David Choi, co-founder of Permian Labs, the builder behind USD.AI. Welcome. Thank you. Thank you for having me Peter. So let's begin with Permian Labs. You guys raised some money from some VCs. Maybe share with us who invested in Permian and why do you think they did? Yeah. So Permian Labs is the technology company that is behind the protocol, is usd.ai. And it was originally called Metastreet in the beginning. And the thesis around then, when we first raised was looking at things beyond the money markets of DeFi. And that was kind of like the priority for us. It's like most of the assets in the world starting in this was really uh started with the NFT like and game file like Meta. where a lot of things just didn't fit into Aave. They didn't fit into Uniswap because most things aren't coins and they don't prioritize liquidity. They prioritize productivity. And that's kind of divide between money markets and capital markets because most capital in the world, the very common example is like a tractor. A tractor doesn't fit into Uniswap ah no matter how many times you fractionalize it. ah And that was kind of like the objective of Permian Labs was how do you rebuild DeFi to fit other illiquid, less liquid, um non-fungible, so oh perfectly divisible assets uh into accessing greater financial markets. And that was why a few of the backers behind us were um Dragonfly, DCG, uh Nascent, Delphi, and a few others, and Alliance, and Big Brain, Solana Ecosystem. But ah just generally, it was a big question to solve. really takes a long while to develop that solution. Yeah, let's get into it. um So Permian Labs is the development company behind USD.AI. Maybe let's get into that transition into like, tell us about USD.AI, what it is, and maybe if you could explain it to kind of the crypto native person, or actually the, maybe how do you explain USD.AI to your parents? And then maybe, and then explain it to the crypto native person. If I were to explain it to my grandma, um after bingo, would explain it as ah it's effectively a way to make AI more scale faster. And AI, as we all know, even grandmas know that it's Nvidia, right? It's usually the GPUs, it's the infra, but it's uh not just uh GPUs, it's also telecom. It's also just all the infrastructure that's needed to make this AI work. And what we effectively do is we help scale that operation in a decentralized way so that financing and liquidity and the ability to acquire those assets and operate them is far easier. um And that's done usually through debt. Pretty much every single data center to solar panels to m energy to telecom and generally all assets and infrastructure stack, they're all financed. um But for some reason, in a very similar category called D-PEN, know, if I have to word it out, know, the decentralized physical infrastructure networks, they don't have any debt penetration, which was a bit of an alarm to me because I think a lot of the participants don't understand that they can get debt financing. But that's how you scale these networks to be 10 times their current size, because, you know, if you can get a 90 % down, that is 10 times the amount of purchasing power they can... get access to. So that's kind what USD.AI is. It's kind of a combination of DeFi and Deepin, where Deepin operators can get access to financing to help them scale their operations. And Deepin protocols are able to outperform because, well, their metrics are higher and, they can get uh lower cost of entry per participant and is less dilutive to their token supply. Yeah, I think that's what kind of a lot of people kind of miss with, you know, some of these depend projects or, you know, decentralized physical infrastructure projects is that they're capital intensive and getting access to capital is absolutely key. You know, I've done some work in the real estate space and, you know, debt financing is like part and parcel of how you do real estate, right? It's only way you can do it, in fact. really is, especially if you want to grow. so these decentralized physical infrastructure projects, it's getting access to capital and credit is key. Tell us more about that and how usd.ai kind of enables that. For sure. Just kind of going off that example of real estate, like in order to get a loan, usually you have to have a tenant issue where a house has to be occupied by participant in order for it to get financed. It's actually an individual financing that's really looking at making sure that the debt can't be paid down. If you're giving a loan against an office, you have to make sure that there's tenants in the office. And it's really just tenant management. And that's the core basis of a lot of the... or financing that happens. That's why people didn't really like WeWork because there are startups and they weren't like law firms that just occupied the location and just provided the ability to repay that debt. And the best comparable I can really give is like I call it the largest energy deep in protocol called Bitcoin. That was in a sense, very much a deep in kind of product. Like it really was the... CPUs to GPUs and then ASICs and around 2013-2014, you actually had the first debt issued against Bitcoin miners. In the US, the most popular one, more well-known one was 21.co. uh That was, know, biology and little you. RCTU actually used to work there as well. And you saw a live like what happens if you structure bad debt because it blows up, right? Like they took all this debt against probably the worst hardware asset you can ever have, which is a Bitcoin miner, right? Like think about it, like it's... 18 % 18 months and then uh it's completely like that. It's already expired. The new AM miners come out. It's completely connected