Frax Finance: The Fractional-Algorithmic Bank

Author: Alec Korman

August 25, 2021

Flawed mechanisms, extreme reflexivity, outright scams – decentralized stablecoins have faced serious challenges. Yet, the reasons for building decentralized stable assets remain clear. Central Banks have explicitly stated their intentions to build CBDCs, enabling them to merge monetary and fiscal policy, and program their issued currencies with undesirable qualities that infringe on privacy and freedom. Since custodial money market funds like Circle and Tether have the capacity to freeze their tokenized dollars, relying too heavily on these centralized stablecoins as an alternative to CBDCs increases systemic risk in decentralized financial markets. Most regard decentralized stablecoins as adding risk to individual portfolios and cryptoeconomic systems; but ideally they will give people more financial security and decrease overall systemic risk by eliminating single points of failure.

Stablecoin Landscape

Despite instances of tribalism, the market needs multiple decentralized stablecoins with varying designs and purposes to properly risk-off. Ideally, numerous quality decentralized stable assets will capture network effects. The strongest projects deserve both support and constructive criticism. With that in mind, let’s briefly summarize the leading protocols: MakerDAO and Terra. 

Maker allows users to mint DAI by locking exogenous collateral in vaults as collateralized debt positions (CDP). Maker is the oldest and most reliable on-chain bank, and has begun expanding the scope of its collateral to include tokenized real-world assets. The Maker token, MKR, controls governance and serves as the lender of last resort in case of shortfall events. DAI currently has a market cap of $5.68 billion. 

Terra is a PoS blockchain built on Tendermint and has cultivated a thriving ecosystem of financial protocols on top of its native stable assets. Terra stablecoins are minted and redeemed through a process of arbitrage swapping with its seigniorage share, LUNA. This process allows the supply of UST to expand and contract according to demand by providing an economic incentive to the market for keeping the price near its intended peg. During expansionary periods for Terra stablecoins, LUNA becomes deflationary and its price typically increases as arbitrageurs buy and burn it — not to mention the traders who simply buy the asset as a speculative bet. Additionally, LUNA has intrinsic value as an asset which secures the Terra blockchain in its PoS protocol and generates cash flow from staking rewards and Terra swap fees. UST currently has a market cap of $2.2 billion. 

Collateralization Extremes

DAI is primarily minted to generate leverage, and Terra stablecoins are primarily minted to participate in the vibrant TeFi ecosystem. Both systems serve valuable use cases, but they also have vulnerabilities. Maker CDPs must be over-collateralized, so the total value of assets locked in the protocol should always exceed the total DAI market cap. This makes DAI capital inefficient since it can only scale with the demand for leverage. The protocol has also developed an increasing reliance on USDC as collateral. The price of DAI is currently stable, but endured an extended period of time last year in which its price traded well above $1. This price deviation occurred because of a deflationary deleveraging spiral after its liquidation system broke down due to the March 2020 crash [1]. 

Since Terra stablecoins rely entirely on endogenous collateral, its system is particularly vulnerable to the risk of bank runs. After the May 2021 crash, the market cap of LUNA rapidly dropped below the market cap of UST. People began trading out of their UST positions, which drove the UST price below its intended peg and reinforced the LUNA price crash. Eventually supporters of the network were able to restore the peg and prevent a serious bank run, but Terra must prepare for this kind of event in the future. This is especially important as the ecosystem continues to rapidly scale, with an upcoming network upgrade and dozens of new protocols creating serious demand for UST. 

Between the collateralization ratios (CR) of CR ≥ 150% and CR = 0% lies a middleground which preserves the strengths of both extremes while reducing their vulnerabilities. This is the key idea behind Frax, which has distinguished itself from other decentralized stablecoin protocols with its dynamically adjusting ratio of endogenous and exogenous collateral, i.e., 0% < CR < 100%. While the protocol hasn’t existed long enough to endure as many anomalous market events as something like Maker, the FRAX price has remained reliably stable, experiencing no serious deviations beyond 1% of the peg:

The protocol has clear aspirations of scalability, robustness, and adaptability. Let’s investigate to what extent Frax possesses these qualities, why it has remained stable, and how it compares with other protocols like Maker and Terra. 

