Loading summary
Transcript74 lines
- [00:01]
A
Hi everyone. Welcome to Unchained, your no hype resource for all things crypto.
- [00:04]
B
I'm your host, Laura Shin.
- [00:06]
A
Thanks for joining this live stream. Before we get started, a quick reminder. Nothing you hear on Unchained is investment advice. This show is for informational and entertainment purposes only, and my guests and I may hold assets discussed in the show. For more disclosures, visit Unchained Crypto.com Bitcoin
- [00:21]
B
changed how money works. Citra changes how Bitcoin scales. With a trust minimized BTC and a native stablecoin ctusd, you can finally access Bitcoin capital markets with lending, private privacy, Bitcoin yield and more. Put your BTC to work at Citraya XYZ. Unchained. EtherFi is giving Unchained listeners 15% cash back on food and ride apps. And that's on top of the 3% you get on everything else your bank is charging you to use your own money. I switched Go to Ether Fi Unchained
- [00:52]
A
to claim your discount. Today's topic is how defi rates can properly compensate for risk. Here to discuss are Tom Dunleavy, head of venture at Verus Capital Capital, and Adrian Kachero Vasilievich, co founder at Steakhouse Financial. Welcome, Tom and Adrian.
- [01:12]
C
Thanks for having me. And us.
- [01:14]
D
Hey, Laura.
- [01:16]
A
So this month is probably going to go down in crypto history for defi hacks. And when I say that, I'm also hoping that that's the case because I'm hoping that this is the peak and not just like an appetizer. So I'm just going to list. I just did a quick search to figure out what are all the hacks that have occurred in this month. So obviously Drift and kelpdao. There's also Rio Finance, Wasabi Sweat Foundation, Pulse Fall. By the way, some of these were like literally in the last 24 hours. Crestao Aftermath, Singularity Finance, Hyperbridge, Notify, Aeroswap, Cowswap, Per Lend, Giddy, Syndicate Commons, Bridge, LendHub, Zetachain, Judao, Gallup, Volo, Quant, Kipseli, Juice Box. Now I'm getting into the really small ones, juice box, V3 and theta nuts. So you guys, that's a total of $606 million in exploits, obviously. So we're mainly going to be talking about defi lending. Obviously not all of these directly impacted lending. But as we saw with kelpdao, that doesn't always matter in a world where, you know, as what happened. In that case, defi lending ended up being part of the step for the hackers to cash out. So naturally, like this has led to a lot of conversations around what is kind of like the appropriate risk or how you even price the risk of, you know, borrowing and. Sorry, of lending in defi. So Tom, you wrote a piece on this that went through some calculations of what you thought the fair yield would be. There's been a lot of response to that, but it wasn't the only piece. There was another one as well, which we'll get into in a bit. But why don't we just have you start with you explaining what inspired you to write the piece and how you think about this problem.
- [03:07]
C
Yeah, thanks for having us Laura, because I think this is a really important conversation. So I spent a number of years in traditional finance, specifically in fixed income, so looking at defi yields for my entire time in crypto, which is now over five years, I've always kind of scratched my head and not really understood why yields were effectively so low. And if they were higher, it was generally because there was some subsidation with points or other mechanisms that we would classify as maybe not true yield for the underlying asset. But this came about specifically because of the hack. So I saw the hack and I said, it's obvious to me at least that we're not assigning risk premia correctly. So why don't we disaggregate that and step through what the correct risk premia in my view potentially could be based on some level of data that we've seen and when. Just to give listeners a framework here, if you're pricing a bond in traditional markets, you look at the risk free rate, which is generally proxied as U.S. treasuries. And as Adrian noted and others noted to my response, and my thread is you could adjust that based on time horizon, so maybe use an overnight rate or something if you're looking at more short term yields. But then you add on risk premia after that. So illiquidity risk, some level of default risk, whether, depending what market you're in, some yields for convenience or what have you. And each borrower is going to have their own subset of risk premia that are important to them. But at a high level there should be a baseline of some level that we as an industry agree are important to understand. And this is how you price risk across Treasuries, corporates, loans, private credit, whatever. You take the risk free yield and you add on levels of risk premia. And that's where the basis of the analysis started.
- [04:56]
A
And Adrian, how about you? How have you been thinking about this problem?
- [05:01]
D
I mean, we don't think about it very differently, actually. So as curators, one of our main responsibilities is to underwrite the collateral markets that we expose the vaults to. And in order to understand what. And the way that we segment the risk of the vaults is broadly speaking, around liquidity. So larger assets, larger, safer, et cetera, assets with better credit risk profiles we would call prime and longer tail or less liquid assets with an element of credit risk, mostly liquidity risk, we would call high yield. So we think about it in very much a similar way. We have this risk management framework, it's published on our documentation site and we update the ratings once a day based on how the issuer changes and so forth. And generally what we try to do is be what we call really boring. So there is a temptation in the curator space to accumulate TVL by pushing the APY as high as possible. And we try to take a more boring approach, be a little bit more conservative and limit the amount of credit risk that we expose the vaults to. Because our view is fundamentally that this primitive of on chain repo is sort of the bedrock for future traditional, that the distinction between traditional finance and defi will eventually sort of dissolve and that this architecture is suitable for doing repo transactions. But nobody talks about breaking the buck on money market repo overnight. And so I do think, I agree with Tom completely that we have to go to a place where these types of transactions can be done with a transparent appreciation of the risk that both sides are taking. I think that's really at the crux of the issue where we agree the most. I have no fundamental issues there with the. To maybe touch on like the article, but I, I feel like the article, like there's a lot of commentary on Twitter. People have written articles back and forth. Like people can read those and like I'm happy to like talk about my response or Tom's article and what, what have you. I feel like maybe the more interesting part is like the conversation between us on where we think there is more or less risk. So like, one of the areas that I felt could have used a little bit more detail was being a bit more specific. Tom had this general framework for defi yields. Our view is with primitives like Moro that allow isolation of risk, this actually makes it easier to price that risk and to allocate accordingly. So one of the things with, with Luca's article as well, when he made his calculation, you know, our benchmark is sort of the, the coinbase defi lend experience where Coinbase users can borrow against their Bitcoin using Defi Rails. This is a pretty substantial loan book priced at like between slightly over SOFR. At the moment that's it's at like 43 or 4, 5. It ranges near SOFR and in our view this is pretty much the market pricing what this repo transaction probably should be. It's not to say that it's riskless. There are certainly a number of risks that could substantially impair the lenders on the other side of the equation. But the main point was fine, but we should probably get more specific and avoid tarring the whole space with a broad brush.
