Bringing traditional risk analysis to DeFi

Transcript

Hello. >> Hello everyone. >> Good to see you all. Um, my name is Andrew O'Neal. I am from S&P Global Ratings. I am a risk guy. Um, which often makes people grown, but it's important. I get up in the morning thinking about innovation in financial markets in particular and how we build for that in a riskaware and robust way.

Now, if you haven't heard of S&P ratings, uh we've been doing risk assessment in uh traditional markets for the last 150 years or so. If you have heard of us, you probably think of us as the TRDFI guys. Now, here's the thing. Risk isn't a tried concept. It's a FI concept. You walk around conference here, you see these bags saying earn 8 to 11% on your USDT, USDC. You don't get that without risk. And so understanding what risk is behind a given investment can help you make the right decisions as to where to put how much money. And that becomes very important when we're talking about, you know, growing

adoption of DeFi. Whether that's whether you're here for like institutional things done by institutional people in an institutional grade way, etc., etc., as you see everywhere, or whether you're here doing crypto things for crypto people. Ultimately, you want to leverage the resilience of DeFi. You want to leverage the transparency of DeFi to build better products and better markets. So what have we been doing in this space? So a couple years ago we started publishing stable coin stability assessments. So ranking the relative risk of stable coins against each other.

Um last year we actually started putting those assessments on chain. Now that's something that I got quite excited about. My background is a as a structured finance analyst. I've looked at many transaction documents where you see what a structure can invest in or not and when it might have to change things and how that interacts with the risk assessments that we produce. Seeing the very beginning of us starting to put that on chains. We put it on base now through through chain link gives me a kind of window of hope into a future where a new audience is using our assessments engaging with us uh in the uh the insights that we provide. But

beyond stable coins, we've also done ratings on tokenized funds. And last year we did our first rating on a uh decentralized finance protocol uh specifically on the Sky protocol formerly known as Maker DAO. Uh I'm going to use that as an example today to talk through some of our perspectives on risk and how we look at that in a DeFi context. But before I do that, wanted to just have a little level set. So let's say I was to call upon one of you to give us all a 30 secondond overview of what the Sky protocol is or Maker Dow.

Who here would not feel comfortable doing it? Okay, everyone's comfortable. Oh, we've got one one hand. Okay, so anyone else I could pick on at random? Two hands. Okay. All right. I won't do that. So maybe you call my bluff for the if if any of you still need a bit of a refresher on it. It is a D5 protocol originally started on on Ethereum and one of the more established protocols out there that focuses on lending. You deposit collateral in the protocol. You take out a loan.

We put a rating out on it last year that looks at its uh solveny risk effectively. And so I'm going to show here a very simplified overview of the protocol's balance sheet. So the first thing that we look about look at when um when when rating risk is looking at what the obligation is that an entity might default on and what that default would be, how it would play out and what are the risks that make it more likely or the structural features that might mitigate some of those risks.

In the context of the Sky Protocol specifically, they issue a stable coin that is used for their lending. The holders of that stable coin are the creditors of the protocol itself. If you want to think about it as you comparing it to a bank, the holders of the USDS stable coin are the depositors in the bank. And those funds are being lent out mostly against crypto collateral increasingly with an element of of real world assets as well. as a holder of USDS, you want that to be worth $1 or more specifically, you want it to be worth one USDC because the protocol also gives you the ability to

swap one for one against USDC. So there is an expectation that you'll be able to do that and that is what we compare to withdrawing from a bank or withdrawing from this protocol would be swapping your USDS for USDC. So then we come to the credit risk of the protocol. What is that? The risk is that there may not be enough USDC to service withdrawals or that there may be losses on the assets of the protocol that mean that a haircut needs to be applied to holders of USDS. And so that is the default scenario that we're looking at. So when we think about a trady context, default scenario would be company fails to make payments on its obligations. In this

context, it's very specifically a haircut gets applied by the protocol to holders of the USDS stable coin. And so our rating is looking at what is the risk of that happening. So to support its liquidity, to have enough cash at hand to support withdrawals, they hold fairly substantial amount of USDC as cash in reserve. And they also hold liquid tokenized money market funds um that they can um uh convert to cash fairly quickly.

They also hold an equity buffer to protect against losses on their assets. So this is like capital on an institution's balance sheet that's intended to be sized to cover the risk of those assets. So if they were to have more more losses on assets than that capital, that would be a default scenario. If they were to have not enough cash to meet withdrawals, that again would be a default scenario. So, a good question at this point would be, do they have enough capital?