Frax Core Stability Mechanism

         Frax is a fractional-algorithmic system in which exogenous USDC collateral and endogenous Frax Shares (FXS) both partially back the stablecoin. FXS serves as a volatility absorbing seigniorage asset, and gives Frax greater capacity for scaling since the system requires less exogenous collateral to grow. A PIDController dynamically adjusts the CR based on changes in a growth ratio which measures FXS liquidity against the total supply of FRAX. 

The CR determines what ratio of exogenous and endogenous collateral is necessary to mint or redeem FRAX. For instance, if the CR is 85%, then minting 1 FRAX requires depositing $0.85 USDC as well as burning $0.15 worth of FXS. As the growth ratio increases, meaning FXS liquidity has increased relative to the supply of FRAX, more FRAX can be redeemed with less of a percentage impact on FXS supply. So, the system can absorb more FXS sell pressure from FRAX redemptions being sold without risking negative feedback loops, and the CR decreases. 

Core Stability Mechanism:

  1. Decollateralize: lower the CR by some increment x every time t if FRAX>$1
  2. Market Operations: don’t change the CR if FRAX=$1
  3. Recollateralize: increase CR by some increment x every time t if FRAX<$1
  4. FXS1559: burn FXS with minted unbacked FRAX, extra collateral, and fees

         The Core Stability Mechanism allows Frax to adapt based on information from the market. Changes in FRAX supply and FXS liquidity, as captured by the growth ratio, serve as signals which indicate how much endogenous collateral ought to back the system. During sustained periods of net positive growth ratio change, the market essentially expresses confidence in the endogenous collateral and signals to the protocol that the CR should be lower in order to better facilitate scaling. This gives Frax a serious advantage over a system like Maker in terms of growth. The total value of assets locked in Frax will always be less than or equal to the FRAX market cap, so Frax can more easily adapt to the pace of increasing stablecoin demand. 

During sustained periods of net negative growth ratio change, the market signals that more exogenous collateral should back the system and that the protocol should become more robust against destabilization risks. Stablecoins backed by even some percentage of endogenous collateral are at risk of bank run events, but Frax aims to minimize the risk of complete ruin. The exogenous collateral within the system dampens the reflexive downward spirals that are more likely to occur in systems entirely reliant upon endogenous collateral. The worst-case scenario for Frax would be if the CR rapidly spiked from a low to high percentage right before an event in which the FXS market cap plummets, triggering a bank run. Then, a certain percentage of FRAX holders could drain all of the collateral from the system through redemptions, leaving the remaining holders with improperly collateralized FRAX. While this scenario is theoretically possible, it’s unlikely. FRAX utility and FXS fundamentals aside, the Frax CR isn’t designed to rapidly fluctuate, so there won’t be extended opportunities in which the CR vastly exceeds the actual percentage of collateral in the system. Also, this divergence typically happens during periods of sustained FRAX demand growth, before returning to an equilibrium as the demand neutralizes. Nevertheless, this risk is theoretically possible and worth considering. 

The critical components of the Core Stability Mechanism can certainly become more sophisticated over time — for instance, the PIDController doesn’t contain integral and derivative terms — but the system’s responsiveness to market information is likely why Frax has so successfully maintained its stability [2]. A CR = 0% assumes the market has complete confidence in the endogenous collateral and no desire for the ability to redeem exogenous collateral. A CR ≥ 100% assumes the market has zero confidence in endogenous collateral and hinders growth. No single developer, economist, or stablecoin holder can know the ideal CR, but collectively they can help the market refine a continuously evolving estimate. Frax’s dynamically adjusting CR aims to capture that emergent market intelligence.

