David Choi & Conor Moore: USD.AI - Financing the Future of AI Infra

Join Jordan Yeakley as he hosts David Choi and Conor Moore from USD.AI to discuss their innovative synthetic dollar backed by AI and DeepEnd hardware assets. Learn about USD.AI's unique approach to on-chain infrastructure financing.

USD.AI: https://usd.ai

🎯 Key Highlights

▸ David and Conor's journey from Deutsche Bank to creating USD.AI

▸ USD.AI's three-pillar structure: Caliber, modular underwriting, and QEV design

▸ How USD.AI enables debt financing for DeepEnd hardware operators

▸ The unique redemption queue mechanism that maintains USD.AI stability

▸ Multiple lines of defense protecting depositors from defaults

▸ How leveraging hardware assets can dramatically increase yields

▸ USD.AI's potential to reduce token emission rates across DeepEnd networks

💡 Subscribe for more crypto & DeFi insights! 🔔

🧠 Follow the Alpha

▸ USD.AI's Twitter: @USDai_Official

▸ David's Twitter: @0xZergs

▸ Conor's Twitter: @_ConorMoore

🔗 Connect with Delphi

🌐 Portal: https://delphidigital.io/

🐦 Twitter: https://twitter.com/delphi_digital

💼 LinkedIn: https://www.linkedin.com/company/delphi-digital

🎧 Listen on

Spotify: https://open.spotify.com/show/62PR1RigLG2YN5Pelq6UY9?si=18ac7ccf36ab4753

Apple Podcasts: https://podcasts.apple.com/us/podcast/the-delphi-podcast/id1438148082

Youtube: https://www.youtube.com/channel/UC9Yy99ZlQIX9-PdG_xHj43Q

Timestamps

00:00 - Intro: David Choi & Conor Moore from USD.AI

01:00 - Team background and journey to USD.AI

03:00 - USD.AI's origin and the DeepEnd hardware financing problem

06:00 - USD.AI's three-pillar structure overview

07:00 - Caliber: Tokenizing hardware assets with on-chain property rights

12:00 - Modular underwriting system with first-loss positions

15:00 - QEV design: Creating an arbitrage venue for redemptions

20:00 - Summary of USD.AI's value proposition

21:00 - Default protection and amortization process

24:00 - Value proposition for borrowers and DeepEnd networks

28:00 - How USD.AI benefits neo-cloud operators

31:00 - Phased approach to scaling USD.AI

34:00 - USD.AI's place in the broader crypto ecosystem

43:00 - Next steps and upcoming launch details

Disclaimer

This podcast is strictly informational and educational and is not investment advice or a solicitation to buy or sell any tokens or securities or to make any financial decisions. Do not trade or invest in any project, tokens, or securities based upon this podcast episode. The host and members at Delphi Ventures may personally own tokens or art that are mentioned on the podcast. Our current show features paid sponsorships which may be featured at the start, middle, and/or the end of the episode. These sponsorships are for informational purposes only and are not a solicitation to use any product, service or token.

Transcript

SPEAKER_00 0:01

You're now plugged into the Delphi podcast. Hey everyone, welcome back to the podcast. I'm your host, Jordan from Delphi Research. I have David Choi and Connor Moore here from USD AI. How are you guys? Great. Doing well. Doing great. USD AI is a synthetic dollar backed by AI and Deep In hardware assets. We released a report last week on the portal that's a great primer. It's definitely it's free right now, so definitely go check that out. Connor and David are here to provide some additional color on the vision for USD AI and how it impacts infrastructure finance. How about we start with some quick intros and background?

SPEAKER_02 0:38

For sure. I guess I'll go over the team real quick and then pass off to Connor for just kind of our um earlier journey. Connor and I, we both met at Deutsche Bank, um kind of coming from TratFi. Um I think it's also the pattern in all of uh Delphi as well, um kind of escaping TratFi. Um but we worked we worked together on real estate investment banking. Um I I personally moved into like the early MEV scene um back in like 2017, 2018, um where it was like the earlier tax surfaces before it was even called MEV, like um gas bombing for ICOs or like LP sniping for like Uniswap. And did a lot of venture over the years. Um and uh yeah, I guess my focus on the team is more on the DeFi and the redemption design um with this like TraFi plus MEV experience. Um Connor's background um is after Deutsche, he he he also focused on real estate private equity and like structured credit. Um so all the way from tenant management to like asset asset distributed asset management to uh fanymate style debt um aggregation, uh more focusing on the physical assets, the cash flows modeling, and like the credit structuring um for the lending primitives that we're inventing for USD AI. And we have a third co-founder as well with Don Alcal. Um he's actually kind of like the perfect um perfect background for USD AI in that he used to work at 21.co uh back in like 2013, 2014, right when like Bitcoin mining was moving away from the GPUs towards ASICs. Um and you saw the first like firsthand industrialization of the crypto mining infra and how like early debt was introduced um to mining um for Bitcoin mining specifically. Um then you worked at uh DRW working on FPGAs for HFT. Um and uh um we we also did MEV together. So kind of he focused on the smart contract system design and like hardware level risk um on the team. Um and yeah, we we all came together um uh in like late 2021 and early 2022 originally to focus on um low liquidity assets, but um Connor will explain that.