It's fairly uncontroversial for me to say no. Not because of any work that I've done, but the protocol on its own measures is targeting an amount of 125 million uh USDS in reserve. As of today, or as of last Friday at least, there was 42 million there. Now that's known. The Sky Protocol is going through a process called the endgame transition where they are building up capital. They are implementing changes to their governance structure. And at the at the end of that transition, they expect to be fully capitalized. But as of today,

even based on their own measures, their own model, they are well behind the targeted amount of capital to cover risks on their balance sheet. So when we think about where does a rating come out in that context, while a rating on the protocol is B minus, which is for those who aren't familiar with our scale, it's a fairly high-risk indication. uh it does mean that the protocol is able to meet its financial obligations absent significant stress but uh it does highlight significant vulnerability. Now that piece of that piece of the puzzle the capital is really key to that and it's getting more complicated.

You may be familiar with the new Sky structure that uses what's referred to as stars or prime agents which are effectively just subsidiaries uh specialized lending subsidiaries that either specialize in given markets or given protocols, given lending types or even different chains. And the new capital structure of the protocol will be that each of those subsidiaries will hold its own capital. And the protocol governance, the Sky Core protocol governance will set the requirements and is setting the requirements for each of those subsidiaries in terms of how much capital they have to hold.

Now that all sounds pretty good. It's common standards across the Sky ecosystem. However, none of that is enforced on chain. So that I'll repeat that. None of that is enforced on chain. That is entirely reliant on off-chain human risk management teams, calculating the amount of capital that's necessary and verifying that that is present. There is no automated or dynamic mechanism through which capital is resized or increased to the extent that um that the that that there is more lending or higher risk lending happening.

You're entirely reliant on a human decision-m process and human action to actually support that. That doesn't mean it won't work, but that's just an important point to note in a D5 context where perhaps there's an expectation of a trustless system. There's certain it when rubber hits the road, there really is always trust and there's and and we need to understand where those trust assumptions lie. Now, the capital framework is very much in transition. So is the governance framework and in particular key person risk.

So just to illustrate the point, I've shown here an example of a governance vote that happened a couple weeks ago. About 7 billion Sky tokens were put in favor of this particular motion. Now, the Sky Governance system uses this delegate uh construct where token holders can delegate their tokens to a specific delegate. So, the top three of those represent 92% of of total votes. That in itself isn't particularly a concern. That's the way it's supposed to work. But when you look one level deeper, more than half of those tokens are coming from a single wallet address.

Um, and that just illustrates the level of dependency that there is. It's not a secret that many D5 protocols remain highly dependent on their founder, their founding team in terms of how they're governed. It's also not particularly opaque. I was able to map this out in little under 15 minutes, but you need to do that work and you need to understand that there is that dependency there. Now, that may be a good thing. I talked about the transition that the protocol is going through today.

What they're effectively doing is a major strategic shift as an organization. And so it makes sense that you would want that governed by a tight team that's aligned on a particular strategic vision. And indeed that strategic vision is intended to decentralize the protocol over time. But we just need to be aware that as of today there is a very substantial um key person risk that needs to be needs to be understood needs to be monitored.

Now capital governance two elements that are perhaps not close to their full degree of maturity yet. Liquidity is a bit more advanced. So, I talked earlier about how they need to maintain enough liquidity that if people want to swap their USDS for USDC, they're able to do so. And so, they hold USDC as cash and they hold tokenized money market funds. Now, as a risk analyst, I love this. I can view their liquidity position in near real time um and uh and bring this up, monitor how they're doing against their liquidity

targets. They've sized their liquidity targets based on historical data and how how many people how many holders of USDS have sought to leave the protocol in earlier times of stress and so they target a certain ratio of um USDC to total assets in a range of 20 to 30%. to uh to support that. And this slide actually been able to show in some more kind of trifly contexts look talking to corporates looking to adopt uh stable coins for treasury management for example. It's a very useful example of how um the composability of DeFi can allow stable coins and tokenized funds to build a an an automated liquidity

management system for corporates beyond the pure DeFi ecosystem. So this this idea really resonates uh with uh w with that particular audience and because of that the Sky protocol is actually by far the largest single holder of tokenized funds. So we talk about tokenized real world assets. Um a lot of that is tokenized funds, tokenized treasuries. A lot of that is being held specifically by uh the the Sky protocol. And so from a TRDFI perspective, we're now seeing money from the the Sky Protocol coming in as a significant form of investment, a significant source of investment into

