Real-World Asset Protocols: Centrifuge & Goldfinch
An overview of Centrifuge and Goldfinch, and the unique complexities and opportunities they bring to on-chain investing.
I cross-posted a similar version of this content to newsletter.stakingrewards.com.
When you hear about real-world assets (“RWAs”) in crypto, it’s common to doom-scroll through Twitter threads like this:
This particular author’s core objection (that integrating fiat law into the programmatic law of crypto dramatically increases fixed costs while taking away a lot of scalability) is not wrong. Since RWAs — loans to small businesses, for example — are ultimately loans to real-world businesses, they must ultimately comply with non-blockchain (fiat) law in order to be legally enforceable. A core use case of the blockchain, the replacement of cumbersome fiat law with lean programmatic law, is thus negated when RWA’s enter the picture.
The issue of yield sustainability is most acute for the blue-chip crypto-native banks (Aave, Maker, Anchor, Venus, etc.) All of them have paid for, or heavily subsidized, yields via inflationary token emissions and/or ‘debatable’ deposit economics. All of them have found a profitable core use-case, such as overcollateralized borrowing against crypto holdings in lieu of incurring capital gains taxes (Maker), flash loans (Aave), or repackaging LUNA staking yield as an overcollateralized loan (Anchor), that’s significantly smaller than their current deposit base. All of them have either run out of new tokens to issue, or (in the case of Anchor and younger deposit protocols) can clearly see that they’re on an unsustainable path. Inflationary token emissions which seem to work well in a bull market aren’t cutting it anymore. They need to find productive uses for their excess capital. And that’s where RWA protocols come in. Think car loans, invoice factoring, and other preferably low-duration, low-regulation, high-interest types of loans to small businesses.
However, RWA borrowing is infinitely more complex than digital-asset borrowing. Digital assets are marked to market to the second. After compressing 50 years of tradfi bull/bear cycles into 5 years, it’s fair to say that crypto handles digital asset risk very well. You either have enough margin to meet your borrowing or you don’t. If you don’t, you’re liquidated immediately. This happens regularly, with little to no ecosystem disturbance, no cries of imminent financial catastrophe, and no bailouts. The system just works.
RWA borrowing introduces a witch’s brew of booby traps into the protocol risk management process:
Liquidity. Real-world SMEs don’t have the capacity to mark liquidity to market to the minute. If liquidity evaporates because a degen took on too much leverage, the secondary market needs to be deep enough to absorb a liquidation.
Duration. SMEs need to be able to lock up borrowed funds for a minimum of 1-3 months at a time. This concept is alien to most of DeFi.
Fiat-law integration. Blockchain contracts need to accommodate real-world law to be enforceable. This is difficult, because smart contract law is alien to most of the world. It also cuts out most of the promised efficiency of smart contracts in the first place, and knocks out a leg of the scalability promised by blockchains, if every smart contract needs a fiat-law lawyer to greenlight it *and* monitor it. If we all have to pay lawyers more money to adjudicate smart-contract law than we do for fiat law, it kind of defeats the purpose of smart-contract law in the first place.
Asset quality. Let’s say I mint an NFT that says I have $1m worth of loans. Who judges whether I really have $1m of loans, or $1m of garbage that’s really worth $1?
As a counterpoint to the above, since everything is by-the-fiat-law-book, RWA debt does have legal recourse of normal legal contracts. In other words, you can theoretically go after someone for misrepresenting themselves. However, this has never been battle-tested.
Fraud resistance. Any decentralized loan platform faces major incentives for collusion at the expense of senior debt holders, with potentially lethal systemic consequences if things go wrong.
RWA Blockchain #1: Centrifuge
The broad category of RWA protocols includes RWA-dedicated blockchains, like Centrifuge and Goldfinch, as well as RWA-capable pools, which you can find on protocols such as Maple, TrueFi, and others. The former category, which offers a blockchain backbone for implementing real-world legal agreements, is what we’ll focus on here.
On Centrifuge, for example, you can of course stake your CFG for an APY of around 11%. However, you can also invest in actual loans to actual businesses: senior debt at a 5-8% stablecoin yield (it depends on which pool you choose) plus a variable CFG yield (which was recently changed, to around 8%), for an all-in very-low-duration, senior debt yield of around 15%. The junior debt typically offers a cash yield of around 15% and an additional CFG-denominated yield of around 25%.
Furthermore, you can drill down into the performance of every single asset in each pool.
A blockchain built around fiat law requires extensive KYC. Any investor needs to submit the same basic KYC information that they’d need to submit at a bank, and Centrifuge is not open to US investors at all unless they submit to an additional investor-accreditation process. Centrifuge’s “Real-World Asset Market with Aave,” in which Aave depositors can invest into some of these pools through a complex flash-loan mechanism, doesn’t allow US investors at all.
