Exploring the Opportunity for DeFi Interest Rate Markets

Interest rates are a critical input in the construction and valuation of any financial asset. Every equation in finance has interest rates in it, either explicitly or implicitly. Mortgages, student loans, business loans, government borrowings, and even equities and commodities are impacted by interest rates.

Businesses can avoid interest rate risk by engaging in fixed-rate borrowing. Fixed-rate borrowing is the most common form of borrowing in traditional markets. Per Notional's terrific whitepaper, “In 2018, there was $15.3 trillion dollars of debt outstanding in U.S. corporate debt and mortgage debt markets. 88% of that debt was in terms of fixed rates.” However, fixed-rate borrowers may be unhappy with the terms of their loan (and/or have the view that rates across the economy will go down), and may swap their exposure with a different borrower who is currently paying variable but wants a fixed rate. Lenders who own fixed-rate bonds may have the view that rates across the economy will go up and desire to hedge that risk.

Today’s DeFi markets are largely limited to variable-rate lending and borrowing. The two largest borrow/lend protocols, Compound and Aave, both support variable rates. While these products work well for users aiming to earn a decent yield, they don’t work for corporations or neo-banks who need predictability—both on the lending and borrowing side. As DeFi grows to serve more use cases, the ability to lock in an interest rate will be a key enabling technology.

Interest rate swaps (IR swaps) are the most popular form of interest-rate derivatives. IR swaps account for $6.5 trillion of daily traded volume, which represents over 80% of the world's total derivatives volume. IR swaps serve a variety of use cases: 1) borrowers use IR swaps to lock in interest rates, 2) speculators can express capital efficient directional views using IR swaps, and 3) portfolio managers and lenders use swaps to manage interest rate risk.

There are a number of teams building fixed-rate lending products and IR derivative protocols in DeFi. In this essay we explore the design space for these primitives and the various choices made by teams building this future.

The Opportunity for IR Markets in Crypto

The vast majority of loans in DeFi today are variable rate. As a point of reference, there are $19 billion of loans outstanding across the three major lending protocols: 1) Aave - $7.9 billion, 2) Compound - $4.9 billion, and 3) MakerDAO - $6.3 billion.

Source: Dune Analytics, @hagaetc

Of these, 100% of Compound and MakerDAO’s loans are completely variable and open term. Additionally, of Aave’s 5 largest markets (DAI, GUSD, SUSD, TUSD, and USDC), only ~1.5% of the loans outstanding were “stable rate loans” as of July 2021.

There are a few DeFi-native fixed-rate platforms live today, such as Yield Protocol and Notional Finance. Notional has $12 million of TVL, and Yield is currently in the process of building out V2. As such, it’s clear that the market for variable-rate loans is much more mature than fixed-rate loans in DeFi. However, we expect fixed-rate loans to become larger than variable rate loans as DeFi rails eventually power traditional financial markets.

We’ve chatted with many institutional funds and neo-banks who are exploring DeFi. One of the main concerns we hear from them is that interest rates in the lending and borrowing markets are quite volatile. There are many reasons for the volatility (basis trade compression and expansion, Grayscale arb closing, new yield farms with really high APYs, etc.), but the most important fact is that even if they get comfortable with custody and regulatory risk, the lending markets don’t offer them maximal utility. The below chart is a sample of some of the most liquid USDC borrowing rates in DeFi over the past few years:

Source: DeFi Pulse

In order for large traditional businesses to borrow in DeFi, rates are going to need to be more predictable and stable. A business who borrows $100M on-chain paying 2% is very unlikely to be ok with the rate spiking to 20% a week later because an unrelated liquidity mining farm is now paying extremely high yield. The business will either want to 1) enter into a fixed-rate, fixed-term loan, or 2) have hedging access for their variable rate exposure from Compound and Aave.

As crypto grows, we also expect the market for on-chain DAO-to-DAO business lending to grow significantly. As businesses mature, debt financing becomes the major source of funding and DAOs will be no different.

We predict that many DAOs, such as the DAOs that manage risk in large DeFi protocols and NFT platforms, will pursue debt financing. Imagine a world in which the SushiSwap DAO can borrow funds from a variable-rate lending protocol and encode logic that says “we will redirect 10% of our revenues from our staked xSUSHI to repay the loan.” This would expose the SushiSwap DAO to interest rate risk from the variable rate protocol, and they could use IR swaps to hedge it.