more traditional uh financial markets through tokenized real world assets. So this liquidity management system works pretty well. There's one major risk that remains in it though, which is depositor concentration or at least the potential for depositor concentration. So you could you can target with historical data saying okay so long as we have 20 to 30% in cash we'll be able to meet withdrawals. Now what if your depositors include you know or a top five that hold almost all of your assets then you know to the extent they all sought to withdraw at once maybe that liquidity position or that liquidity assumption doesn't hold

anymore. Maybe you're more likely to be in trouble. And that's where this intersects with the concept of permissionless finance because if you have a purely permissionless protocol, anyone can deposit however much they want into the protocol at any time. Now, if that's you or me putting in a few hundred, maybe a few thousand dollars, maybe there's some high rollers in here, a few tens of thousands of dollars, that'll be fine. If you're putting in several billion dollars in cash overnight, then you might have sort of liquidity concentration that could hurt. It creates risk

and that happened. So this happened last year, February time. May remember February 2025. bit of a start to the month in crypto markets particularly in pers and when you look at um funding rates go well going down or going negative is a um protocol heavily exposed to that that tends to react to that situation by moving out of crypto carry positions and into um into stables to earn yield. And so here what you see is start of February 2025, Etheina moving one and a half billion

dollars into the Sky protocol um within a few days. Most of that within one day so that overnight Athena became the well the largest depositor and uh representing about 20% of the protocol's total balance sheet. So long as that is only one that's okay. The cash that they've put in still there. So long as they remain lower than 20%. The buffer will be enough to cover them.

But say you have three, four, five of those that come in within a short space of time and then want to withdraw at once. That's where you have that liquidity risk. So depositor concentration and this is why I come back to that fact. It's not about DeFi. That's not about trfy. This is justifi concentration leverage contagion. These are the things that we all need to be mindful of. In this case, the exposure eventually came down, but the fact remains that anyone can deposit several billion at any time. And so thinking about how you reconcile that desire for a permissionless architecture with risk gating that can prevent you from kind of

ecosystem level risks that might harm all stakeholders is a key part of the risk discussion. But again what's somewhat beautiful with all this is the level of transparency. The way in which this can be monitored in real time. But I'm going to move away from Sky now and talk about when that stops, when risk moves offchain and specifically looking at the stream finance um event that happened in November last year.

So, DeFi is no different from Treadfi in the sense that investors chase yield and to the extent that rates across the market are declining, investors will seek increasingly risky investments to earn yield. That may take the form of looping. It may take the form of less liquid forms of collateral, but it can go as far as off-chain yield strategies.

And when you move offchain, you move away from some of that transparency benefit of DeFi. You may have exposures to leverage, to counterparty risk that the end investor is not able to monitor. And when that happens, you you have the potential for capital to flow into common buckets. Everyone's trying to improve their APY. They see one, two specific vaults that are um supporting that and so the money tends to flow towards the same honeypot.

And so you have multiple protocols, multiple vaults exposed to the same underlying risk that not everyone is able to monitor at the speed at which they should be or they think they are. So you have losses that happen offchain that flow into a given protocol that flow into vaults that are invested in that protocol that flow into vaults that are exposed to each other and and ultimately you have um maybe end investors who weren't expecting to be affected by that kind of contagion event.

Transparency. Transparency is key to the DeFi story. And we can bring it then to what happened just over a week ago with Resolve because here we can see something quite interesting. I'll go over the the salient details quickly if anyone wasn't following. Uh but um you know what what happened on um Sunday 22nd was that a hacker managed to mint unbacked tokens unbacked stable coins from the um the resolve uh protocol. So US stable coins managed to stake those and uh um and

then swap out those uh stake tokens against uh ETH collateral in particular. But where it's so there's maybe one key message from that is security around minting processes is always critical. But that's probably not new knowledge. What's more interesting here is how that then intersects with DeFi lending because within minutes of that event happening, you saw a collapse in the value of the uh staked us token and there was an arbitrage opportunity that existed because that that drop in value was not synchronously reflected in

various lending protocols. Morpho among others. Now, Morpho itself doesn't um doesn't select the forms of collateral that are eligible. That's that's a choice made by curators. But what we see here is that you staked USR was onboarded as collateral in several vaults a few months ago. The total liquidity on the staked USR USDC market was around $5,000 at the time. It's never been higher than 50,000.