- [08:35]
A
Sorry, yeah, we're going to get into those weeds more toward the end, but let's just start with because I noticed both Tom and Luca's piece. So for the audience who didn't hear. So Luca Process Prosper wrote on a substack Rhodes about this topic as well, and both he and Tom started with calculations of how Tradfi calculates this type of risk, or equations, I should say, of how Tradfi calculates this type of risk. So why don't we just like start with, with that? Because I think having that contrast between how tradfi does it and then what is different about Defi and what factors need to be added into the equation to come up with a fair value. I think like that process in and of itself is really, you know, educational. So yeah, either one of you can go first here, you know, talk a little bit about how Tradfi typically, typically calculates this type of risk.
- [09:35]
C
So I think there's, there's two things here. There's the empirical academic observation which is useful as kind of a baseline, and then there's the actual market rate that folks are trading against, which some would call more true rates. Now let's call the academic observation looks at the probability of default and then the loss given default. So what is the probability of default for each individual risk premia that we can assign and then what is the loss given default of each one of those individual risk premias? So we have a call it 10 ish year history of DeFi. Fantastic. Obviously there's a lot of regime changes in there, there's a lot of technology changes in there. So that's certainly not as reliable as a traditional finance bond market which has had 100 plus years. Despite some changes in there, folks would be, I think surprised to learn that most bonds still trade bilaterally and a lot of times through spreadsheets and phone calls. So there's still A lot of still inherent bias in there in the traditional markets that hasn't changed over time. So that's generally the framework and then the way I thought about it was taking that risk free rate and then assigning individual risk premium on top of that. And we can get into that as it makes sense Laura, but you know, the framework is loss given default and how often these are going to default depending on what you're looking at.
- [10:52]
D
So sort of like yeah, so I, I, I, I completely like this is a completely orthodox framework. Like it's very rational way to think about the marginal lender. So if you're a lender with $100, at what rate would you not be willing to lend provided you believe the assumption for expected losses is X? Right. That's kind of what Tom's approach was. And here again like I do think it's important to try and narrow down. So like okay, not to slam ave, I'm not super personally familiar with aave, but it's also a model that we haven't spent a lot of time thinking about because it's pooled. So it has a characteristic where a lot of these factors are confounding because you have cross collateralization in this lending market it provides for a lot of capital efficiency for the user because they can take out a loan with the collateral that somebody else can borrow essentially, which was part of the reason the contagion was able to spread so quickly. We've only really worked with the morpho isolated model which kind of meets our philosophy of what defi could be, which is, or what it should be, which is, you know, it needs to be really ossified, it needs to be really small and hard. These risks are real. Like all of the risk premier that Tom is alluding to in his article Oracle, you know, admin keys, all of this where defi. What defi does best is very little and as stupidly as possible with as many guardrails as possible. And we call this distinction a crypto guarantee versus a social guarantee, which would be an admin key, for example. And in Tradify you have very good safeguards against. You have very good social guarantees. So if something goes wrong, you take them to court, you recover your assets. Crypto doesn't have that dependency because you depend on a settlement layer that does everything very quickly, finalizes very quickly and doesn't require an intermediary to settle the transaction. That unfortunately also means that if something goes wrong, it goes wrong big and immediately and it propagates extremely fast, which is why we believe that having primitives that are smaller in surface, more difficult to manipulate and do fewer things, but just consistently and better is kind of like the Uniswap V2 and V3 model of like it's just an AMM, there's no governance or minimum governance that doesn't really change how the thing works. And Morpho is kind of in that vein. And what that allows is actually more transparent pricing around risk because candidly I would not have been able, if I had the time, I still would not have been able to come up with a view on what the rate for AAVE should have been. I have quite a clear view on what I think the over collateralized bitcoin lending rate should be, which is pretty much where it is.
- [13:49]
A
Yeah, yeah, just something more contained and easy to parse. So now let's talk about what risk exists in DEFI that doesn't exist in TradFi, which Tom's piece went into and so did Lucas. And there's probably more, more risks than what you guys identified. But Tom, why don't you go ahead and start with what you called out.
- [14:14]
C
Yeah. And just to briefly extend Adrian's point, I think it's very important to disaggregate what is curated Defi and Adrian and his team do a fantastic job avoiding a lot of the risk parameters that we're going to touch on here because they're a curator and they are effectively paid to underwrite risk. But you know, that's somewhere like 2 to 5% of the overall DEFI market. And correct me if I'm wrong there Adrian, but you know, the majority of DEFI is, actually does not have that risk profile. And that's kind of what I was getting to. Right. You have the majority of DEFI that needs kind of this blended rate to appropriately assess where a baseline level of risk should be up and above these, let's call them very high quality quasi corporate bond type loans that Adrian's underwriting. So just to get into a few of the things here that I identified and I'm sure there are more that I may have missed. But you start with that risk free rate and then you add on, let's call it, I called it technical expected loss, which is, you know, hacks or exploits or whatever. And I just basically use the same framework that I outlined before. I took the aggregate amount of DeFi TVL and I said annually, what is the default rate on that? And historically it's been about 0.5 to. Unfortunately this year we're looking to towards a 2% annualized rate on defi defaults. And while recovery rate, the other part of the equation has been really strong. And Adrian's pools. Right. I think you said it was almost zero. Sorry, almost 99.99% in terms of recovery rate in broader defi. That's actually not the case. Right. I mean North Korea is not returning anything and then you have white hats and others who maybe return 10 to 20% historically. So that got me to about another 1.5% risk premium on top of the risk free rate. Oracle manipulation, another one. I don't know if you guys remember mango markets, cream, etc. Same calculus there. How many exploits and how many hacks or have we had specifically due to Oracles and outlining what you've had on the recovery rate side. And I think this is distinctly different than sort of the technical, you know, code related issues. So you know, folks said I may have been double counting there. I still don't think so. I think there's code related issues and there's specific Oracle related issues and those are two different attack vectors. Next one I added on top was the governance and risk premia for the social layer. So this is everything from Drift to Ronin to Sweatcoin. I mean I think it was like literally this morning or yesterday. And you can extend that to Buybit and others. Same sort of calculus there that is in my mind clearly different than the others. And then there are a few new risks which we're starting to appreciate right now. So these are the rehypothecation and collateral risks that you have when you have this, let's call it exotic composability collateral. So insert your few letters and then eth after it. Right? And then like using that as collateral and another protocol. Those risks are really, really hard to quantify and underwrite, particularly in a business like Adrian's, which is why I think he doesn't accept some of these collateral like rightfully so, but a lot of defi does and they don't understand those risks and they're just taking them on. Then I added a few others, you know, kind of regulatory asymmetry which I think you could argue about if that's a real risk or not. And is that diminishing as you have clarity with reg with clarity and Genius and others and then just sort of like a plug figure that you have in most traditional financial models which incorporates a level of uncertainty in your model and you could just think of that as like long term capital management. If folks don't remember that, right. Like, these are the smartest hedge fund managers in the world. They, you know, went to like 12 sigmas and how they did the risk models, but there still was that like complete uncertainty that they had at the tail end of the model that cause the entire thing to collapse. And most models have a traditional plug figure at, you know, 1, 2, 3% for that. So that's how I got to my 12 and a half percent number. And there's certainly a range there. And as I outlined, I think it's really dependent on who you're doing business with and your counterparty risk and the protocols you're using. But it felt like a pretty good baseline number from my assumptions that I kind of outlined there.