SPEAKER_01 2:30

Yeah, I can give some more kind of background on on our path. So when we started the company, we were really looking at what was happening in DeFi, particularly Ave, Uniswap, Compound, as sort of these zero-to-one moments in financial innovation where you actually had you know tremendous improvement, 10x, 100x improvement in in execution, but it was always for you know AMMs for swapping liquid tokens, um uh CDPs and um uh lending markets for barring against liquid tokens. It was all kind of catered to liquid tokens. And we thought there would be sort of a next wave of financial innovation happening on crypto rails, particularly for illiquid assets. So think capital markets, debt capital markets, you know, equipment financing, um, financing businesses, the things that we had experience with in our past life in traditional finance. And so we started the company originally, it was called MetaStreet, started in 2021. Um, basically, over the last three years, have done about $400 million of loan volume. But we've always struggled to kind of find the right uh collateral pair for what we're doing. And what you've seen in the RWA space is there's a lot of adverse selection problems where, you know, for example, if someone's tokenizing real estate and trying to get a loan on-chain, you know, you typically would ask, like, why are they not getting a loan from the most robust debt market in the world, which is like real estate debt? Um, and so, you know, we've really struggled to find like the right collateral pair. And I think in the last kind of six to 12 months, what we've realized is D-Pin networks and specifically the hardware that underpins these networks is a really natural fit for on-chain financing for illiquid assets. And so we started this uh separate vehicle called Tactical Compute, which is uh basically an AI mining network in partnership with Aether and Beam. And the whole idea there was to basically raise capital and buy GPUs or get access to GPUs to earn yields from these decentralized AI networks. And in that process, what we realized was even though the cash flows are quite good and we were super well capitalized, it was basically impossible for us to finance that vehicle because you're earning cash flows on-chain. So you you're precluded from what a bank would be able to do or private credit fund, because fundamentally, when you're doing things on-chain, you really need an on-chain capital source to pair with what your use is. Um, and that was sort of the the inception point for USD AI. You know, the previous few years of doing structured credit products for illiquid assets paired with kind of this new vertical within D Pin, specifically for AI, but also for telecom, you know, D pin assets or energy um networks, where all of them are really capex intensive. And at the same time, you can't actually finance these assets because you're earning cash flows on chain. Um, and so that was basically what led us into wanting to build USD AI, which fundamentally you can think of USD AI as it's a yield-bearing synthetic dollar. Um, and the the yield is coming from hardware assets that are being collateralized into loans. So GPUs, telecom, energy, you know, these are CapEx intensive asset types. The operators of those assets can basically get a loan from USDAI and pay interest. And that interest is what goes to paying yield out to the USDAI token holders. And those USDAI tokens can then be you know proliferated through the rest of DeFi, borrowed against, swapped, etc. Um, and so it's really a crypto native structured credit product for illiquid asset types within kind of the these deep pin network ecosystems.

SPEAKER_00 6:36

So this is DAI meets internet capital markets, and this is this is real capital, productive capital, not you know uh crypto native capital markets. Um can you help me establish the market for this a little bit? What D pin sectors, what what uh what sectors is this applicable to, and what's the the total size that um of this market?

SPEAKER_02 6:57

So I guess we can start from like all RWAs, like crypto native RWAs, and just really focus on deep in SS sector. And that heuristic is just really finding that green field of like, hey, good credit, but like um kind of under underbanked in a sense. Um and among that, even um you've seen this, you've seen this behavior among uh deep in protocols where they're focusing on scale. Um, we call it like a bit of a deep deep in trilemma where they either balance out scale, price, or uh data fidelity, or just like quality of of output within the within the asset itself, whether they're providing compute or they're providing energy. Um, a lot of them are trying to lower the barrier of entry from the incremental user by lowering the price, which naturally lowers the sacrifices quality. Um so the the the areas that we really focus on are um sectors that have expensive hardware, um, things that are actually valuable, like you this is a there's a good reason why LLMs are not powered by um consumer grade hardware. It's it's run by H100s and things like that, because um it's it's designed to be productive, it's not designed to be cheap. Um, and that that kind of heuristic is exactly what that is really good for, because it will make good cash flows in the future, but um it will definitely have higher barriers of entry. Um, and so we really underwrite the hardware itself and like what it's actually valuable for. Um, we don't necessarily underwrite the deep in network. That certainly helps in terms of bolstering the cash flows, but we we really look at the hardware and hardware quality because the hardware is mostly semi-fungible in a sense. So once you underwrite one, um there are some nuances between maybe one uh H100 and a data center in Texas versus one in Europe, but generally they're pretty similar. Um so we probably look at things that are above that 10k per asset range um to start off, because it's just the most scalable, and it's also where the most need is in terms of um this capex story where it is cash intensive. These these protocol these protocols or these um miners want to grow, they can't get financing. And these deep in networks are just inflating their token um all the way down because they're trying to compensate for that initial growth hurdle of like um participants putting up that much money in order to um get their money back, which they call the payback period when it comes when it comes to the concept of mining.