Frax AMOs

         Frax v2 introduced the concept of Algorithmic Market Operations (AMO). Since the protocol controls a nonzero amount of exogenous collateral from FRAX minting, it can deploy contracts which use those assets to implement various strategies. Generalizing from the Core Stability Mechanism, each AMO can be broken down into the following four essential properties:

  1. Decollateralize: action which lowers the CR
  2. Market Operations: action which runs in equilibrium and doesn’t change the CR
  3. Recollateralize: action which increases CR
  4. FXS1559: formalized accounting of the balance sheet of the AMO which defines exactly how much FXS can be bought and burned with profits above that CR

This framework enables the Frax community to select from an arbitrarily large set of modular banking strategies to generate revenue, upgrade the system, and enable partnerships – all without undermining the price stability of FRAX. The active AMOs, which the developers describe as financial money LEGOs, essentially amount to Frax monetary policy. The protocol has deployed a set of successful AMO strategies:

The Investor AMO deploys Frax collateral to yield aggregators and money markets. Frax currently deploys a portion of its USDC into Yearn, Aave, Compound, and OlympusDAO. The AMO does not allocate funds into strategies that have a time delay for withdrawals, so it never lowers the CR and can pull out the collateral if needed for FRAX redemptions.

The Curve AMO deploys idle and newly minted FRAX into the FRAX3CRV pool. Frax utilizes this AMO not only to earn revenue, but to increase FRAX liquidity and strengthen its peg. Frax also sets and collects admin fees since the AMO contract deployed the FRAX3CRV metapool and owns admin privileges. The AMO also utilizes Convex and StakeDAO to maximize LP rewards. 

The Lending AMO uses Rari Capital and CREAM. Since all borrowed FRAX is over-collateralized, this strategy never lowers the CR. Directly lending FRAX into money markets gives this AMO the ability to increase/decrease FRAX interest rates through minting and burning. So, for instance, Frax can change the cost of shorting FRAX, or incentivize FRAX adoption by offering highly competitive borrowing rates. Note: this AMO essentially replicates the functionality of MakerDAO by allowing people to mint stablecoins against an over-collateralized deposit in order to generate leverage.

         AMOs give the protocol a fractal structure since each individual AMO has the same essential features and price stability invariant as the system as a whole. The massive design space for AMOs, and their modular nature, makes Frax highly adaptive. The protocol can regularly upgrade or create AMOs based on its current state and general market conditions. Another nice feature is that the protocol should always have a nonzero amount of collateral assets at its disposal for AMOs. This is important since these strategies can increase stability and incentivize adoption. 

Frax, like Maker, has received criticism for its reliance upon USDC. At this time, USDC is the only collateral asset in Frax, but there’s no reason that must be the case. Not only does Frax intend on utilizing a basket of crypto assets like other decentralized stablecoins, as well as ETH and WBTC, for its exogenous collateral, but the developers are designing a metric referred to as the decentralization ratio which will measure the extent to which a decentralized stablecoin system is dependent upon centralized components. In order to adopt a basket of crypto assets as collateral without risking the stability of the system, the protocol will deploy AMOs which manage the volatility of the non-stable collateral assets. For instance, an AMO can run a neutral strategy by depositing ETH onto Synthetix and converting half of the supply into iETH (inverse ETH). AMOs enable Frax to select from a larger set of possible collateral assets and grow more decentralized. 

FXS1559: AMO Revenue

The FXS1559 property of AMOs turns FXS into a revenue generating asset. Importantly, FXS holders have claim over all AMO revenue. Some of the profit is held indefinitely, like farmed tokens which give Frax increased influence over the governance of partner protocols, but most goes towards buying and burning FXS on the open market. To get a relative sense of FXS cash flow, we ought to compare Frax with the leading stablecoin issuer: Circle. Since both entities are essentially money market funds – Circle a custodial MM fund with ERC20 receipts and Frax a non-custodial automated on-chain MM fund – we can gain good insight by comparing them on a relative basis. This past month, Circle announced its intention to go public via SPAC at a $4.5b valuation. Circle projects $115m in revenue this year, meaning Circle equity is set to trade at 39.13 P/S.