And yet users were able to deposit $5 million or more, I think 6 million in total, um, as collateral and take out loans in USDC. How does that happen? Well, you have the asset onboarded as eligible collateral among a long list of assets in a specific vault. Perhaps an expectation that there's never going to be a large amount deposited because there's only $5,000 in daily liquidity. But because of this permissionless nature where anyone can go and deposit then if you have that large dislocation in the market where there is a

difference in the price feed going into a morpho vault and the price at which the token can be acquired say on unis swap you have this arbitrage opportunity. There are vaults now that are stuck with nominally $5 million worth of um that token as collateral likely will lead to losses for depositors into those vaults. And again, it comes back to this question of reconciling permissionless architecture with effective risk management.

So if you're building for a system that is intended to be permissionless, that is intended to be used by anyone anywhere all the time, it is still relevant to consider that if anyone any time means factors of thousands of the daily liquidity of an asset, then that may need to be somewhat constrained. This all gets a lot more complicated when you bring in real world assets.

Now, I'm going to be brief on real world assets because I know that the next uh the next presentation will be covering it. So, if anything, this might just give you bit of a warm-up, a few questions to ask them. But I wanted to start with some fundamental differences. So if I hold ETH in my wallet, I'm holding the asset. It is not the representation of an asset that sits elsewhere. It is the asset itself.

When I'm holding a tokenized real world asset, by definition, it is the representation of something that exists elsewhere. And that brings many considerations that make the whole tokenization of real world assets complex, not impossible, and certainly not short of benefits, but complex. And so to if people are frustrated at how maybe slow the development is of an established real world asset market, that's why these things take time. There are legal considerations, operational considerations liquidity considerations. So when I buy a tokenized share in a fund,

is that does that make me a shareholder in the fund? Is that the same as if I'd bought a traditional share in the fund or am I buying it through a vault in DeFi? And if I do that, well, what do I actually own? Because you may well own nothing more from a legal perspective than an IOU from some shell company in a sunny island somewhere that says they hold the asset and promises to pass on the um uh promises pass on the yield from that asset.

If you were to have to contest that in court, well, it might make for a nice holiday to that particular jurisdiction, but your chances of success might be somewhat constrained. Beyond legal considerations, they're operational considerations. Real world assets don't usually price at the same pace as crypto assets would price, and so using them as collateral in DeFi becomes problematic. If there's only a new price of the fund once a day, how do some of the architectures that exist in DeFi around liquidations work?

If that particular asset can only be held by a per permission set of investors because it can only be issued in a specific jurisdiction or because the fund manager needs to keep a record of their investors. How can you assume liquidity in that asset? So when we look at the tokenized funds that we rate for example, they're all limited to onboarded whitelisted investors, you can't if someone breaches their liquidation threshold, you can't just liquidate that permissionlessly in DeFi.

You can it can only go to another holder in the fund. So when we think about vaults backed by real world assets, either you have to move away from that liquidation concept which you know is a possibility or you rely on liquid source of liquidity for example the sponsor of the transaction contributing to uh to the fund to do that now that can work but again introduces dependency so there's a whole architecture of legal operational liquidity considerations that are different to traditional I'll a traditional crypto assets

that need to be reconciled with the composability of DeFi and that needs to be understood from a risk risk picture and I don't want that to sound like I'm ending on a negative note because I think risk is good. I think risk analysis is good. I think robust conversations around risk is what helps us as a community build innovation, new better financial markets. Whether that's the institutional grade stuff, whether that's the crypto stuff, robust conversations about risk, please.

And with that, I will take any questions. Yes please. >> So, if um people didn't hear the question, uh question was would we give a single A rating to anything in DeFi? Uh I think I would I would say one day probably but but that's where it's important kind to understand what the what the current state of affairs is. Um so for example we rate tokenized funds. Uh well now

some of them get have now got to AAA rating um on our on our fund rating scale. So it's not necessarily just a tokenization or a blockchain pro. It's not it's not a blockchain risk that necessarily or a technology risk I should say that that that that that really drives ratings. But when you're looking at DeFi specifically, you have to appreciate that a lot of these systems are still maturing. Um and that's the example that I walked through on Sky showed you maybe flaws in the capital framework, flaws in the governance framework, similar flaws exist elsewhere. Um so

you know when I think about financial institutions for example that get to single A ratings that's generally because they operate in systems that provide a substantial amount of government support and access to emergency liquidity. That's the sort of thing that isn't as of today readily available in DeFi. So I think we're probably some distance away from that for a rating on on a D5 protocol. But I think what we are seeing is gradual maturing of the ecosystem. And so that's why what we've tried to do in our rating on on Sky in particular is to point out not just where the risk picture is today, but how the protocol is looking at implementing structures that might reduce some of those risks.