to a single token, but for some reason, TradFi loves it, right? Because it's a new kind of yield, new kind of product, a new kind of thing to oh finance and equipment's always like a little better than a business loan because it's really just very simple. Like what is the equipment worth? ah And ended up becoming a $25 to $50 billion of debt industry. It's much bigger in itself. And that actually like 2014 to 2025 now that changed economics of how Bitcoin is mined where the operators now took on debt. They didn't actually have to finance everything with equity with like, you know, straight up cash. And then they had to sell the Bitcoin instantly to cover the cost of the operation. It's kind of like a hidden history that a lot of people miss. But that is the same issue that a lot of deepening operators face today is that A lot of the miners, well, first off, the operators have to really increase their rewards. And then all the operators immediately sell, which is why the economics of deepens kind of flawed, kind of like early Bitcoin. um And um this element of the larger capital markets to help finance an extremely capital intensive industry, literally, I don't know which equipment industry in the world isn't financed, like, frankly, like even, um you know, residential solar panels are 70 % financed all the way to data centers. Some are even over 100 % financed in a sense. um But ah that's what's been missing in Deepin, in my opinion. It's not really Deepin. Deepin to me is the revenue layer, but the bigger stack is the financial layer. And I actually call it InfraFi because it's really focusing on combining the infrastructure with the Deepin, with the DeFi and the Deepin and all those together, which ah kind of creates, I guess, a deepened 2.0 sort of narrative, which focuses more on scale uh without sacrificing quality. Yeah, infra, infraphy. Is that the category that you guys are trying to build right now? Yeah, and improvise on top of deep end. Got it. I think that's pretty fascinating. You're seeing it more and more in crypto, right? Where rather than the MeToo type of protocols, um we're seeing more protocols trying to create their own category. And it makes sense, especially in the case of InfraFi, which is on top of D-Pen. um How are you positioning the USD.AI to, you've got Infra, or I guess, deep in projects that need access to credit and capital. And then is uh there a two-sided market in this space? I guess it's also the financing partners on the other side. Is that correct? Exactly. So we've done loans ourselves against, we've structured the entire borrower and lender with the two-sided marketplace design. And we realized how inefficient it was. It was just so hard to scale. It's kind of like, you're a VC and then for every deal you do, you raise an SBV. Like every deal is underwritten. So we knew this wouldn't scale that well. And we looked at ways... is a special purpose vehicle for the audience. and that's more a metaphor for like, hey, every time you do a deal, it really like, should people just like to pause for each deal, there's a lot of fragmentation, the underwriting is not very scalable and the product itself, like it's tough to get to billion dollars of TVL here. So the question became like, what is a scalable way of uh sharing that yield, sharing that risk with multiple participants and also creating oh a singular vehicle in which we can socialize the capital markets here. And that's that was really through the synthetic dollar stable coin like design. um So I don't like to stable coin because I think that's gonna be a regulated word. In reality, we're very much like a yield coin. So we say yield bearing synthetic dollar as the longer phrase. And this was really pioneered by like kind of Athena, like kind of socializing the basis trade, which anybody could have done, but it's just a much more compact and scalable product if it's a synthetic dollar kind of vehicle where a single In this case, USDAI and staked USDAI can be put on Pendle. So you have PT deep in yields um or PT deep in debt yields and YT as well. And these integrations in DeFi is that somebody has to do the dirty work of taking things that are really non-fungible like hardware, whether it's uh solar panels all the way to H100s and somehow converting it into a simple yield bearing synthetic dollar design. And then after that, it can finally access DeFi. And that's how we figure out how to scale this operation so that it becomes that billion-dollar protocol. got it. How is that done? Maybe walk us through the process. Let's say I'm a deep end project and I need access to capital so can grow. So I can share it from the perspective of a deepened operator and also just share it from the perspective of a mentor, of a USDA mentor. So for the borrower, it's quite simple. uh They effectively enter into a tokenization agreement. So they tokenize the hardware itself. We have a few methods to make sure that it really isolates the hardware from the operator. So we really, we don't give a loan against the operator. We don't give business loans. Because if you're underwriting a business, like if you're giving a loan against the business, you're um really just giving a loan against an SPV, against like an entity. And you just hope that they'd repay in a sense. Whereas if we isolate the hardware away, um we use something called a bailment contract where it's kind of like a sale lease back where they sell us the hardware and then we lease it back to them and they have the right to buy it back. It's a very common structure in real estate as you may know. um That makes it so that it's almost impossible to bankrupt from our side. know, run away with the hardware. It's already