SPEAKER_00 9:17

Interesting. Yeah, one of the one of the coolest parts of USD AI, I think, is is the unique problem set that you guys had when when building this. So this hardware doesn't mark to market. There's a lot of unique um issues with the game theory and the default process that you have to solve for. So you guys sort of created this Oracle-free DeFi structure from the ground up and from a product perspective, not from a philosophical need, right? Um so can we can we go over like USDAI structure at high level and and what makes up USD AI?

SPEAKER_02 9:47

Yeah, um, I can go over the like very high levels and the details we can kind of split up. Um, but there's really three pillars that kind of like establishes that. Um the you know, Oracle is lending is the base structure, um, but that's really supported by these three categories that makes it all possible. Um, because you don't have an Oracle like a single source of truth based on like a very you know marked market, which is like how frequently something's marked in the market, right? Um we have three mechanisms. Um one, and they can be kind of separated as the entry, um, the risk scaling, and then the exit as like the um the allegory here. And the first one is we call caliber, is like uh it's an acronym for like how we tokenize assets in a way that I think is superior to many other methods because we actually deal with a physical asset, something you can pick up and like put away, um, and not no SPVs and no like shrouds of uh of uncertainty. Secondly, we have uh this modular underwriting system, um, which um we can go into in a minute, but that really focuses on how to scale up the various underwriting processes of different hardware um globally in the world that actually powers these AI networks and what we give that physical loans to. And lastly, we have the uh the QEV design, which is uh an acronym for Q extractable value. This is inspired by you know Flashboss's um you know MEV boost, um, which kind of mimics um this Q that is invented. Um it's we kind of view like these liquidity um uh liquidity blocks as like uh block production in a sense, where people are competing in the blind market to exit the redemption. Um that way um this is a little more game theoretical in how you exit the pool. Um but yeah, those are the three main pillars which can which we can go into. Um but uh I'll let Connor take the first one.

SPEAKER_01 11:28

Yeah, so let me talk a little bit about caliber because this is a a really kind of critical component for this tall work. So what what we did here is we basically, in order to execute a loan, the loan is happening on-chain. And in order to execute a loan, we have to have an initial legal contract that's basically tokenizing assets. Um, and so what caliber stands for is collateralized asset ledger with insurance, bailment, evaluation, and redemption. But basically, all we're doing is in order to execute a loan, the borrower has to sell the assets to a subsidiary of Permian Labs, which is our Deller C Corp, in order to get their loan. And the reason we do that is because this allows us to get a master insurance policy over all of the assets, because the assets are actually technically legally owned by us. And then once they've sold the assets to us, they sign a sec a separate contract that's a bailment that basically gives them the rights to operate and profit from the assets. And in that bailment contract, what we do is we say the bailout, who's the the rightful owner of this, is actually this holder of this NFT. And so there's no loan agreement that's ever happening off-chain. The only off-chain component is what we're doing with this tokenization and subsequent bailment contract. But then what this does is it gives us the ability to put property rights completely on-chain. So that NFT is called an electronic document of title and it fits under UCC Article 7. And it basically is just a way for people to transact and have an electronically native deed to an asset. So once that that exists, then the borrower can post that deed as collateral and borrow from the pool. From there on, every point of property rights exists solely on chain. And the reason this is really important to us is because when you think about the RWA space, it's a bit hand wavy in terms of what tokenization really means. And what I mean by that is you'll see a lot of products that have some underlying loan agreement that then gets rolled up into a fund structure. And then that fund is then selling some portion of its fund into some tokenized entity to then sell on-chain. The point is, like your property rights do not exist on-chain, they exist off-chain, and then hopefully these various intermediaries that you're interacting with are going to honor that. And we don't really like that. So, what we're doing here instead is putting 100% of the property rights on-chain. And then when you think about what's happening ultimately, is you're you're putting this, you know, tokenized um yield product out into the world. But we want the users, the buyers of that token to be able to drill down into every single loan, and you'll be able to see exactly what's happening at all times and see that that is completely on-chain and you have complete, you know, actual property rights in that case. Um the the next part to mention, so that's basically sort of this legal construct that's underpinning everything we're doing here, which is called caliber. The other part to um to kind of like really dig into is what the the actual underwriting process looks like, because USDAI is going to be supporting you know everything from neo clouds and GPUs all the way to telecom assets, power assets, EV chargers, so any of these deep end networks that have sort of this hardware infrastructure component to it. And so obviously, it would be you know a lie to you know, basically present ourselves as experts in every one of these possible categories. I was just talking to guys yesterday that are doing weather stations. Like obviously, I'm not an expert in weather stations. So, what we're basically doing, and this kind of builds on the original Meta struct uh Meta Street pool structure, is for every primary origination market, there's two tranches. The first tranche is the first loss position, and that's basically 5% of the capital stack. And then there's the second tranche, which is what is actually going into the USDAI backings. So that first tranche is the first loss, and that the capital that goes into that first loss position is also the underwriter and originator of those loans. So that you have some alignment of interest in that that first loss position gets some outsized return in in exchange for you know originating these deals and underwriting them and putting up that first loss capital, but they also bear first loss. If there's defaults, they're the first ones that would lose their money. So that position is basically held by you know experts in the specific subject matter that we're looking at. So, you know, there's a first loss underwriter for the GPU space, there's first loss for EV chargers, et cetera, et cetera. Um, and then David probably should talk about QEV because that was really his his baby.