- [18:20]
A
So, Adrian, what. Go ahead and outline the risks that you feel tradfi. Sorry that DEFI has that tradfi doesn't.
- [18:31]
D
I mean, I don't fundamentally disagree with the list. It's. I think Tom did a pretty decent job in being quite exhaustive about all of the sources of risk. Enumerating them is very complicated because you often, like, you don't know if there are some new risks around the corner that you hadn't, that you hadn't imagined before. I think the probability of default and loss given default model is maybe more intuitive for like a simple explanation. And I think it illustrates very much the sort of unique risks of DeFi, which is, you know, you have a situation where the probability of default is very difficult to define. You don't know at what point, like if you're holding the kelp eth, you don't know at what point an attacker is going to exploit one or many vectors to eliminate it and to create the insolvency. And the issue with DeFi, given the fact that the crypto rail is decentralized and uncensorable, is that the loss given default is almost total. So this makes it very, very difficult to actually quantify in that way. And I would say from this perspective, I think the operating model relative to traditional finance has to be completely different. And one of the reasons that we've perhaps seen so many of these hacks is that you've had operators who, in wanting to build essentially a fintech product, applied a mentality or point of view or perspective that's relevant in tradify and probably would have been just fine as a neobank product. But when translated to defi exposes the vulnerabilities in a. In a catastrophic way immediately as soon as there is one. Right. It doesn't. You can present this as A risk or as an opportunity. Because I also think that the flip, this is the dark side of the coin of Defi, which is actually with Defi you can build systems that execute automatically with no intermediaries, with full margin compression from disintermediation. Where for example, one of the. Well, we, we recently worked with S and P to issue a formal credit action, a credit rating on, on sky ecosystem as an issuer. And one of the early sort of mind shift that we had to sort of coach them through was when they were asking who where's the court? Who's enforcing. The reality is, is that in the premise of a crypto defi settlement layer is that you don't need that second order of social guarantee enforcement. The rules are defined in the ideal. The rules are defined up front, the participants know what they are agreeing to and the enforcement is automatic. I think again, coming back to sort of the Uniswap or the morpho model, when you have a system that works very well, like over collateralized bitcoin lending with stablecoins, you have a market has predefined risk parameters, it can't be changed. When a borrower or a lender come in, they know exactly what the rules are and therefore are able to price it much more efficiently. In my view, one of the ways that we can actually get to a more transparent pricing around the whole of Defi, which Tom is sort of looking for, is by doing that, like actually going simpler and then building more complexity on top through these simpler modules. Like you can build very complicated, you can build a very wide range of rich complexity with simple primitives. But the primitive itself needs to be super hard. It needs to be ossified. Like the Nikolai from Maker had this concept of incentive compatibility. And what can, what can go wrong, will go wrong. As, as, as part of like his framework of thinking of building and Defi. I think that mentality, I think can move the space forward. But as to whether the rate is like 12 and a half or seven, we take more of the view that yeah, we don't spend that much time looking at what other curators or what other venues do. There's enough, enough to do with our pools as it is. We feel like the market rate, like you can argue with the market, but the market is the market rate is the rate. And so the high yield rate is 6 or 7. So like that is what the rate is. There is maybe a question of what is the rate that I, that the next marginal lender will require in order to get in I think that's a reasonable question to ask. Our view is that they will eventually come, and they are eventually coming, but it won't be with fintech hybrids ported onto crypto because they create these operational security nightmares. It'll more likely be, for example, replacing the ledger for an asset backed financing platform and having it run on smart contracts to create that efficiency, but keeping the governance and the. In keeping the governance as minimal as possible and maximizing the immutability of that layer so that at least the transaction between two parties can be said to be. I hate using the word safe, but. Safe.
- [23:52]
A
Okay, well, I wanted to ask you to elaborate on, you know, what your objections were to Tom's piece because I definitely saw like, you know, so Tom walked us through how he calculated it for his essay and I saw you felt like he was double counting some things. So explain, you know, what your views are on how he calculated it and how you would do it differently.
- [24:18]
D
Yeah, what I, what I tried to do, I picked up his. And I don't want to dwell on this too much because I, I don't think necessarily that the framework is inherently wrong or anything. One of the issues that I had with Jautical was in calculating the so on the lot given default and the probability of default, it seemed to me that he had taken elements from other risk premia and computed them in the first term. And if you strip those out, it, it so happened that you actually came across the, the rate for the high yield for our high yield vaults. Like in my framework, when I say prime, I mean steakhouse prime. I'm. I'm sorry to shill on your podcast, Laura, but like when I say prime I mean steakhouse prime. When I say high yield, I mean steakhouse high yield. I don't know what anybody else is doing. And yeah, when I stripped out these two terms, it seemed to me that, oh, I guess this is what the marginal lender is actually pricing high yield. I don't disagree at all with the way that this might be priced for a pooled lending market or X, Y, Z eth or what have you. On which I agree there is far too like there are so many hours in the day, I can't, you know, I just can't.