insured because of this design is held in, you know, this issuing code, which, you know, holds the hardware itself. um Very inspired by, like, you know, Athena's like off-chain settlement where you try to isolate that risk and really uh lock boxes so that it's hard to break. um And uh this allows us to scale this operation quite fast because we have a single master insurance policy. So there's no theft. There's no... catastrophic damages because that's all insured. And at the same time, it's properly represented on chain, unlike most other RWA protocols, which is just like an IOU. This is truly a one-to-one representation. um So that's step one, which is the hardest part, which is really making sure that one H100 is in fact one H100. And uh once we do that, we kind of use this product we invented called Metastreet, which is a way to take illiquid like really non-fungible assets and uh try to make it more more liquid, which is really done through that structuring. It's a structured finance product where we take all these layers of abstraction, which I can get to across like five experiments we've done. But the main idea is that it's oracle-less, it's modular, it's underwriting, and it's a lending protocol. And then there are tranches of debt that help secure the layers of uh financial certainty on like what your risk is to what your reward. um But it's a we call like underneath MetaStreet we have a few other names like the automatic ranch maker which makes it so that your risk is specific to your um to your reward and multiple participants can come in to participate. um got it. so maybe we lay out the you've named I think a couple of products here that I want to kind of clarify. So MetaStreet, first of all, there's usd.ai. And then how does that relate to MetaStreet? And in premium labs is kind of like the the that's the builder behind all of it. Yeah, so I guess for that this is more a structuring thing. Metastree is like the technology protocol and Premier Labs is the technology code. But the final product that most users and participants will see will be USD.AI, which is the synthetic dollar. got it. Okay. And how do you want the rest of the conversation to go? Do I guess we have a number of things that we could focus on? um Would you like to focus on USD.ai? Yeah, exactly. But at the end of the day, it's all the same kind of product where it was kind of layers of iteration to get to that point where USDA wouldn't have existed without Metastreet and Metastreet wouldn't have existed without Permian Labs. And the core idea is that you have the structure finance layer and then the final output of the structure financing was really as simple of a UI as possible, which is mint and redeem. Yeah. all the complicated steps to modular underwriting all the way to the way we represent underlying assets on chain and then also structure that debt to make it into a pure like very simple vehicle. That is, I guess, more technical details that the audience might be interested in, but something that they don't need to engage in. Maybe we can go back to like, what was the first aha moment behind USD.AI? Like, was there a clear uh kind of insight that you had that, you know, you need to start something, iDavid needs to start something in this space because this is a clear issue that these decentralized infrastructure projects have. Yeah, I guess the aha moment for me was Oracle is lending as kind of like the aha moment for me. There was a little wave in like 2023 where a lot of people were trying to do this Oracle is lending originally to avoid Oracle attacks on DFI protocols. But oftentimes the Oracle is design was more dangerous than something that was just well tested and more robust like chain link Oracle. But the aha moment wasn't that. we should just replace money markets like Aave without an Oracle and then people just choose their own prices. The aha moment for me was that it actually wasn't that you're trying to solve money markets like Aave without an Oracle. It was that you enabled new assets to be lent against with the exclusion of the Oracle. So things like NFTs, things like GameFi, things like Deepin, things like RWAs can suddenly integrate into a protocol like this without And that's only done with an Oracle-less learning protocol. That was kind of like the big aha moment for me, where you have this, I call it object-oriented finance versus something that was more function-based, right? And that's a comp-site comparable. But really focusing on the unique properties of the object instead of shoving an object, like a GPU through uh an obvious protocol, which just wouldn't work at all. Because that's... Oracle-based protocols depend on extremely high liquidity on the spot markets. And that if you want to dump, like you see with the markets recently, like hundreds of millions of dollars, you really couldn't do that within instantaneously. um It's just like a very, well, illiquid market. This often happened with uh NFT protocols, where people try to put like JPEGs onto an Auvio-like product. um And it blew up because they didn't think about the design, didn't think about, you know. uh, mark to market. didn't think about the real complexities of being Oracle-less. Um, so that was the first aha moment. Uh, the second aha moment was, um, I was actually chatting with this, uh, um, with this helium miner at just some deep end event at a conference. Um, and you know, you go to a conference and, uh, usually you talk to people who are like building an industry. This guy just ran one helium miner, like just, you know, $500 note. And that's all he did in the space. He's just more curious. I was like, it's cool to talk to somebody who's kind of dabbling. And