation is therefore limited: the pool can improve liquidity diversification for crvUSD and create a new arbitrage route, but it does not guarantee that either stablecoin will maintain a one-dollar price under stress.

Yield Evolution: sGHO and scrvUSD Are Separate Products

Both protocols have introduced native savings wrappers, but their yield models should not be treated as interchangeable.

Aave launched Savings GHO, or sGHO, on May 16, 2026. The product offered an annual percentage rate of approximately 4.25% as of mid-2026. Users deposit GHO into the wrapper and receive exposure to the product’s yield distribution without converting the asset into a different volatile collateral position.

Curve offers Savings crvUSD, or scrvUSD. Its return is linked to protocol revenue and the share allocated to depositors through governance-approved parameters. The exact long-term target APY is not fixed because it depends on realized revenue and the approved revenue-share percentage. A February 2026 proposal considered increasing the maximum share of crvUSD revenue distributed to scrvUSD from 50% to 80%.

The difference is important for comparing risk-adjusted returns:

sGHO is not the same as stkGHO. Savings GHO is a non-slashable yield product under the stated product design. That distinction matters because users should not infer staking-related slashing exposure from a savings wrapper.

For scrvUSD, the primary uncertainty is not a fixed coupon. Its return depends on how much revenue the crvUSD system generates and what portion governance directs to depositors. A high revenue-share ceiling does not guarantee that the ceiling will be reached, nor does it establish a permanent APY.

The wrappers also create a secondary peg consideration. A stablecoin savings product can reduce liquid circulating supply if users deposit the underlying token and hold the wrapper. That may support market scarcity, but it can also reduce immediately available liquidity during a redemption wave. Yield design therefore interacts with peg mechanics rather than operating independently from them.

Market Scaling: What GHO’s Supply Growth Does and Does Not Prove

GHO’s circulating supply exceeded 580 million tokens by March 2026. The token had deployments across Ethereum, Arbitrum, Base, Avalanche, Gnosis, and Lens.

Multi-chain deployment expands the number of venues where GHO can be borrowed, traded, and used as collateral or settlement liquidity. It can also distribute demand across different lending markets and decentralized exchanges. That is useful for adoption, but it increases the number of components that must remain synchronized.

A cross-chain stablecoin system introduces operational questions:

• Are minting and bridging limits consistent across networks?
• Can a liquidity shortage on one chain create a discount that does not appear elsewhere?
• Does the GSM provide equivalent conversion access across all deployments?
• Are oracle and liquidation parameters calibrated for each chain’s liquidity?
• Can a bridge or messaging failure isolate supply between networks?

Supply growth by itself does not answer these questions. A larger circulating balance can improve network effects, but it can also increase the amount of debt and liquidity that the stability system must manage during a contraction.

The comparison with crvUSD is not simply a comparison between 580 million GHO and the current crvUSD supply. GHO’s design emphasizes Aave’s lending infrastructure, governance-set rates, and the GSM. crvUSD’s design emphasizes Curve liquidity, automated rate adjustment, LLAMMA positions, and PegKeeper operations. Scale changes the magnitude of the risks, but not their basic direction.

Theoretical Limits and Stress-Test Vulnerabilities

The two systems solve different parts of the stablecoin problem. Neither removes the need for collateral, liquidity, governance, or reliable market data.

GHO’s limiting conditions

GHO’s stability is constrained when one or more of the following conditions appears:

• GSM capacity is too small relative to the outstanding market imbalance.
• The accepted reserve stablecoins lose liquidity or confidence.
• Governance responds too slowly to a persistent supply-demand mismatch.
• The borrow-rate discount creates demand that is not supported by durable use cases.
• Cross-chain liquidity fragments the arbitrage loop.
• Aave collateral markets experience a rapid liquidation cascade.

The GSM can establish a strong arbitrage route, but it cannot force traders to use that route if transaction costs, bridge risk, or reserve concerns are too high. Manual governance can set a coherent policy, but it cannot compress decision time below the protocol’s social and operational coordination limits.

crvUSD’s limiting conditions

crvUSD’s automated design has a different set of boundaries:

• LLAMMA depends on sufficient AMM liquidity and reliable price movement.
• Soft liquidation can still realize losses during a sustained or discontinuous decline.
• PegKeepers can accumulate debt when pools remain imbalanced.
• Dynamic borrow rates can overshoot if the feedback signal is noisy.
• EMA smoothing improves stability of the control input but also delays the response.
• A severe collateral sell-off can link liquidation activity to broader market stress.

Automation reduces governance latency but does not guarantee correct behavior under every price path. An algorithm can respond precisely to the signal it receives and still produce an undesirable result if the signal is delayed, manipulated, or generated by an illiquid market.

The January 2026 EMA update addresses one specific vulnerability: excessive rate volatility. It does not eliminate the underlying feedback problem. Smoothing makes the rate path more predictable, but the system must still choose between speed and stability. A very fast response can destabilize borrowers; a very slow response can allow peg debt or market imbalance to grow.

Which Model Is More Robust?

The answer depends on the stress being measured.

GHO has a clearer governance boundary. The GSM provides an explicit 1:1 conversion mechanism, while Aave Governance can adjust rates, discounts, accepted assets, and capacity. This structure is easier to map conceptually, but its effectiveness depends on governance speed and the available room inside the stability module.

crvUSD has a more integrated automatic response. LLAMMA manages collateral through a price range, PegKeepers adjust pool-side supply, and the borrow rate changes in response to deviations and accumulated intervention debt. This reduces reliance on frequent governance action, but it increases the number of interacting feedback loops.

This is not a ranking of stablecoin quality. It is a classification of control architectures. GHO concentrates discretion in governance and conversion modules. crvUSD distributes control across liquidation algorithms, pool contracts, and automated monetary parameters.

The most important metric for both is not the headline APY or circulating supply. It is the amount of usable liquidity available when the peg moves, the speed at which the system can alter incentives, and the quality of collateral supporting the outstanding debt.

The GHO/crvUSD integration may improve those conditions by expanding liquidity routes and diversifying PegKeeper counterparties. It can also transmit stress between the systems. A shared pool is an additional arbitrage path, not a substitute for independent risk controls.

Protocol analysis also requires separating market attention from mechanism quality. The distinction is less suited to news about singers, rappers, and bands than to balance-sheet analysis, because a stablecoin’s resilience is determined by contract capacity, collateral paths, and liquidation thresholds rather than headline visibility.

Conclusion

GHO and crvUSD maintain their dollar references through different engineering choices.

GHO uses the GSM to support 1:1 swaps and relies on Aave Governance to set borrowing costs and adjust key parameters. Its strength is a direct and legible conversion route. Its main constraint is governance latency and the finite capacity of the module.

crvUSD uses LLAMMA to soften collateral liquidation, PegKeepers to adjust liquidity-pool supply, and dynamic borrow rates that respond to peg deviation and intervention debt. Its strength is an automated control loop. Its main constraint is feedback complexity: the system must manage collateral volatility, pool imbalance, rate sensitivity, and the timing introduced by EMA smoothing.

The practical comparison is therefore conditional. GHO is more policy-driven. crvUSD is more algorithmically reactive. Neither model is self-sufficient. Under stress, each depends on arbitrage capital, functioning markets, adequate collateral, and parameters that match the speed of the shock.

A stablecoin peg is not a single switch. It is the output of a control system. GHO places that system closer to governance and reserve conversion. crvUSD places it closer to automated liquidity management and progressive liquidation. The theoretical limit for both arrives when the available liquidity, collateral value, or response capacity becomes smaller than the imbalance the protocol is attempting to absorb.