Fixed-rate lending and IR derivative protocols are about to explode in popularity as new products launch.

An Overview of Fixed-Rate Lending and IR Derivative Protocols

There are a handful of DeFi constructions that can produce fixed-rate loans and IR derivatives. Each construction has pros and cons.

Zero-Coupon Bonds (ZCBs)

In this model, borrowers create bond tokens (yDAI-2021-12-31, for example) by depositing collateral into smart contracts and borrowing yDAI against it. They then sell the yDAI on the open market which settles on a specific date. The effective interest rate paid by the borrower is determined by the discount at which yTokens can be sold on the open market, and the length of the loan. For example, Alice can mint 100 yDAI backed by $200 of ETH and sell the 100 yDAI in exchange for 97 DAI to Bob. Hence, Bob is lending $97 and will receive $100 at contract expiry.

There are many implementations of this approach, such as 1) trading bond tokens primarily on an AMM or CLOB 2) physical vs cash settlement 3) liquidation management.

The zero-coupon bond model is used by Hifi, UMA, Notional, and Yield Protocol.

Pros:

1. Simple construction
2. Supports both fixed-rate lending and borrowing
3. Overcollateralization improves safety for lenders
4. Can build a yield curve on top of the protocol
5. Native token value capture is clear through backstopping liquidation risk

Cons:

1. Fragments liquidity across expiries
2. Need oracle for the collateral/debt ratio
3. Requires depositing collateral into a new smart contract system
4. Capital inefficient for speculators and liquidity providers
5. Liquidation risk

Yield Stripping

In this model, users can deposit a money market token such as Compound’s cUSDC into the yield stripping protocol’s smart contracts. The protocol then splits the deposited cUSDC into two tokens: 1) the principal token (PT), and 2) the yield token (YT).

Depositors can then sell the YTs for cash upfront, allowing them to lock in a fixed rate for the duration of the contract. As an example, let’s assume that the current lending rate on Compound is 10%. If Alice deposits $100, her annual yield—should the blended Compound rate stay the same for the next 12 months—would be $10. She can trade this unpredictable $10 YT for $8 cash today.

Bob, who buys the YT from Alice, is taking a leveraged long view on Compound’s lending rate. To illustrate, imagine that Bob thinks Compound’s interest rate will double to average 20% for the next 12 months. He can deposit $100 into Compound and earn $20 over the next year. Or instead, he can buy 12.5 YTs for $100 ($100 / $8 = 12.5). If Bob is right and the blended rate is 20%, he will get back ($100 20% 12.5) = $250, a 150% return on capital. Of course, if he is wrong and the rate decreases to 5%, he will only receive ($100 5% 12.5) = $62.50, a 37.5% loss on capital.

Yield stripping is used by Swivel, Pendle, Tempus, Element, Sense, and APWine. We are investors in Swivel, and are very excited about combining the yield stripping construction with a central limit order book.

Pros:

1. Enables fixed-rate lending
2. Enables leveraged long speculation on interest rates
3. Requires no liquidation or oracle because principal is deposited upfront and leverage is implied
4. Capital efficient for speculators
5. Safer than ZCBs because collateral is contained with Compound, Aave, etc., which are trusted and established protocols
6. The market capitalizations and liquidity of AAVE and COMP (which act as backstops to protocol risk) have achieved a level that is difficult to replicate for new protocols.

Cons:

1. Does not allow for explicitly shorting of rates
2. Somewhat capital inefficient for rate trading because it requires the entire principal amount up front
3. Limited to principal assets and rates (no exotic rate exposure)
4. Does not allow for perfect hedging on the floating side because of the spread between Compound and Aave’s lending and borrowing rates
5. Fragments liquidity across expiries

Stable Rates

Aave piggybacks the protocol’s variable interest rates with their stable-lending feature. The basic idea is that when a borrower opens a new loan, she has the opportunity to select “stable rate.” Her origination rate will be higher (often substantially higher) than the variable rate she can borrow at. For example, the variable borrow rate for USDC is 5.4% and the stable borrow rate is 11.9%. The variable borrow rate for ETH is 0.25% and the stable borrow rate is 3.3%. This buffer ensures that the protocol remains solvent before a “rebalancing event” in which the stable rate changes for existing borrowers.