- [25:27]
A
Yeah, I'm guessing that you don't allow rse or you didn't for. For backing or.
- [25:34]
D
We had a, in one marginal market we maybe had some dust on Unichain which we exited very quickly. Nothing else. The biggest risk for us was honestly mostly the contagion from AAVE because so many collateral issuers were using AAVE as a source of funding, the liquidity drying up. We were reasonably confident that the solvency issues were quite constrained because we mostly focus on stablecoins and the AAVE issue was like kelp ETH to eth. But there was always the risk of cross collateral contagion or the liquidity freeze putting pressure on a, on a redemption mechanism for an issuer. Like we saw a bunch of issuers see a lot of outflows and that much pressure on like the liquid asset, if it's locked up in a pending resolution of the KELT thing can create a solvency issue on collaterals downstream.
- [26:32]
C
Yeah.
- [26:32]
D
For us the stress was more around that because other collateral issuers had exposure to aave, which. So yeah, Tom is completely right in the sense that the composability of DEFI creates a lot of these connected services which you have to constantly be on your guard on.
- [26:50]
A
Okay, so you don't object to the basic structure of his equation. It's more like just on the margins. There were a few things you were quibbling with.
- [27:02]
D
Yeah, I mean my biggest objection was the resolution. So we strongly felt that there was more like it would be more beneficial to try and have a little bit more resolution to what exactly you are trying to price that it's not necessarily this. And on this I maybe have the biggest disagreement with Tom that I don't think you can actually say that there is a defi yield. Like there are different types of defi yield, just like there are different asset classes in tradified. So it's very difficult to just like paint the whole thing with one brush. When you have like bitcoin over collateralized and you have like high yield and then you have everything that's not steakhouse, which I completely agree should have a huge risk premium. Like for sure.
- [27:45]
C
Yeah. I think this speaks a little bit to the evolution of the asset class.
- [27:49]
D
Right.
- [27:49]
C
If you just think about like high yield bonds in the 80s they were materially mispriced for a decade plus and that's how Mike Milk, Milken and all these others made a ton of money. Mortgage backed securities obviously clearly mispriced going into 2008. And now we're at this new asset class that has a history of maybe a decade if that. And we're trying to price individual risk premium for really short data series. So there's going to be large percentage point differences. I think it's just important to understand where the broader risk premia sit. And then the debate and discussion should be how do we reduce each one of those risk premia across the board? So Adrian's done a great job in saying, okay, here is maybe the highest high quality defi that you have in this pool. But I do think it's still important to agree on what the broader subset of the market should have as a baseline yield. And you have this in traditional finance. You have, okay, the corporate bond, let's call it ETF or mutual fund or whatever it is that has a blended rate. Then you have the aggregate, the U.S. aggregate index, which is actually the biggest proxy for bonds in the US and actually contains a bunch of different credit risk profiles. It contains a third Treasuries, a third corporate bonds, and then a third asset backed securities. So that is the rate I'm thinking of here, is if we were to take a general blended yield in bond markets, that's the US aggregate index in DeFi. Maybe we need to get something a little closer to that for folks to understand as they're putting incremental dollars in, like, how does this compare to what I would be doing off chain or even elsewhere on chain. And if we don't have like that general risk profile, it's really hard to make underwriting decisions across the board.
- [29:27]
A
All right, so in a moment, we're going to talk about some of the other ways in which calculating risk in defi is different from TradFi. But first we're going to take a quick word from the sponsors who make the show possible.
- [29:38]
B
If you gave me 50 right now, the first thing I'm buying is a pair of espadrilles. Coinbase One member month starts with 20% off your first year of Coinbase One, plus a $50 Bitcoin bonus when you spend $100 with a new Coinbase One card in your first 30 days. It's one month of more more rewards and prizes all month long. Coinbase One is the ultimate membership to make the most of your money. And I know because I'm a happy Coinbase One member. It gives you zero trading fees on thousands of crypto assets, 3.5% APY on USDC boosted staking and lending rewards, and up to 4% Bitcoin back with the Coinbase One card. If you trade crypto regularly, the basic annual membership can pay for itself, enjoy bigger rewards, exclusive drops and experiences that money can't buy, all with Coinbase One. Sign up now to get a 20% discount on the annual plan. And so you're locked in for the weekly rewards drops Starting on May 4th, visit coinbase.com Unchained to get 20% off the first year of your annual plan today. Offers valid until May 31st terms apply. Coinbase OneCard is offered through Coinbase Inc. And Cardless Inc. Cards issued by First Electronic Bank. Bitcoin back rates are based on cardholders assets on Coinbase. Bitcoin changed how Money Works Citraya changes how Bitcoin scales Citraya uses Bitcoin as both the settlement and and data availability layer as Bitcoin's application layer. Citraya gives you access to the First Trust minimized BTC on a fully programmable platform and a native stablecoin for Bitcoin. You can now participate in Bitcoin capital markets with lending privacy payments, Bitcoin yield trading and predictions. You get expanded Bitcoin utility without sacrificing its security. Citraya Mainnet is live put your BTC to work at XYZ/ Unchained Etherfy is giving Unchained listeners 15% cash back on rideshares, groceries and restaurants right now, which honestly is kind of wild for a card like this. On top of that I'm getting 3% cash back on every single transaction using my actual crypto. No conversion fees, no nonsense. My bank never once did that. And it goes beyond just spending. You can borrow against your holdings at 4% or less, which is super useful if you don't want to sell your assets. You can also earn on all major assets up to 8% APY just by holding and moving money is just easy. No hidden transfer fees, no friction. It just works globally.
- [32:21]
A
If you want to check it out,
- [32:22]
B
go to Ether Fi unchanged to claim your offer. That's Ether Fi Unchained.