Additionally, because a Centrifuge smart contract needs to handle many more variables than, for example, an Anchor protocol smart contract, it needs to make many more smart contract calls. These calls are insanely expensive. I have heard that even minting a single NFT on Centrifuge (which is how Centrifuge memorializes a real-world lender/borrower agreement onto the Centrifuge blockchain) can cost $500-$2000. A single pool of loans might have 10 or 20 such NFTs rolling over every month, and Centrifuge eats most of the cost. This must be insanely expensive for Centrifuge to operate, driving lots of CFG emissions and doubtless also driving Centrifuge’s decision to Ethereum for Polkadot: it’s way too expensive, even with low gas fees, to run a real business with anything resembling “ordinary” transaction sizes (sub $100k) on Ethereum. In the distant future, when oracles monitor real-time performance of real-world assets and match it to loan covenants & asset performance, the data load of RWA smart-contract self-sufficiency will be dramatically higher than it is today, when such asset quality diligence is done off-chain.
But for all these faults, Centrifuge offers something very few other protocols do: an unambiguous source of sustainable, real-world, non-tokenomic investor yield. They have also made the first real attempt at bridging fiat and programmatic law in a way that drives real-world value. I also think Centrifuge’s Polkadot migration (I would guess it takes effect by the end of May) will be a big positive catalyst for Centrifuge if they can pull it off, since it will probably slash Centrifuge’s opex dramatically, allow them to charge fees, and also reduce fees for users.
RWA Blockchain #2: Goldfinch
Centrifuge’s implementation of on-chain structured credit puts the developer first: data composability and transparency are high, but so is operating complexity and structural transaction cost. Goldfinch’s approach arguably offers less composability, but also streamlines the process, resulting in a smoother user experience. As a result, Goldfinch has somewhat higher TVL ($114m) than Centrifuge ($85m) despite being significantly younger, and a very vibrant community. (The DeFiLlama-type TVL stats for these two protocols are wrong, and understate their TVLs significantly.)
Some necessary elements of real-world credit, like a default legal structure that’s flexible and easily adaptable to different countries, don’t exist at all on Centrifuge. True decentralization of the platform is a very long way off. (Permissioned-ness is highly centralized by definition, but many individual functions, like asset diligence, credit ratings, or asset-originator vetting, could theoretically be decentralized more quickly.)
As a significantly younger protocol, Goldfinch seems technologically behind Centrifuge in some respects. However, Goldfinch made up for its late start with several significant iterations on the Centrifuge model which, in my view, better balanced user-friendliness with user flexibility.
Instead of having a senior and junior token for every pool (like Centrifuge’s pool-specific DROP and TIN tokens), Goldfinch has one senior debt token for all its pools (FIDU) whose exposure to all Goldfinch pools is programmatically adjusted based on the skin in the game of the pool sponsors.
Goldfinch leveraged its founders’ experience with emerging-market debt investing to create a robust off-chain legal structure, which significantly reduces operational overhead for users.
Goldfinch pool landing page:
An example Goldfinch pool overview is as follows. (The author has no involvement or investment in this pool or any other pool or platform featured in this review. The author has limited experience interacting with both protocols as an accredited investor.)
Because Goldfinch and Centrifuge took philosophically different approaches, one’s strength is the other’s weakness. Centrifuge pools offer daily liquidity for either tranche — conditional on someone else being able to take your place. Goldfinch junior pools, on the other hand, seem extremely illiquid, ie, your funds are locked up for 12-36 months (36 months in the above case). Your junior token is represented by an NFT which you can’t seem to exchange. I guess this is a good way to source very high-conviction backers, but 45-75% APYs for 3-year-locked-up EM debt (with no in-between in terms of secondary liquidity) seems very suboptimal to me. However, Goldfinch’s simplified approach dramatically reduces operational overhead for users.
On Centrifuge, you can see whether every single loan was repaid or not. Goldfinch doesn’t seem to have this feature. In general, Centrifuge traded significantly more granularity/data composability in the future for user difficulty in the present day.
Based on my limited experience thus far, one thing both platforms could use is a data room which offers meaningful, current detail on asset performance (a key tradfi structured credit requirement). I think both platforms could benefit from enforcing a mandatory form and cadence of asset quality reporting. Their current data rooms seem to have highly variable quality. I would assume that their borrowers are less familiar with tradfi reporting expectations, but some sort of tutorial to set asset-reporting expectations and cadence would be nice.
Then again, Rome wasn’t built in a day. I would assume both platforms have asset-quality modules somewhere in their project pipelines, and this feature currently belongs much more in the ‘nice to have’ bucket than in the ‘must-have’ bucket relative to more basic priorities, like managing transaction costs down.
Goldfinch and Centrifuge represent the two furthest-evolved and most-ambitious attempts at sustainable yields at scale for DeFi, beyond the familiar use cases of overcollateralized borrowing. By attempting to bridge fiat law and programmatic law, they are taking on a much more ambitious ask than many other DeFi protocols. As they introduce new challenges, though, they’re also offering the first scalable solution to DeFi’s biggest headache: how do we scale DeFi yield without seigniorage?
thanks - this was really helpful. would love to connect and chat if you're open to it!