Unlike the borrow rate for a variable user, the borrow rate for stable borrowers in most market conditions will not fluctuate with new borrows, new deposits, new liquidations, etc. into the core Aave market.

This is the model used by Aave.

Pros:

1. Trusted protocol in the Ethereum ecosystem, and AAVE has a large market cap and sufficient liquidity to backstop the protocol in size -- as such, likely safer than other models. See this post for more detail.
2. Enables users to borrow at a stable rate (though not guaranteed fixed in weird market conditions)
3. Lots of liquidity because it leverages Aave TVL
4. Highly utilized in more volatile rate markets on lower market cap and less liquid tokens
5. Liquidity concentration because of the loan’s perpetual nature (no fragmentation across expiries)

Cons:

1. No ability to explicitly speculate on rates
2. Only applies to borrowers and not lenders
3. In adverse market conditions users’ rates may be “rebalanced”
4. Does not allow for synthetic rate exposure
5. Borrow rates for stable users are significantly higher than variable

Contracts for Difference (CFDs) and Interest Rate Perpetual Contracts

In this model, users deposit any form of collateral into a margin account and go long or short rates. This construction looks very similar to perpetual contracts on FTX and BitMEX. Funding is paid based on the underlying reference rate the contract points to versus the current mark interest rate on the protocol.

These products offer leverage, and thus require a liquidation engine. In the case of a trader’s account margin falling below its maintenance margin, the system unwinds the position against the perpetual contract AMM (or CLOB).

This is the model used by Strips Finance and YieldSwap.

Pros:

1. Explicit rate shorting is possible
2. Ability to get exposure to synthetic rates that are not derived on chain (e.g., LIBOR or FTX’s BTC perp funding rate)
3. Capital efficient because it does not require principal token up front and allows for two-sided leverage
4. Liquidity concentration because of the contract’s perpetual nature

Cons:

1. Requires an oracle for off-chain rates
2. Liquidations are necessary because there is leverage
3. Requires separate perp contracts for Compound/Aave borrow and lend rates in order to create perfect hedging

Exploring the Trade-offs in Fixed-Rate Lending and IR Derivatives

In examining the various types of fixed-rate lending and IR derivative constructions in DeFi, it’s clear that there are many trade-offs based on design decisions in these systems. In our view, each of these types of protocols will carve out a niche for itself.

DeFi Perpetual Contracts for Interest Rates

The flexibility of the synthetic nature of the perpetual contract enables speculators—and hedgers—to trade rates on any theoretical interest rate (e.g., Compound USDC lending rate, LIBOR interest rate, Fed Funds rate, FTX funding rate, ETH 2 staking yield, etc). Notably, there is no need to roll over a position because these contracts are perpetual in nature and do not expire. As such, there is no liquidity fragmentation across expiries. Speculators drive liquidity, and that drives value for hedgers and other organic users.

Imagine Alice is borrowing XYZ token from Aave to use for her trading strategy. She can, via interest rate perpetual contracts, go long the Aave interest rate for XYZ token such that she is protected in the event that her borrow rate increases and eats into the profitability of her trading strategy.

The perpetual swap contract is not perfect, but it offers a capital efficient way for market participants to bet on rates.

In our earlier post about the design space for DeFi derivatives (highly recommended read) we concluded that the perpetual swap contract will likely be the winning construction for DeFi derivatives. We believe this is likely also true for IR swaps. While the perpetual swap contract is not perfect, its weaknesses are manageable. While the stable rates model is safe for lenders in that capital is kept within Aave, it does not allow for two-sided leveraged exposure to rates and this reduces its utility. ZCBs enable perfectly predictable fixed-rate lending and borrowing, but don’t enable capital efficient rate speculation and therefore have less liquidity.

While the perpetual contract construction for interest rates in DeFi is not yet live in meaningful size, we see a massive opportunity for it. We expect that the perpetual contract, which is already the dominant trading product in crypto, will also work very well for CeFi and Defi interest rates.