SPEAKER_02 16:25

For sure. Um, and QEV was kind of an analysis that was made from past attempts in terms of uh a lot of these blue chip RWA protocols that claim to have good collateral, but oftentimes and do end up depegging. Um, because even blue chip collateral can become pretty toxic if it's not managed correctly with the maturity schedules. Um and QEV, I guess in summary, is uh is really an arbitrage venue where participants can actually profit um from DPEGs. Um and that's kind of like the main heuristic in terms of how we design the entire um redemption queue. Um oftentimes redemption queues are first come, first serve, which is whatever available liquidity there is, um, it's whoever like redeems first, or it's like some whitelist system, which is a little more opaque. Um for us, it's uh it's we turned it into a market in a sense, where every 30 because these assets that we deal with have extremely high amortization, um, that's very that's more of a feature than it is a bug. Um, because what that means is that lower the the loans amortize as well um at a at equal or faster speed, we prefer faster. Um and that means that every single month um there is a large cash up um from the principal uh repayments and as well as the depreciation, which mimics the depreciation of the hardware. Um so I guess it's best to give an example. Um if, for example, this loan is amortized as about 5% a month, um, because uh because that's just the way that the loans are structured, the straight lines of zero, say after like 24 months or 18 to 24 months, um then what happens is that the liquidity that's cashed up again from that loan principal, um decreasing every month with increasing cash every month, um participants can then enter into an auction that happens every 30 days. Because every loan payment happens every 30 days, what you do is aggregate all these depreciation repayments um and you put it up for an auction. Uh we call these uh QEV boost auctions, which is effectively every 30 days is a synchronous auction where you can put in a blind bid that nobody knows what your amount is, um, whether it's what and you can bid like you can bid like one BIP, two BIPs, um, and whoever has the highest bid gets the highest parada amount of liquidity into that redemption. So this is again a very big arbitrage venue where a participant can buy, say, SUSD AI, which is our stakes version, with like a long uh redemption queue, like um this 30 days specifically, um, and then enter into this queue um even though they bought it at a cheaper price and they've redeemed it at par. Um and what we're really trying to uh mimic is just understanding the perspective of effectively an arbitrageur, uh stablecoin market maker, or uh any participant, and buying these assets in the in the secondary, but um making a profit in the primary. Um and that's why we made the redemption queue into a market where it is a little more game theoretical. The question is what's the Nash in terms of how much a participant should bid in order to uh get the most amount of uh you know liquidity um for the position. Now, if you don't bid at all, all that you know bidding amounts in terms of the this uh redemption queue is then just given back to all the participants in the in the market. So if you do nothing, you just get supplemental yield, um profiting off this like traffic attacks. Um and the blind nature of it is what helps amplify and stabilize the entire um entire stability of the synthetic dollar. Um yeah, so that's kind of how QAV works um at a high level. We have more details uh obviously in your report and uh in our docs as well.

SPEAKER_00 19:57

Nice. So to summarize this, USDAI is a vehicle for on-chain users to provide liquidity to D Pin operators who are essentially looping. And Caliber and the the rest of your process puts as much of this on-chain as possible, and as as we've seen with other sort of real-world asset projects. Um, there's a modular underwriting process that allows specialized actors to bring into the USDAI umbrella various um hardware types from various like deep deep end sectors. Um, and the yield for USDAI is the interest paid on these loans from the deep end loopers. Is that correct?

SPEAKER_02 20:39

Yeah, exactly. Um and we're really trying to fit in, we're what we're really trying to do is like take a capital markets or I know we call it ICM nowadays, but uh capital markets is assets that are actually productive and configuring it so that it becomes um fittable into the money markets that we have, because you can't take an NFT and put an Ave. You can't take a piece of hardware and put an Ave. So this is kind of this uh fungifying process, I guess.

SPEAKER_00 21:03

Nice. So um we we spoke a little bit about the game theory and like the default process. Can we double-click on that a little bit as far as okay, the the equity cushion and the amortization process and like how for for people who you know may be concerned about defaults and losses for LPs, like how can we do a quick overview of that process and and how we prevent um bad actors?