Circle’s balance sheet includes revenue sources beyond its USDC business, like Transaction and Treasury Services (TTS). If we were to treat Circle purely as an entity which only generates revenue from its USDC business, then Circle equity would trade at 112.5 P/S.

FXS currently has a $592,000,000 market cap, and Frax is on pace to generate over $38m in revenue in the coming year. So FXS trades at 15.45 P/S. 

Despite USDC clearly existing at a much larger scale than FRAX, Frax is on pace to generate almost the same amount of revenue off of FRAX as Circle makes off of USDC. Of course, Frax’s impressive revenue is largely due to the massive yields in DeFi, and Circle has a more predictable revenue stream with a superior risk profile. Still, on a relative basis, Frax is significantly more profitable than Circle. In fact, Circle doesn’t project profits until 2023, so their 2021 and 2022 P/E is negative [3]. For Frax, on the other hand, P/S and P/E are the same quantities since the balance sheet is owned by token holders with all revenue counting as profit to shareholders. While the FRAX market cap may seem far off from that of USDC at this point, the conversation may dramatically change in the near future as the threat of regulations on centralized stablecoins looms large.

Multi-Network Liquidity & Crypto Native CPI

AMOs have enabled Frax to entrench itself as a vital contributor to many other DeFi protocols. For example, the protocol is one of the largest CVX farmers, and FRAX is one of twelve exclusive tokens used as a routing pair for Uniswaps v2 and v3. Frax has continued this approach by providing FRAX liquidity on multiple networks beyond Ethereum, like Avalanche, Solana, and BSC. By eventually bridging the liquidity between these separate networks, Frax can increase the overall utility and frictionless movement of the stablecoin. This focus on pervasiveness across protocols and networks is largely responsible for the $25m to $306m growth in FRAX market cap in the 8 months since its launch.

In the spirit of integrations and partnerships, Frax has also spearheaded the development of a public utility that both decentralized and centralized stablecoin systems may utilize. For context, recall that the vast majority of the stablecoin market cap is made up of assets pegged to the value of USD. This comes as no surprise since USD is the global reserve currency and the most common medium of exchange. Yet, with unprecedented quantitative easing and early signs of inflation, relying on USD to retain its purchasing power over time may grow increasingly precarious. 

To provide an alternative to the binary choice of temporary stability vs. extreme volatility, Frax is creating the first crypto native CPI. This new unit of account will not only bring the traditional CPI on-chain using Chainlink oracles, but also incorporate essential cryptoeconomic elements within the price index, like the Ethereum base fee and Bitcoin hashrate energy expenditure. Critically, several stablecoin protocols have contributed to the development of this project. The Frax team realized that in order for this unit of account to gain network effects, multiple stablecoins must peg to its value – so the Frax Price Index (the Frax stablecoin which pegs to this new CPI) will be just one instantiation of this abstract value. The more stablecoin protocols that peg to the CPI, the more frictionlessly those assets can be transmitted. For instance, if several protocols embrace this CPI, Curve pools can form with tight liquidity around stablecoins with this new value. Just as de-risking crypto markets requires many quality decentralized stablecoins, creating a new value which retains its purchasing power over time requires many stablecoin protocols embracing this crypto native CPI. 

Understanding Risk

The extent to which Frax reduces the risks inherent to other stablecoins systems, particularly those which utilize endogenous collateral, remains an open research area. Forecasting the dynamics of any crypto asset is difficult, but FRAX and FXS are particularly challenging to model. The Core Stability Mechanism, along with the AMOs, are simple and elegant, but simulating how all of the components evolve through time is highly nontrivial since some of these components impact each other nonlinearly. This task is essentially a physics problem since Frax is best represented as a dynamical system governed by a set of nonlinear difference equations. Properly simulating the system would allow for testing out new AMOs, potential PIDController upgrades, stressful market events, and so on. The beauty of Frax is that the system is highly adaptive, upgradable, and poised to grow more robust as our understanding of the risks gains clarity. 

Thanks to Andrew Kang, Ben Simon, and Sam Kazemian for their helpful feedback and comments.

Nothing in this article constitutes investment advice.