- [32:31]
A
Back to my conversation with Tom and Adrian. So as I mentioned earlier, you know Luca Prosperi had written longer piece as well on how to calculate this risk and he identified a few other issues that are relevant in Defi that are not relevant in tradfi. He talked about how in tradfi in the equation you would assume that the debt matures at a specific point in the future. He talked about how in defi with a collateralized debt obligation you would see that the value of the collateral is continuously being observed so that that point of of time is like not not something that's really existing really in crypto. He talked about how in tradfi you would assume there's this sort of like continuous path in asset prices, but obviously with crypto prices that have limited liquidity they can violate that assumption. And he noted that In Defi, you have to account like, oh, sorry, this is actually repeating the point, but, but anyway, so, so basically he, he's just talking about like different, you know, distinctions that exist in defi. Oh, we talked about how in defi there's the act of rebalancing, et cetera. So you know, I wanted to like just have you expand a little bit on how those kind of factors should be accounted for. The one last thing I'll say is just, you know, Luca basically came to the same conclusion that Tom did where he said, you know, even on blue chip crypto collateral like eth.usdc, he felt lenders should be demanding like 250 to 450 basis point spreads over the risk free rate. And then he said, quote, over collateralized lending on liquid crypto. Native collateral is morpho's bread and butter and represents the vast majority of its TVL. However, we observe consistently narrower spreads by 5 to 10x over the risk free rate versus what rational lenders would require based on appropriate mathematical specifications. So I was curious, like, for your thoughts, if you agreed or disagreed,
- [34:47]
D
I can go first. I had the biggest issue with luca. He's a friend. Yeah. But this is where I disagreed with him the most,
- [34:56]
C
where he.
- [34:59]
D
So you brought up a few interesting points first, like what Luca fundamentally did was try and model. Everybody's trying to come up with a model of here's what I, here's like the spherical cow, right. Like we're going to model just this perfect thing and then try and see how it matches reality and adapt the model afterwards. And his view was that defi lending or repo lending in general was kind of like a put option on the underlying collateral. And he used the pricing models from that world, from that universe to try and price it. And where the biggest methodology, methodological error that I had with it was he assumed that there was a loss to the lender every time there was a liquidation. Fundamentally where in reality it's an incentive balancing mechanism in the way that Morpher works, where if you have a liquidation, you have to reward the liquidator so that you don't have to depend on anyone and so that the incentives between all participants allow the lender to stay whole. And yes, the borrower gets wiped out, but you know, they took the risk, so that's kind of their fault. The other aspect was, yes, for us, liquidity is same block. Like this is the other major difference and change in perspective coming from traditional finance where Treasuries are liquid because they are T +0 or T +1. For us, instant liquidity means same block. No money market fund to date has like absolute same block liquidity for now. We're hopefully going to be able to do something about this in the near future. Until then, there will be a fundamental inefficiency between moving between SOFR and DEFI because you don't have that transmissibility. And so from our perspective actually bitcoin and ETH over collateralized lending is more liquid. Even the newest Treasuries in DEFI just because of this. And then the, so that's really the biggest. If you, if you do away with this methodological disagreement, you realize that actually bitcoin and ETH over collateralized lending in variable rate repo markets like Morpho, when they're isolated, are very liquid and very safe. Historically there are risks. So the biggest risk is if the price of Bitcoin and the price of ETH were to gap in the same block more than the haircut, the lenders would experience a loss. For that to happen, the Oracle system would have to fail or there would have to be an actual price, single block price drop of that magnitude. But the more that Bitcoin and ETH get adopted as an asset, as an institutional asset in ETFs, what have you, the lower the volatility is and the less likely this outcome becomes. So for from our perspective, this prime lending was very much out of, out of range from what LUCA was arguing.
- [37:49]
A
Oh great. Tom.
- [37:51]
C
Yeah, no, I, I, Luca and Adrian both know a lot more about the nuances there than I do. I, I would just point out that over collateralization is not a long term solution in my mind to any sort of credit markets. So you could assign a higher level of risk premia there as and when we move to at least something on par or maybe even under collateralized in the future.
- [38:14]
A
Okay, well I did also want to ask like when we were talking about how Kelp Dow didn't even start with lending, like how do you guys think about that aspect of it? Like I, like I guess. Tom, in your equation would that be sort of that initial, I forget what you called it in your equation, but would that be the initial risk? Or like how do you guys even quantify that? Because you know, just like it almost feels like, you know, there's sort of the universe of lending itself which you should over time be able to come up with models for. Especially you know, as Adrian talked about, just like each of the specifics, like you know, different types of vaults would have different equations, whatever, but Then yeah, when you end up in this situation where there can be something. Yeah. That that just sort of comes in another direction like, like or, or is that something that is easily quantifiable? I. I don't really know. So.
- [39:18]
C
And this, this relates to the point you mentioned about more traditional risks that don't present themselves in traditional finance that do present themselves in defi. I mean there's no, there's minimal social layer, like yes, I could upload something on a USB device in my Morgan Stanley office and take it home and maybe it gets hacked or whatever. But the instances of that are few and far between. I mean there's a very clear attack vector on a social layer in defi that needs to be priced and that presents itself in kelp dao. And then there's the, let's call it rehypothecation or exotic asset risk that is a distinct and separate risk in most of traditional finance and usually contained in a certain sense. And maybe it wasn't in 2008, but large majority of the time it is, is not contained traditionally in defi. Right. Adrian was just saying he didn't really have exposure to this incident and still had some effects, knock on effects. Because defi is so highly integrated, particularly with these exotic assets that have hundreds of millions of dollars. Right. You have people coming out of the woodwork saying Ruben heard of kilt out whatever 300 million, $200 million. So those two risks are distinctly different and something we should certainly assign another level of risk premia to. Then there are other things like you know, stablecoin DPEG risk that I accounted for that is call it minor today. But USD has, USDC has debegged in the past. Obviously things like Terra Luna have materially. Deepak, there are other risks there that are, that are present, that are not in traditional finance that we have to just add on to this risk free rate. And then more broadly as we kind of think about this, this baseline level of rate that we have today in defi, is it a true rate? Is it a battle tested rate? And I would argue no. I mean we're trying to disaggregate the little pieces on the edge and folks will say okay, but the market rate is the market rate. It doesn't really matter if you think it's 12 or 13%, it's actually 7 or 8. To which I would say okay, we have, you know, a few billion dollars trading against these amounts, you know, daily in turnover. Right. You know, you guys atrium have what, 2, 3 billion, something like that I mean total defi is like 20 billion. The battle tested rates in traditional finance trade trillions daily, literally trillions. So you know, there are a number of reasons to be on chain that can compress these yields. So I, I would argue once you get institutional players who actually come in and net new capital, it's demanding a
- [41:40]
D
level of
- [41:42]
C
sort of light on each one of these individual risk premia that the rates would materially variate higher.