SPEAKER_01 21:27

Yeah, so we basically think of everything in lines of defense. So you have you know the the LTVs on the loans are less than 100%. So you have um, you know, 70%, basically 50 to 70 LTVs day one. That means that you have a 30%, you know, 50 to 30% equity cushion. And that equity cushion is really the first loss, right? If a borrower defaults, the first person that's gonna lose money is the borrower because they put up that equity cushion. The next line of defense is the first loss position. So the um the underwriter is gonna then take the next you know portion of loss, and and then finally you have sort of the the actual exposure of the protocol. And so what happens is every 30 days, let's say the loans are three years long. So every 30 days, the loans are gonna amortize by about 3%. And so that means that you know you basically are using these loans to finance acquisition of new hardware. And so you're buying, you know, day one, you have a 30% cushion. Over the subsequent three years, those assets are presumably going to depreciate by some amount. But because your loan is being paid down every month, you're continuing to maintain this cushion. And so even if the borrower ended up defaulting, you should have a profitable default scenario because the uh loan can basically the rights to the collateral can basically be sold at a discount to what the fair value is. And so what happens is every 30 days, these borrowers have to make these payments. If they miss a payment, then that deed, that electronic document of title, is basically auctioned on-chain. And there are a number of resellers, whether it's for GPUs or other equipment, that basically are part of our network that would come in and buy those assets. And so those folks then have this opportunity where I get to buy this GPU that typically costs $100 for $70, and I can resell it for $95, right? Or whatever those margins end up becoming. And so that's basically the whole flow. So you have sort of multiple lines of defense, and then you also have on the depositor side this ability to basically exit pre loan maturity because the borrowers have secondary liquidity, or sorry, the lenders have secondary liquidity for these tokens. And so you have kind of the benefit of what looks like you know a publicly traded token, um, but has access to these high yields where you can still get out, you know, whenever you want based on secondary liquidity and and these uh redemption proceeds.

SPEAKER_00 24:18

Got it. Okay. So how about the value proposition for borrowers? Can we touch on that a little bit? A lot of that's a source of confusion, I think, for a lot of these types of projects. So how much more profitable is it for a borrower to lever up on the D pin operation than um like a vanilla non-levered position? Yeah, they're gonna be paying interest rates that are quite high, right? We're targeting what 15 to 20 percent um interest on this.

SPEAKER_01 24:48

Yeah, so it's it's honestly it's a super easy math equation where you just say, as long as your unlevered yield, the yield that you're making without debt, is in excess of your interest payment, that's always going to be accretive, meaning it's always going to increase your overall returns. So if you have a 15% cost of capital and you can get a 70% LTV loan, uh, that basically means you can add about two and a half turns of leverage to your returns. So if you were making 30 unlevered, you're gonna end up making close to 70 or 80 with leverage. Um, and the main benefit is really not only for the borrowers, but also for the deep end networks themselves, because the deep end networks are solving for a cost of capital. And right now, in order to attract hardware operators to that network, they have to give them these crazy yields. And those yields are inflationary. And as David was talking about, the you know, end result is every single network token just gets punished because you're giving out so much an incentive and your network hasn't really stabilized yet. And so if you can access leverage, let's say that the market clearing rate for these operators is you know 80%. Well, if you are if you have no leverage, then the deep end network has to pay out 80% APRs and tokens. But if you can access a debt market, then you only have to pay out a effectively a 30% APR. So you the deep end network ends up reducing its emission rate by over 50% just because you have access to this debt market. Um, the other thing that I think is really interesting as you kind of look not only to deep in related hardware, but just general neo clouds and hardware operators, what we've seen is um, you know, there's probably about 200 neo clouds. Neo clouds are just companies that buy GPUs, put them in a data center, and rent them to AI startups. So there's about 200 of those, and all of them have raised raised somewhere between five on the low end to a couple hundred million dollars worth of venture capital money. They're basically raising the most expensive cost of capital to then go and buy depreciating assets. Makes no sense whatsoever. But they don't have access to debt otherwise. Um, the credit market off-chain does not really exist for this space, barring a couple of really large players like Core Weave. So, you know, in that case, you also have a huge benefit, right? Because you shouldn't be raising venture money to buy depreciating assets. You should be building out your software platform, hiring people, you know, doing high return on cost activities. And so, you know, not even within the defense space, but also just general GPU as a service platforms are also like huge customers of ours because ultimately they can access credit. And then for them, it's really just a scale game. The the number one neo cloud will end up being the biggest buy assets, and that's just all it takes for them to win or lose. So if you have access to debt markets, then that means you can buy more assets, which means you can bully out the other guys or acquire them or you know, do whatever to consolidate power. Um, so you know, I I think there's huge benefits, obviously, to borrowers to have access to debt markets because it just means that you can be more efficient with your dollars.

SPEAKER_00 28:11

Got it. Yeah. Um can we touch on um the phased approach to scaling USD AI um over the coming months and and years? Yeah.