yeah, he took a... I was talking to him like what we doing is like, oh, actually what I did was I bought my healing miner with a chase card and then I borrowed against it on Klarna and he was just arbitraging the rates because was... Because, you on Klarna it was like 0 % for a few months and then it's like up to 20. He's like, even with the 20, like I'm still making like, you know, 40 to 80 % or something. And I'm like, wow, you are... you are actually arbing as a retail user. That's kind like my home. like, well, if this guy's doing it, how come all the other decentralized protocols aren't doing it? So we launched this entity called TacticalCompute.ai, which is like AI mining corp, which just uses the latest NVIDIA chips to mine AI networks like uh Acer and a others. It's like a 40 mil mining corp out of Abu Dhabi. And that was to get our hands dirty. And we just kind of know, raise debt to make this possible. And that was getting our hands dirty moment. I'm really inspired by that one conversation I had with a user. we kind of combined a lot of these experiences together to create the simplest vehicle to scale what I think is probably the most important commodity in humanity today, which is the commodities around AI. uh I think it's as effectively what Petrodollar was in the 70s, where this is why like the US kind of constrains access to AI uh assets such as graphics cards. And the real important part is going to be how you finance and control interest rate of these assets that underpin everything. So that was kind of like the mission being combined with the use case being combined with a technology platform. I think speaking with that Helium miner is really instructive of, when you spend time with users, and users I'm using very generally, you could really come up with some serious insights. That led you to creating a whole new project and uh learning about the process that they went through to buy a Helium node um and being able to arb. you know, do some small arbitrage with that. Like you can, created a whole entire business out of that. Yeah, and that's kind of like the, I like to say the first useful RWA protocol, ah because I don't think a lot of people realize Deepin is RWAs in a sense, mostly because when you have an RWA asset, you're proxying a network exposure. So you don't really think it's suddenly an RWA because it's physical. You think it's still a Deepin asset, and it doesn't even cross your mind that it's RWA. But uh the fact that you're kind of tokenizing the future rather than tokenizing the past, m kind of represents a different value proposition. And the fact that there's a game to play where there's a mining rates versus the debt rates, and you can scale it based on how you control those levers, um this to me is like the new like yield farming in a sense. It's the same thing as yield farming, but a different proxy where it's not just money into a protocol and you farm rewards, it's money into hardware, which is financed through another layer. um And like it just gets more interesting how this ecosystem can evolve very rapidly. Yeah, I agree with you. I've met with a couple of uh real world asset RWA projects recently, and I'm reminded of one that I met with about a month ago. And they're in the business of uh putting a bunch of life insurance claims on chain. And that's kind of a really good example of like the past, know, just whatever's currently on paper, which surprisingly like it's all on paper, putting it on paper or putting it on chain. but it's still looking backwards and it's not revenue generating. Whereas what you guys are doing or decentralized infrastructure, D-PEN projects, it's really revenue generating and it's looking towards the future. And I think that's a really good distinction, but you're right. D-PEN is RWA. I struggle with a lot of RWA use cases myself because I feel like it's like a litmus test of somebody who's been in this space for a while. You know, like you've seen the IBM hyperledgers all the way to like the early RWA attempts. None of, mean, the only real use case, like you can look it up and you can qualify me or not, but like most RWAs are just T-bills, right? T-bills are entirely riskless. And of course that being on chain, like wouldn't have that much issue scaling. But actually looking at risk being underwritten, like why is the RWA, like all the way from real estate to reinsurance, all these traditional industries coming to crypto to get financed? And that is by definition um adverse election. Like they're coming here because the greatest capital markets in the world, for example, real estate, literally the greatest, ah if you can't get financing for your, in what we call the golden age of private. private credit, where bank financing and private credit financing is now massive. It means your assets not worth financing in traditional rails. And this theme about like, hey, there's greater credit access. I'm like, there isn't. There's only like 70 billion of assets in DeFi. It's nothing. That's like one credit fund in the DeFi space. uh It just means that. not good assets. So you either have to deal with something that truly has a market fragmentation issue that crypto for some reason solves, which is not just it being on chain and you tokenizing it. That's like a 5 % improvement. It has to be something new, something that just can't get financed. One category that I think it's worth relooking at, shout out to like Wildcat or um a few other products, the crypto market makers were actually in fact a very good credit market to tackle. It's just that we just had one sour apple, three arrows. and a few others where the underwriting wasn't set in place. But they were like a revenue generating