- [41:49]
A
Yeah, so actually let's talk about that because I think that was like one of the main criticisms I saw to your piece. So for instance, Dan Robinson of Paradigm responded, quote, I agree that defi lenders seem to be underpricing risk. Most blue chip yields are lower than the risk free rate you can earn from money market funds. Even for lenders who don't have access to treasury yields, it surprises me that they take those risks to earn roughly 3% on stable coins. Then he said, but that doesn't mean rates should be higher. You have to take demand into account. I saw Bill Barheit had like some other ones which we can get to. But you know, I just wondered like, you know, you, you didn't really seem to factor that demand bit into your equation. So how do you think about that?
- [42:37]
C
There's a real convenience yield for having assets on chain and that could be if you're a crypto native fund or defi power user or whatever and you want the optionality to keep your money liquid on chain available for opportunities. Absolutely. There's the censorship resistant angle. Right. If you live in Iran or something and you want access to stable coins or something, it's much more convenient to have that than trying to find t bill rates somewhere else. There's a real tax advantage for most people. Right. There's a lot of profits that have been made on chain that folks are less likely to off board depending on their tax jurisdiction. I mean, there are a number of individual reasons why you would structurally look for a yield on chain that could satisfy some level of demand up and above the straight risk free rate you had by moving it to your US bank account and buying a money market fund or something like that. So I think those are totally legitimate arguments, but they would actually make the actual assumed rate higher if we were to make all us equal.
- [43:35]
A
Adrian, what about you?
- [43:37]
D
But that's why I think the, I think the universe, the universe of current defi users has a convenience yield aspect to it. But I don't think it's the whole universe of crypto native users. I mean one of the biggest use cases on Morho is borrowing against Bitcoin for users within the exchange. None of these users are presumably in Iran and lending to those same users on a DEFI venue on base. Again, also none of these users are in Iran because the settlement layer is quicker, cheaper, more convenient and has the benefit of speed, finality and et cetera, et cetera, et cetera. And when those markets can clear and meet each other without the need for intermediation, that would press those margins higher. So an interesting benchmark that actually I didn't pick up on, and I don't think anyone else in the discourse picked up on, was the simplest comparison you could make of how efficiently DEFI is pricing these yields is how does the rates compare for Defi Bitcoin over collateralized lending to off chain institutional Bitcoin over collateralized lending. There's a significant difference and it's not because DEFI is on chain that it's special, or there's a convenience yield that people are willing to sacrifice and not lend to a user in custody. But variable rate crypto, for all its, you know, volatility and instant liquidation and so forth, is more liquid. And rather than being. This is one thing that I didn't pick up on when I was having a chat with Luca actually where he was suggesting, yes, there's supply and demand imbalance and that could explain part of the difference. But actually something that people are discounting is that rather than there being a liquidity premium demanded for lending against over collateralized Bitcoin, there could well be a liquidity discount by virtue of the fact that these markets are so big and so liquid that you can actually lend at 4% for 12 seconds and come out for hundreds of millions of dollars. And that's very difficult to replicate in traditional finance. So that liquidity again has like that, that flip side where it can actually explain why the rate is so much closer to sofr. So like if, if you can get DEFI coinbase Bitcoin, borrow at four and a half percent or whatever, and the comparison then the comparison is can you borrow from a market maker? And what, what is that rate? And it tends to be significantly higher because there's intermediation and margin extraction and rent extraction essentially. And that's one thing that DEFI does well in compressing that layer away. And that's our sort of view for the direction of travel of this space. And that's how you can get to more efficiency. And in the future, like, yeah, we're in this space because we want to make finance open and transparent. We really believe that this is a better way of doing finance. For all of its defaults and curiosities Tom's, you know, loss given default rate, I think it was like one and a half or 2%. It's not that bad. Sorry. Compared to traditional finance it stacks up decently well as a whole. And it's not like traditional finance is absent of Infinite Mint issues either. You know I'm in Switzerland at the moment. We used to have two global GCIP banks. Now we have one because it was, it faced very much a similar like the Archegos blow up that destroyed Cuddy Suisse was very much an Infinite Mint hack that the hedge fund manager had unlocked because he found that he could Infinite mint his assets by just not telling other people that he was doing it. And the Khalisus were the last to exit and you know, other banks were quicker and faced no losses. But it's no different. And now Switzerland has one g sib instead of two, you know, to potentially far more catastrophic or potentially far more catastrophic consequences. And in a system that if, if it had been mediated on defi could have been resolved more transparently and perhaps quicker.
- [48:02]
A
Tom, I just noticed in the chat you wanted to talk about steakhouse financials risk model.
- [48:07]
D
Yeah, sorry I had asked the.
- [48:09]
A
But that was new.
- [48:11]
D
Yeah, because it was relating to an earlier point to show that we had very much the same risk dimensions as Tom. We just arranged in different places. Yeah, but if it's not relevant to the point we don't have to bring it up. Yeah. So this is on our docs page, we have another page which we update once a day and it's very much, you know, it'll see, it'll look very familiar because it's very much like what's the source of social risk to the asset? What's the technical risk, linear risk, the market, the liquidity and so forth. And this is kind of we don't come up with the risk premium per se, but it's how we underwrite an asset to qualify it as prime or high yield stakeholders prime or stakehouse high yield.
- [48:52]
A
Okay. There was one other response to Tom's piece I wanted to talk about which was Bill Barheit. So he, you know, like the others felt like the market was, was deciding what the, what the yield should be. Some of the things that he disputed were he thought that 10 year treasury was not the right risk free base. He again felt you were double counting some of the, you know, possible, you know, failing Failing possibilities. And then he also thought that you were sort of like annualizing some of the, what he called six weeks of panic into a permanent risk rate. I don't know if you, you know, had any thoughts on, on his response or, you know, if you wanted to.