SPEAKER_02 28:24

Um so initially, I guess uh really the first few days of our into our launch will be mostly in treasuries just to reduce the trash uh cash drag. So whenever somebody deposits uh or mints USD AI with a USDC or USDT, um they're minting treasuries as kind of the like the base layer. Um and that's again, whenever there's still capital, we want participants to at least earn the minimum yield. Um and then that's being then staked. And that's once it's staked, that's when it's viable for actual infrastructure loans. Um and our initial target's gonna be 40% utilization during this initial um uh initial period where we're kind of um testing out um a few of the parameters in terms of how the protocol works. Uh our target utilization will likely be 80%. Um we don't necessarily have like a peg stability module like any money market would, where it's like an like automatically adjusting to a specific uh target utilization. So again, this is uh this is kind of learning in terms of how the market responds. Um but right now we we will keep it at 40% utilization, um, unless there's some flexibility and some obvious changes that could be made at the protocol level. Um but that's probably our uh our like stage two um when it gets to that maturity state. Um and once uh QAV uh launches with that as well. Now if there is a governance token, which obviously we're currently structuring um and where the protocol is actually decentralized, um uh that's probably the next step after that, where these parameters can be controlled by decentralized governance. Um a lot of our protocols already are very well decentralized. Um technically, um, anybody can launch the protocol permissionlessly in terms of like different pools, they won't have access to this um like large um capital vehicle, but they can launch their own like individual borrowing pools um permissionlessly. Um so in terms of uh the steps after that, um, we will start decentralizing the income generated as well as the protocol management, um, in terms uh which is typical for most projects in this space. Um as for the underwriting and the modularity in terms of other uh collateral types that we onboard, um, that's also going to be a stepwise process as we develop um these various uh sectors. The three main sectors that we're focusing on the beginning um is going to be um compute, uh telecom, and energy, um, where um the underwriting scales with um obviously more participation and more demand needs. And I think a lot of those use cases kind of emerge once you do have the money as well, or deposited and being utilized.

SPEAKER_00 30:56

Got it. So out of the gate, it'll be highly liquid, but the yield will be lower. And then as it becomes um increasingly composed of hardware, the yield will increase and it will become less liquid. And that's where the QEB comes in and potentially a governance token. Um can you share any thoughts that you guys have had as far as how you would implement a governance token and and what that would look like as far as utility and stuff like that?

SPEAKER_02 31:21

Yeah. And I just want to highlight the yields that correspond to the stages that I described about a minute ago to your point. Um it depends really on the staking ratio because USDEI doesn't generate any yield if you hold it, but you you're you're in points under points program. We call it cores. Um so that doesn't make any yield, but you get um obviously um the our our points program. And then the staked version, if it's like a 50% ratio, then it's just two X treasuries. Um at a 40% diversified into like uh hardware loans, which you know range from 15 to 30, um, predominantly in the higher teens, um, then the average weighting gets closer to like you know double the double digits, like mid teens. Um and at maturity be something closer to say like uh 20%. Um and that's uh that's all contingent again on like uh multiple variables, but and the type of loans that are issued. Um but generally all the loans that we do are have a hurdle rate of at least 15%. Um and that's um very competitive even in Web2 like our TradFi lending markets. Um as for the governance token, it is uh it the goal is that it does earn fees for maintaining the protocol. Um, for example, there'll be like an interest clip um all the way to um an insurance fund structure as well, um, which is still kind of being thought out um in terms of how that uh how those how that yield aggregates. Um but the more interesting part, besides just you know, um uh fee uh profit profit sharing is uh is how it kind of participates in terms of the governance of liquidity toward sectors that it wants to help grow. Um it's a little curve-esque and curve inspired, um, but um you know kind of gating like in terms of how you underwrite these assets all the way to feed participation and risk tolerance. Um, I think there'll be a point in time where a lot of the best hardware has to be financed by that. Like, can you imagine a solar panel industry or data center industry without that? It's just nearly impossible to envision. Um, and where deep in protocols, if they want to scale rapidly, once they start proving out the initial hardware designs and that it's actually industrializable, much like Bitcoin when it moved away from GPUs toward like ASICs, you know, like even though it was more niche, um, it was far more productive, even though it's far more expensive. You saw a lot, a lot of um um the growth story being shared with uh traditional finance, and people really believe in that CapEx story. Um, and that becomes a really sellable kind of uh um well scale scalable form of credit. Um and and that and that's kind of what we want to imitate in that um in that same story as Bitcoin, where um you saw this massive industrialization of the hardware sector, and that really stabilized the the Bitcoin price because now all the miners were now just holding Bitcoin because they had access to credit. Um so what we hope is that a lot of these different participants, um you know, miners all the way from um BitTensor all the way to um say like Dawn Internet um and other like uh other emerging platforms, especially in this DAI wave, um ranging from you know Prime Intellect to like news and these other ones, which really do depend on like high quality um compute, um they will not only want access to compute, but also credit, um, which helped scale um the wide availability of um of these resources on-chain. Um so uh in terms of how it routes, um, I think that's a pretty interesting use case. Um, but of course, there's the base level, which is at least the fees that's generated from the origination all the way to the inter sharing.