company that could get financing traditionally. I think that's interesting credit. Because you actually have sour apples and good apples, whereas some RWA categories is just pure bad apples, or like 95%. Whereas in something like Deepin, which is crypto native, if you tell me you're mining BitTensor to a bank, they're not going to give you financing at all. But if you go to like a crypto native underwriter who actually understands the risks and surprisingly they do for mining operations even like casuals who buy Bitcoin They kind of get the Bitcoin mining industry for some reason They're a little more comfortable with the risk and they they're excited by the idea of a crypto native RWA Being brought to tratify rather than tritify our double ways being brought to crypto This is why people like the Athena Athena to me is an RWA. It's an CD five, which is off chain being brought to institutionals um through this proxy. um So Athena is also an RWA, just people don't view it like that. Deepin is also an RWA, but people don't view it like that. Bitcoin mining is definitely an RWA, being proxied to TradFi um and almost as big as DeFi, it's like total like TBL, um but we don't really view it as RWA. um You seem to be very comfortable speaking about credit markets and debt and crypto. Tell us about your background. What led you to... What is it about your background that makes you the right person to be building this right now? Yeah, I'll give my background as well as my teams. So my background, my first job was actually doing art lending. So giving loans against artworks like, you know, Frank Stella's, Warhol's, Robert Indiana's and stuff. Then worked at Deutsche Bank uh doing investment making for like, uh real estate investment making for like casinos, shopping malls and stuff like that, which generally, you know, asset heavy industries like that are entirely based on debt. like Deutsche and JPM are quite big. And that specific sectors. And then I joined crypto because they wouldn't let me trade anything besides currencies, not stocks, bonds, or a lot of commodities. like that's low, feel so I'm like, what can I trade these currencies? And I found cryptocurrency because it the most risk on currency they could find. And it was a way to kill time. And that was like 2016. And then that kind of led me to like, what is the game to play in that moment in time? And then, you you could just buy tokens or you could play the ICO game, which led me to MEV because it ended up being like, can get into the ICO first? This was before like venture investing wasn't really a thing yet. And like venture friends really didn't exist at that point in time besides like one or two funds. And that got me to MEV. We didn't call them MEV back then. It was just trying to make money in a sense and by playing with like some launching elements and kind of dabbled in that, did a lot of angel investing over the years as well. And uh after a while, I kind of wanted to be a builder in space after seeing so much vaporware, I guess, in the early days of the ICRO. By really leveraging the technology that we have, uh Uniswap was like a big inspiration to me, like the idea of Oracle is like financial, like primitives being built and wanted to build something like that, but for the capital markets instead of the money markets. That's really cool. The rest of our team like our CTO he's work at 21.co Which was the first I know Bitcoin mining operation that was supposed to be the largest back then in 2014 2013 and he saw it kind of like that Bitcoin mining that blow up happened real time after that He was also working at HFT doing DRW and then Connor my other co-founder. He used to work at Deutsche with me where he was also working on real estate debt and they worked at Rock Point, which it's like one of the larger real estate private equity firms where he kind saw the idea of structuring debt in large scale, real time, and packaging it up and then making it so that small houses can't become $6 billion of debt um if you just have the right structuring. So that's kind of our team background of uh evolving over the years, but really our time building MetaStreet um as a first iteration of this product, which was to deal with um NFT and gain by liquidity, which is actually um how all these other protocols want to tokenize their assets as NFTs. They think it'll offend the DeFi and I'm like, it won't. There's nothing in DeFi that will take ingest in NFT. building on that experience kind of brought us here, which is like, hey, we've operated a hardware before, we're mining networks right now, all the way to we know how to structure debt in creative on-chain ways that pushes the frontier of what you can do on-chain. And now we're collaborating with the deep networks to see if we can evolve not just a way to increase their sales, but maybe even change your tokenomics, where it's not going to be as dilutive anymore. It's going to be, don't finance everything with tokens and token rewards. Finance it with debt, then tokens definitely help with the economics. But at the end of the day, the hurdle rate is the interest rate. That way, they don't really have that much. They can scale at the cost of capital. You said something that I want to kind of double click into. That the current kind of mindset is to, if you want something liquid, know, make, or if you want to make something liquid from an illiquid market, like NFTs, you know, or make them into NFTs. I think you said something like that. And then that's a bad idea because maybe you can explain, or yeah. in fact the only way to do it. The question is, the only issue is that there's no bridge of turning the NFT into something liquid. Not yet. that because