Hopefully that's didn't completely sidestep your question. Yes please. >> Hi. Um, there's there are credit ratings, fund ratings, stablecoin ratings. There could be tokenized credit, tokenized funds, tokenized stable coins. How would how would how can you interpret a rating across those? Is an A rating the same across those? credit obviously is senior in the capital structure. So I'm wondering if it's different than a rating across the others.

Sure. So so we have um separate rating scales um that uh have different definitions. So for example a even in within tradfi a fund rating is not the same as a credit rating um because it's it's looking at a slightly different aspect and that point that I started at u as in what what would be a default so in the context of money market fund a default would be uh you know dropping I think 50 bas more than 50 basis points uh below below the buck um whereas in uh a bond context it's failing to meet interest or or principal payments. So our rating on the Sky protocol is on our credit rating

scale and so it would be com it's comparable for example to a bank that would be rated single A. Um it follows that kind of conceptual framework. Um our stable coin assessments are a relative ranking of stable coins against each other. And so that that was quite deliberate to have kind of uh and I guess it depends on really the use case for for the assessment. So what the question we asked ourselves is what are people going to be comparing to? So when we started doing stable coin assessments two years ago our sense from our conversations with market participants was that people are looking to compare stable coins against each other and so uh and and so that's what uh what we built there. um in whereas in the I

guess in the in the Sky context, we're looking at something that's much more much closer to a a credit rating and something that's um uh you know provided I think a good degree of comparability. >> Okay, we got one more maybe I'll go first. Uh >> yeah, sure. >> Yeah. So, uh are the ratings public? Like can we read the ratings and the logic behind them somewhere? >> Yes. and and then maybe second question uh like uh who are the investors that um you see coming into the class and how is that changing so I think I mentioned already in one

direction so like say money that's flowing from sky in particular into traditional ecosystem and in particular I think about like the US CLO market it's an area of structured finance Um I think about 3 years ago the in the US um CLLO ETFs became a thing. Uh and there's about 30 billion now of of CL ETFs. It's about 10% of the total CLO market in the US. About one or one and a half billion of that is now owned by Sky. So you're talking about you about half a percentage point of the total US CLLO market is now owned by a D5 protocol which you know is I think interesting in itself. When we flip that around and think about like institutional money coming into DeFi, I think it's very early, very specialized still in terms

of um being being more I guess cryptofocused cryptonative investment uh and cryptonative funds looking to uh place money into DeFi. as we see more adoption of let's say vault technology around real world assets that will ch that will change but we're really not there that uh that example that I gave about only holding a IOU from a shell company in a sunny island that is the reality of a lot of uh current implementations of real world assets into actual DeFi so whilst that's the case It's only going to be a kind of experiment experimentation playground.

I think I have to wrap up. Can I squeeze one in? I'll squeeze one in. He's not looking. >> Yeah. Thank you. Thank you for the talk. Uh I have one question about I'm wondering if this serious talks between risk and permissionless architecture which are important could be uh like an obstacle you know because could be like a bottleneck. So what are your insights in how to make risk not be the bottleneck of this massive adoption of the permissionless architecture? >> So I think you can the easy way to not have a bottleneck is to just ignore it.

Um and then you end up in a situation where something blows um and uh and I think then you're set back maybe permanently. So I think bringing those conversations up front um and kind learning from events like the ones that that that I was describing here and and seeing seeing what can be implemented in terms of uh you know restrictions. You know one of the reasons and and you know here's the reality. One of the reasons why there aren't more codified restrictions on how much you can invest is not just an ideological thing. It's also because that increases computation, increases gas, and so increases cost. Um, so for

someone who's trying to advertise the best APY they can, then it's not particularly tempting. Uh, but that's where that becomes relevant in in a when you take a longer term view. Um, maybe taking on that initial cost up front, but uh, but but being more resilient. And with that, I think I'm going to get pulled off. Thank you very much. Well Love.