SPEAKER_00 34:48

Yeah, absolutely. Um the the yield component of this is is very intriguing. I'm I'm curious how you see yourselves fitting into the broader crypto landscape. Like, do you see this um being more of a like intuitively categorized as a D pin project or a D F DeFi project? I mean, you guys, I believe, are um launching on Arbitrum, so that gives you sort of the yield and the DeFi angle, and then like a lot of the Solana ecosystem is this D-pin stuff. So, like, what's your vision for how this fits in to the the broader crypto ecosystem?

SPEAKER_02 35:22

We do we do all the dirty work in the in the physical space to try to fit in the DeFi space. And I think generally, I guess the best patterns I've seen in DeFi that you know, that the the most recent big success being Athena is really having vertical control over not only the origination, but all the way to the distribution. So origination, structuring, and the distribution itself. You also saw this with with uh with Lido and why kind of that was like the market leader by a significant margin. So because you had the P2P validator for the struct for the origination and structuring all the way to the distribution of you know um uh Steeth and Rapstick D in terms of the wide ecosystem. Um and much like how we thought LSTFi was different from uh DeFi, everything's just DeFi at the end of the day. So uh we we do view ourselves as a DeFi protocol. Um now where we get the source of yield might lean a little more RWAs, might lean a little more deepen, but uh um a lot of the um our objective is to distribute the yield that we generate to the broader ecosystem. And I we're trying to make the rates a little more competitive. Now there is a bit of a sacrifice where there's a longer redemption period versus what we what we're normally used to. Um but I think the DeFi ecosystem is maturing a little bit. Well, we we already saw like the seven-day unlock periods for the vault, so as well as with Athena. And so participants getting a little more comfortable with a little longer durations of redemption. Um, and this is also why we deal with specifically hardware, and we don't do any form business loans because a business loan is truly a you know a long, a long-term loan, and you can't really um redeem any value like during that period of a significant value. Um, so all the integrations that we have in DeFi, um, we we are trying to be like the best rates on Fendle on the PT side, all the way to making the looping a little more sensible because there is a hurdle rate for DeFi um that you know will move move up and down. But for us, we are pretty close to fixed income because our loans are a little longer term, but at the same time um uh um far more liquid than many other designs. And I guess this is also kind of like the um the way we view DeFi is that people usually say DeFi money markets in a single phrase. Um and I think that's kind of a flawed thinking, thinking that money markets is the end goal is really just a the first step into a much broader ecosystem. Um, because money markets traditionally move at that it it's capital that moves at the speed of money, which means that uh the yields almost always compressed to you know the risk-free rate um generally, because everything gets arbitraged very, very quickly. And you see this um if you're a T Fey farmer, like everything just gets a single digit super fast. That's because we're giving loans against stuff that we call internet capital markets, but it's not capital. Um money focuses on liquidity, it focuses on um uh how much uh is traded and how much volume, whereas capital does not focus on that at all. Um, capital focuses on things that are actually productive, businesses all the way to real estate to say hardware, um, equipment, things that are focused on productivity rather than liquidity. And oftentimes they're very, very illiquid, um, where the daily trading volume is significantly lower than this total uh total total asset like representation, um, in terms of you know, thousands of times bigger. Um and that's like a nice ratio to kind of measure is this really a capital asset or a money market? And I I think oftentimes when you look at RWAs, people try, and that's why a lot of people are leaning toward RBAs because they're looking for this exogenous forms of yield to bring into DeFi to loop and to do interesting stuff with. Um but to Connor's point earlier, you're getting a lot of adverse selection when it comes to RWAs to come to crypto. Um, whereas the best RWAs in crypto that people have really liked is actually when TradFi buys an RWA from crypto rather than the other way around, where crypto's buying an RWA from TradFi, if that makes sense. Um and the two best or three best RWAs, quote unquote, has been a T-Bills, which is a great use case, steel capital and like trading platforms. So people just get passive yield. Um that's what that's just seems fairly obvious. Bitcoin mining. Um it's almost like $25 to $50 billion is kind of hard to really measure. Like it's a lot of debt that TriFi absolutely loves. And like to the point where even MicroStrategy kind of took this idea of convertible debt, I made a which was really popularized by Bitcoin mining. Um, and people love that CapEx story. And you know, Athena was really cool too because it was a yield that TriFi has never even heard of, right? This idea of uh Delta Neutra perp balancing because of uh of this new um trading platform and inefficiency um where preps didn't really exist uh outside of crypto. Um and the this new category of equipment financing kind of follows the same idea of that Bitcoin mining story, whereas the high capex industry, you're mining BitTensor or you're mining something that's on-chain, all the way to um being familiar with that uh that CapEx story. And I think it's something that TraffI would actually want to buy along with crypto natives.