- [49:45]
C
Yeah, I'll do them quickly because I feel like I've hit on them a little bit here. So the risk free rate, that's fine if you want to use a SOFR overnight. Right. It makes sense. I think there's some, the 10 year yield captures some time value, some credit risk components that are not inherent to the overnight rate that I think are actually present in a lot of defi. So I think it's fair to use the 10 year rate and that's the traditional risk free rate of finance. So if you want to use sohere, it totally makes sense. You just got to take 50 to 65 basis points off my yield. Totally reasonable assumption there. I don't think I've double counted anything. I think I outlined that, but I'm going to release the full detail on all of the different risk premium after that because I didn't do that do a good enough job there. But I don't think I'm double counting there. If anything, I think I'm probably under counting some of the, under some of the other components that I mentioned. And then Sarah was the third point he brought up.
- [50:35]
A
Oh. He said he felt you were annualizing six weeks of panic into a permanent risk rate.
- [50:43]
D
Yeah.
- [50:44]
C
So I think that's short sighted in that when you have an event or a month like we've had now, you have to materially change the risk profile. And I just think back to my TradFi days when we were underwriting private credit and was like, okay, give me your history of what you've done for the last 10 to 15 years of underwriting private credit. Okay. You have a 99% rate of non default, so 1% default rate. Okay. Your recovery given default is about 95%. Fantastic. You've done that for like 10 or 15 years. Here's the mix of your loan composition and then all of a sudden you have something like Blue Owl happen where it's like, oh shoot, you had most of these AS software and SaaS based loans and all of a sudden the default rate ticks way up. You have to re underwrite your models based on that. You can't just say, oh, that was, you know, we can't just go back to the old model because this thing happened. So you have to reinterate what the risk profile is when you have 600 millions of dollars of DeFi hacks in one month. You have to rewrite your models. Now do you have to say, you know, the higher yield for this one particular month is completely what you should use going forward? No, but you have to assume that some level of that was actually real because it did happen and you have to price it into your models. And this is a fundamental challenge across cryptos because our time series are so short for all assets, for all risk profiles, for volatility across the board that every single negative data point just makes the, the default rates and, and sort of the overall risk profile that much higher for traditional investors who are looking at time series data over a number of years and decades.
- [52:12]
D
Yeah, I completely agree with Tom. Sorry to interrupt, Lola. I completely agree with Tom on this actually, like for us it's a reflection of like we would describe it as something like the cost of capital for going on chain spikes whenever these events take place. And for good reason, because it is nerve wracking and because these losses are sort of very public, very transparent and catastrophic and large. And yes, it would be unreasonable not to have a marginal lender re underwrite their expectation for the cost of capital for going in.
- [52:47]
A
Yeah, fundamental disagreement there. Yeah, I'm going to add like a few that Luca pointed out or. Sorry, sorry, this was, this was. These were some of his explanations for why he felt that for instance, morpho depositors weren't demanding more in terms of, you know, yield. Some of the hypotheses were depositor misperception, as in they didn't understand the risk they were taking. Survivorship bias because some of the losses only hit specific vaults. Bull market masking, token subsidies. There's one other one that I really thought was interesting. He called it regulatory arbitrage. He talked about how basically the Clarity act, since it's likely going to forbid stablecoin issuers from passing along yield. He said that that results in a total pass through risk transfer onto depositors, which I thought was really interesting. If there's any congress people who are listening to that, I hope you paid attention to his point there. Okay, so I think we kind of covered why it is that maybe the market is actually at a lower rate than some people think. But I did also want to then just go to RWAs because this is something that we're seeing. Like there's an uptick in interest here. But Luca called this out and we have another show, Bits and beps where some people have Also been calling this out basically, you know, with, with RWAs use it being used as collateral. Luke is saying that when there's volatility in the price there, it's not observable because that's only updated, you know, like quarterly or whatever the cadence is. And then basically like the Oracle updates. But the collateral price can be out of sync. One is just updating a lot faster than the other. And then he also talked about how liquidation just can't be immediate in that kind of situation. It can even involve like a legal process. It takes months or you know, whatever. So you know, he's very bearish on non crypto native collateralized lending. And I was curious to hear your opinion on that.
- [55:03]
C
Depends on the collateral. If you're looking at like private market or real estate stuff that is marked quarterly and is very long process involves legality to actually liquidate as real trouble. And I think that's a legitimate argument from the folks who may not like in Congress why you want to put this stuff in your 401ks. I mean long term investment vehicle. But if you ever needed to take a loan against it or try to liquidate it in some circumstances, it is not something that you're going to unwind pretty quickly. Now if it's gold or oil or something you can easily trade or hedge, I would disagree. I think those are very well established and very easy to write collateral against.
- [55:38]
D
Yeah, echo Tom completely. It depends. And you know, there are some, you know, assets that are not easily fungible, not easily easily measurable, update their price twice a year. They make for very difficult matches with on chain variable rate repo. With instant liquidations, it's a very dangerous mix. There's usually a substantial duration mismatch that you can only patch at best. Doesn't mean that you can't eventually do something like on chain receivables financing as long as for example the entire chain is running on stablecoins. Because then presumably the secondary market for these individual receivables is much more liquid and you could update your models more frequently than once a month. You could do it as every time there was a trade or swap or whatever. But yeah, something like gold tokenized gold works fine. You know, in tokenized stocks they're still sort of the weekend bridge, which is a question mark for variable rate repo because onchain repo on variable rate works throughout the weekend but obviously the stock markets don't. So there is a more pronounced gap risk there. It's not impossible. And there is a substantial margin that could be disintermediated with DEFI infrastructure, provided it's coming back to sort of the earlier points that I was trying to make more aligned with Nikolai's philosophy of make things more like Uniswap, make them harder, make them more difficult, stupider, with less governance, so that they can just do fewer things better. And that's actually the value proposition of DeFi.
- [57:19]
A
Okay, so last question. So one of the responses to Tom's post was from Michael Igorov of Curve, and he said, so this is what it's saying. If you supplied in random stable coins blindly to a random protocol on defi blindly, your breakeven risk adjusted rate is 12.55%. And then if you start being selective, these risk premiums go lower. So this is sort of where we've been trending in this conversation. There's sort of like the broad brush stroke way of defining this. But then when you dig down into the details, it could be like specific vault, specific types of collateral, like whatever it is. So how do you think about that? Which protocols or types of protocols or ways that defi lending can be structured? What are some of the ones that you think about as ratcheting the risk up or down? And then how would you tweak your equations or models to account for that?