I've met with a number of game developers that have, you know, of course, NFTs in their game. And they always talk about like, how do we create a marketplace? did these, these NFTs make it liquid? And, and that's always kind of the, the trouble is, is both on the, really the demand side. There's a ton of supply, but the demand side and there's really no market makers in the NFT space. Yeah. There are attempts. Yeah. Let's talk about that. It doesn't work because of inventory risk. yeah. That's the full conversation to have on another time. I can walk you through the exact thinking of how we got to where we are based on iterative experiments and on-chain deductions rather than hypotheticals. We actually did these structurings. You started with Peter Poole, which is very obvious, where you have an Oracle-based system where everybody deposits money. It's just USDC Vault and it's giving loans against underlying assets. And typically how they work is that they liquidate um when the price goes down. And that's what the work goes for. Thing is like, there is no price of a piece of hardware or really a price of even like CryptoPunks, right? It's kind of like the last sale. Like how many times does it trade a day? It trades about like a million dollars a day for like a lot of these JPEGs. Thing is the way that these Peter Poole designs work is that they require you to sell everything that's held in the book instantly if the price goes down. So you end up selling 100 mil of an asset that trades a million dollars a day. That's about 100 days of volume being done in a single hour. Doesn't really work. um It works for Bitcoin or it works for Ethereum. Like the largest event for Ethereum was the MakerDAO liquidation on chain. That was like when they had to liquidate a lot, like a lot, a lot of Ethereum, but it ended up being like 50 % of daily volume, which is doable. Like it's manageable. Look at exchanges, they're the same concept. There's a price that they refer to, but it's easy to liquidate because there's so much volume and market makers and stuff like that. um For things like real estate all the way to equipment, there's not that much liquidity. So you end up coming to a P2P design um where it's bilateral. If you default on the loan, the underlying lender gives you the asset. That also doesn't scale. So you have P2P, is limited in terms of the scale and becomes extremely niche because the only people that know how to lend to that pool are like specialists. They know how to like actually deal with like multi-variable lending. um So it comes back to like, how do you do P2Pool, but with the underwriting ability of P2P? And the conclusion that we had was that effectively it came to tranches where people can actually ascertain their risk levels. And this circumvents the need for an Oracle, whereas somebody is like, hey, The LTV is like, I'm to give like a 60 % LTV and this other guy is going to give a 30 % LTV. So it ended up being Peter Poole with an Oracle through a truncheon mechanism. And I can get to like more steps here, but effectively we went through a few more steps where people don't actually want like 100 truncheons, they actually just want two. um They want the truncheons to be tradable, like a bond, and you speculate on the value of those assets. And then you can modularize the risk as well where somebody can create those ones that are, that's more specialist that somebody can join that pool for. um All in all, I want to highlight like the way you develop the solutions for low liquidity markets is a sequential steps of structured finance, which is that you have these layers of liquidity that makes it so slightly more liquid for each layer you develop. All of that's done with the ministry protocol. um And that's what USDA is kind of built on top of. It's also why we don't give loans against operators. We give loans against hardware itself. Things that are truly like um not multivariable based on very speculative underwriting components, but really just purely a semi-liquid asset. um And because of that nature, um it makes us become a viable synthetic dollar where, well, we're underwriting um the slight liquidity of um hardware, the productivity, but also at the same time. ownership of underlying assets so we can actually liquidate in the case that things don't get repaid. I'm glad you brought things back to usd.ai. In terms of product development, I guess what stage are you in the product right now? It's not live yet. Maybe tell us about where you are in the product development process. Yeah, so in the product development, we... The launch of USC.AI in terms of the synthetic dollar and the hardware loans that we have coming up um into that launch, that's probably the first week of May as our target right now. We've entered audit with our auditors and we're getting pretty much everything, the last checks and boxes. There's a few of the mechanisms in the pipeline after the launch. One of them was this concept we have called QExtractableValue, which effectively allows redemptions to become a market. uh It's kind of inspired by flash bots. But effectively, one of the core issues is um with the liquidity assets, it takes a longer time to get out of your position. um And first come first serve isn't always the best methodology. So we invented something that's kind of like an auction to get out of your position to help with the stability peg. Because um we can't rely on a king curve like Aave to develop fixed income markets like this. um So we've invented our own, which is called QEV. um That's not out. That won't. Q extractable value. extractable value got it and the queue refers to you want to get out of your position you have to wait in line got