SPEAKER_01 40:25

And I think like David and I are both very much like uh decentralization maxis and believe that decentralized networks scale much better um and much more quickly than centralized players do. And so, like when you think about what success looks like here, you have sort of the one of the best use cases of crypto, bar none, is capital formation at rapid speed, right? We've seen this many times over. So you have the ability to have capital formation rapidly with just one click deposit and mint a token, right, on the depositor side. And then you have this modular underwriting. So you have a variety of underwriters and originators that can plug into this network and specialize in what their subject matter expertise is. The outcome here is tremendous scale. I mean, much more scale than anything that we've seen in crypto, except for maybe tokenized T bills. Um, you know, Core Weave did a $7 billion loan with Blackstone. To give you a sense, uh Core Weave now public, but before they were publicly traded, they did this loan at SOFR plus 11. So basically 15-ish percent cost of capital for $7 billion. So the scale in this game is so large. Um, you know, there are very few D5 protocols that ever get over $7 billion, and that was a single deal. And so, you know, I think what success could look like here is really you have the benefit of rapid capital formation plugging into a monstrous TAM of assets that also can't really get, you know, great financing options. Um, you know, we we even like have been talking to some family offices who are really interested in buying GPUs and financing those acquisitions through USDAI. And the interesting thing is the main thing they care about is the depreciation benefit because they just want to take tax write-offs on these big bulky assets. So it's its own financial product right there, just creating a depreciation product for people to offset, you know, capital gains and income. Um, but yeah, I think like the scale is just really important to highlight because it's so much bigger than what we've seen in DeFi. And so for us to be, you know, getting to billions of dollars of uh deposits would still be a tiny fraction of what the kind of TAM potential is here.

SPEAKER_00 42:49

Yeah, so so the big idea here is that similarly to Athena and some of these other um, you know, major DeFi protocols that have come out over the years, it's it's you we get this fundamental new source of yield that is less cyclical than a lot of these other sources. And then on the other side of it, you have a much better way to scale um capital in the real world with this, with these deep, deep end projects. Um so it hits both hits both sides. That's really interesting. Um, I'm curious. So, what are what are the immediate next steps for USCAI? What are um when are you guys launching? I I think I heard something about a private beta from the X account. Um, can you give me a a little bit of a picture of the next next couple weeks and months and and how you guys are scaling up to launch?

SPEAKER_02 43:37

Yeah, absolutely. Um over the last uh few months, um, we've just kind of been building the borrower pipeline. Um and you know, even though it's been months that these people have been looking for borrow, they they're still looking for borrow because a solution just like fundamentally does not exist. Um we're finding use cases that um we realize we can be the salt the solution for, um, that just kind of emerge the movement. You get um your hands dirty in the space um all the way down, even still in the GPU market, which you would assume is pretty efficient, but is so extremely inefficient. Um, not even for Web3. D pin might be the wedge, but even many Web2 use cases, um like major major brands, major companies, um, and pretty very good credit in terms of uh it's a very greenfield of credit. There's adverse selection is not really something you think about uh because you just get really good performing counterparties that just don't have that systemic um access. Um is is systemically underbanked. Um and frankly, it's because the industry is so new. It's um the concept of using GPUs as like I would argue the most currently the most com important commodity in the world. And like people hear more about like how uh NVIDIA chips are distributed around the world than oil, um, especially given the AI race that happens today. Um so as probably the most important commodity in the world, um it's uh the demand for it has been extremely high. Um so the the the demand side in terms of like the yield generation, like how do we generate the yield, that's been building up over time. We're still obviously open to chatting with participants on the borrower side who are looking for higher um higher yield generation through the use of that. Now, as for the you know, the day five participants, um, I'm guessing majority of your listeners on on Delphi. Um, we are um currently um on ramp uh in process of launching that private beta. So our contracts haven't deployed um now just finished the audit, but now doing a few more steps in terms of making sure the front end everything is uh is uh um up to our standard. Um and that should be launching uh if everything goes right next week. Um and this this program we have a course program, uh which is our points program for participants who want to generate this to access this new form of yield. Um and we are trying to make it very DeFi friendly. So we do want to have the integration with Pendle, um having all the oracles up, all the uh money markets, um, especially the modular ones, um, so yield organizations and uh many of the vaults in participation for people who want to either farm the yield itself or they want to farm the points, um, or both in a sense. Um and that timeline in terms of pub launch will probably be closer to the end of May. Uh yeah, I guess people can keep their ears out and kind of join our wait list. We have a telegram group and we have a Twitter so they can follow follow us there. Um but yeah, the launch is very soon and uh hopefully they'll see it on the recommended farming list on their favorite uh yield farming platforms.

SPEAKER_00 46:27

Yeah, I can't wait to see how this um how CT handles all the new yield opportunities with these uh DeFi integrations. That's really cool. Uh Connor, David, thank you guys so much for coming on the show. Um, this was great. Very um, very interesting stuff. Thank you guys.

unknown 46:44

Thanks, Dorothy.