- [58:28]
C
Until a few months ago, I would have said lendiness is a clear one. But then you have something like Balancer and he just kind of goes out the window. Right. Like, you know, you would think that we have some level of assurance that these risk profiles are reducing, but in fact they're increasing with AI. So I think it's really important for folks to critique what I said and look at more and more risk premia and then individually try to assign some level of risk. So if you're working with steakhouse, I'm very sure a lot of the risk premia I associated with are materially lower than I set for the broader market because they're curating for you. And I think in the short term that's probably the answer is to look at curated vaults with clear collateral backers that have a track record in history. And unfortunately, I know a lot of people in DEFI won't like to hear this, but we're effectively having a bailout fund for this recent hack and socializing some of the losses on the other end. And we're sort of stepping into this permissionless financial system. And you could argue that's what curators are doing as well by self selecting these assets. So I think it's more of like the progressive decentralization route that folks would say, I think we're Tom, aren't we here after this many years? I'd say probably not, because, you know, you've seen a lot of these hacks and folks want to share since they're getting their money back if there's. There's an issue. So I think it's more like defi on guardrails for the time being until we can effectively really reduce a lot of the risk profile across the board here.
- [60:01]
D
That's what the, to me, that's what the DEFI is, guardrails. So properly designed, Defi is nothing but a set of guardrails. You know, Uniswap and Morpho Blue, like the Singleton Smart contract don't do anything except constrain what the users can do with it. That's kind of their strength. That's really where the future of the ecosystem lies, in our view. And you know, I, I want to make finance open and transparent. I don't want to be underwriting boot tokens for the rest of my life. You know, like, the point of this is like, I do want to eventually have a meaningful impact. We're starting to get there. You know, rent, rent, minimization, disintermediation, these are real benefits that are occurring now to users of Defi, but they lean on the things that are being constrained more. So a bad outcome in my view would be guardrail, meaning the law has to come in and step in and say DEFI is only defi if it comes from this region of France. The best outcome is like, the best outcome is something like you have small primitives, you have Uniswap, you have morpho. They're extremely constraining. They provide for a set of rules that can't be broken because they're run on a decentralized network that can't be censored. And then on top of that, people will do what they may, but at, at a minimum, the base level is really, really hard. The. The. A bad outcome is something like, yeah, like a KELP exploit or drift, which is where you have like these mixes of intech mixed with Defi with all of the vulnerabilities that that ensues. Obviously that's much more risky and much more difficult to price.
- [61:54]
C
Yeah, Steakhouse, the champion of Defi, I think is a great slogan for you guys.
- [61:59]
D
But yeah, this whole castle will burn down and we will still be inside saying people will be the last out the door.
- [62:08]
C
I think we probably need, we agree on a lot here. I think we probably need different definitions for these things. Kind of like we're trying to write the definition of MEV again because it doesn't really make sense in its current context. The way you say it, if you say defi more broadly, I guess it's like saying credit. But what you really mean is, okay, you have Treasuries investment grade and then high yield and then just like random crap on the side here. And it's all blended together right now. But really we need to segment these things and borrowers should look at it in that light because when you have the CEO of CalPERS coming out and saying we can't touch defi because it has all these crazy risks, it's like they're obviously not referring to you guys or these curated yields, they're referring to the complex more broadly. So as we continue to grow, I think it's just more nuanced. We got to add in the conversation.
- [62:51]
A
Yeah, yeah, Honestly, I envision a future where, I mean, it's just such a basic thing to say, but the transparency of it all is really going to actually make this financial system stronger in my opinion than the traditional financial system. So yeah, people will be able to see kind of more clearly what the risks are. But yeah, we need more like experience with it all to even calculate it
- [63:20]
D
so well, to give you in. One of the big themes of the 2008 credit crisis and the global financial. And the beginning of the global financial crisis was defined by a complete evacuation of trust in the system where nobody dared face any other counterparty because they had no idea what they were exposed to and to what degree. And one of the main liquidity outflows was just people pulling into their idle markets. Use the defi vault term. Like all of these banks pulling into idle markets and waiting to see who is going to die. Whereas even in the heat of the contagion around AAVE and KELP and all this, I mean, all things considered, it has resolved. It has been able to resolve reasonably quickly given the size of and prominence of this exploit relative to the size of its space. Like, you know, Archegos exploded over months and NODI still probably don't know what the exposure was. But even in the case where you had a question about specific issuer and their exposure to aave, you can. Oh well, let me just check. And I think that goes a very, very long way in restoring trust in the system and bringing it back into bonds absent the absence of the, you know, Defi United and all this which is great as well.
- [64:40]
A
Yeah, actually, but just hearing what you said, it made me realize that the way things are headed, we will likely end up with more privacy in defi. So then would the system rely more like on CK proofs to. To. For people to have comfort or like. Yeah. How do you feel like that part's going to work?
- [65:06]
D
I don't. I don't know the first thing about ZK proofs. Like, I would qualify it as. Is it, am I depending on a social guarantee or am I depending on a crypto guarantee? Like, if it's something that's hard and I can check it by myself with my own node and. And so forth, then I have reasonable assurances that that's there because I don't have to worry and do the calculations of whether Ethereum is censoring me. I don't like, that's kind of the point. If I have to depend on the social guarantee, then that looks far more like a fintech or a Tri product than it does like a crypto product. And that will require significantly higher like, I. For the love of God, I hope collateral issuers are not going to start privacy shielding their asset reserves. Like, if anything, I felt we were going in the opposite direction after. If somebody wants to shield their stablecoin transfers between their friends, then. Okay.
- [65:59]
A
All right, well, we'll have to see how that all goes. You guys, this has been great. We're a little bit over time, but thank you both so much for doing the show.
- [66:08]
D
Thanks, Laura. And next, Tom.
- [66:11]
C
Thanks, Adrian.
- [66:12]
A
And for those of you who watch, thanks for joining us and we will be back next week. Catch you later, everyone. It.