it it's modularized by something called the textability module on Aave and a few others where it's on a kink curve. So that target utilization, it's kind of keeps you toward that utilization rate. That's harder to do with fixed income markets like us because we're getting a loan against hardware and it's like fixed term because that's how the borrower would prefer it. We use something called QXDraudableValue, which is every amortized amount of the loan is then put up for auction every month. So participants can exit out of their position um at that moment in time based on a bit that they have for that production rate. And it happens every month. um Those interesting new concepts that we're developing, that won't be out at launch. They'll be out maybe a few months after, ideally one or two. um But for our launch itself, if you want to mint USE AI, put that on Pendle, put it on Uniswap, put it on Morpho, um that's going to be launching in, end of this month um or early next month. And then participants will also be able to see through the UI itself what kind of loans are being provided against um based on the underwriters that are onboard our platform. Because we're modularized, so the loans can come in. They're underwritten. um They're collateralized on our platform. And then the loans are provided and the yield is generated. But yeah. So if we look at the market that USD.AI serves, you've got folks that need a loan, and then you've got folks that are minting. um How does, guess, USD.AI get minted? Maybe we can... Okay. And that could be, so that's available to anyone who wants to do that. And then they get a yield on top of that. for doing that. Got it. If they mint it, they don't get a yield. They have to stake it after they mint it. Got it. Okay. I'm trying to describe the, it's really a multi-sided market. It's more than just two sides at this point. You've got people that need a loan so that they can grow their, their business. Right. If I could just generalize it like that. And then you've got folks that provide USD.AI. They mint USD.AI. Um, and then so that USD.AI is, is, it's, it's underwritten. It's collateralized by Tether or whatever. ah that you'll accept. What else am I missing? And the third category would be underwriters, people that help originate the debt, um the people that look at the hardware and see what rates would be acceptable based on the market clearing rates. um And that's a marginalized nature where they get a fee for doing that activity. It's kind of like, I guess the best comparable would be like a Morpho. um Morpho services more DeFi like money markets. For us, there'll be more servicing capital markets for DePen. um But. that takes quite a bit of skill because they'll need to know some, be able to do some due diligence on the hardware itself. Yeah, well, it's really just one time because hardware is mostly equal to other hardware, just generally, even if it's geographically displaced. One Bitcoin miner is equivalent to one Bitcoin miner. One GPU is equivalent to one GPU, depending on the specs that you have. One antenna is the same as the one antenna. I just highlight we're not doing Bitcoin mining as an inner protocol. I think that's too developed of a market for it to be justified. And the rates become quite... similar to there might be some rates that are different based on geography. But at the end of the day, if we have this really robust tokenization structure where you know for sure one asset is one asset, the law of one price prevails where like it ends up being like quite similar depending on the hardware asset that you have. So the underwriting isn't as complicated as people would think. It's much, much more simple than the new underwriting businesses. because that's like analyzing cash flows, it's analyzing all these other risk factors. This is really just hardware in and of itself. And that's why we are more asset-backed than we are business-backed. And it puts us at, I think, the very limit of where you can see synthetic dollars being pushed, because the risk factors are more uniform, and uh the redemption schemes are a little more ah more structured. ah And that's why we went with this design rather than uh marketplace. The, how does your go-to-market look like that you've got really three kind of parties that you're trying to court? You've got the mentors and then you've got the underwriters and then you also have those that need loans. So how does your go-to-market look like and where do you spend your time? ah I've been spending more time on the on the borrowers and the depositors. The underwriters um is fairly uniform, where, like I said, it might be tough the first time, but afterwards, it should be pretty easy to set up and underwrite the specific assets for. And that's going through a whitelist process, and I think it scales with governance. For the depositors, um that's also extremely simple, which is you just mint usc.ai, but of course, um We are competing in the yield markets. So you look at like stuff like the basis trade all the way to the defy money market lending rates and making sure that yields are competitive with those rates. Ideally, if we beat those rates, they'll naturally deposit. And the borrower is also like, fortunately for us, it's a green field where a lot of the participants in the space just don't have any borrow that are looking for borrow. And it's also easy for us to distinguish uh which operators are somewhat uh um easier to onboard than others. um Frankly for us, like uh the risk here um is uh mostly like extinguished because we have this tokenization structure where the sale leads back makes it extremely secure. Like they don't actually own the assets in the loan. We own it in the underlying process and it's instantly insured in that sequentially right after. So there is no like attack factor here, which most people would assume. Whereas if you give a loan against the business, they're just going to walk away uh if they can't pay it. For us, um we have collateral um every single time. you have borrowers, unique use cases. um We've looked at deals from, like I said, uh obviously data center compute all the way to Tesla battery chargers um to, well, the big world of anything expensive deepened, things that just required that to one make money and also acquire that hard asset in of itself. And in creating very interesting new product line offerings as well, where lot of these deep protocols thought they were limited and hampered by lowering the cost, know, like having these cell phones that they have a million users, but now they don't have to sacrifice quality. So we're inventing new products that a lot of these deep operators can have imagined before. And that's what excites me the most is that like we're creating an ecosystem here. So those are the really two categories I've been focusing on. How are borrowers learning about about you about USD.ai? uh We try to work with the Deepin protocols as distribution channel partners. um So our customers aren't actually necessarily the Deepin protocols itself. It's the operators um inside the networks that are like the miners of Tau. They're operating H100s, but they also need Tau and a few other specs. um But there's pockets where they mine. And a lot of them actually mine multiple networks, AI networks, for example, for compute. um And there's even a crossover with, uh like, if you're mining Chia, you're also mining, like, this other storage network and a few others. So um the operators are, um they're multifaceted in a sense. um And it's also, yeah. so the deep end protocols, they refer their miners to you. If you want to grow, go to these guys and they can provide capital. Exactly. And I'm hoping that in the future, there'll just be direct offerings where the ability to borrow would be native to a lot of these websites or lot of these distribution channels where it really only makes sense if you take out that much like a house. Like a lot of the times, like, it doesn't really make sense to buy it with cash because one, lot of users don't have it. And that really opens up that purchasing power. Yeah. When you look at borrowers, you've got kind of the hobbyist that are, you know, just like the individual that you met who has, I think, one antenna, right? Or one Wi-Fi node. But I guess where does it make sense to provide capital? You've got the hobbyists, you've got the big businesses, and then what does it look like in the... How do you segment that market? So yeah, that that Helium node example wasn't like something that we would probably do right now, not yet. Because we're really targeting things that are held in a third party, things where we actually have a right to an underlying asset. So it wouldn't be in a household. It'd be something closer to data center. It'd be something like an antenna or energy and a few other things. Those are three sectors we tackle first, compute, telecom, and energy. And they're generally things that are expensive that makes it so that normal day participants normally wouldn't want to afford unless they have a way of lowering that cost. So our ideal range is like, know, the one to 5K as like the starting point and then upwards of really like 100 to 150K of like, of cost and operations or like, sorry, hardware cost in and of itself because that's when you definitely need that or not many people are going to participate because that's very large participants. But operators-wise, we can finance one small cluster of H100s all the way to a very large data scale operation, as long as liquidity is there and the hardware is being represented on chain correctly. Yeah. How do you throttle the balance between available USD.AI to kind of on the borrower side? Like, could you be in a situation where you underwrite ah more debt than uh USD.AI can provide? Yeah, and they just it's probably going to mature to the point where. um If you're talking about like if there's too much USDA state and not enough bar demand, that's a scenario that less people will just take. And in order for the yield to be, utilization comes down to like, well, sorry, the total supply goes down and utilization increases naturally. And then the rates are competitive to the market all the way to like, there's too much bar demand and then utilization is always going to be 100. And that's great. That's what we prefer. But frankly, there's the... The use cases that we're seeing scales very rapidly. Because once you do one POC with one piece of hardware, it can easily scale to represent the rest of the market. um And um it becomes a very repeatable process that scales quite well. And the rates are quite high for this too, because um these deepen operators are making quite a bit. They don't even say APR in a lot of uh deepened protocols. They say payback period, which is how long it takes to get all your money back. um To finance folks, um a six month payback period is a 200 % APR because you make all your money back in half a year. It's more than 200%. So the interest rates are very high in the DeFi world because again, you're giving loans against productive assets. Giving a loan against a tractor is very different from giving a loan against Apple stock, something that's already liquid or looking at the liquidity rather than looking at the productivity. So the rates are the highest in as of today, it would definitely be the highest in all of DeFi. um And then for the borrowers, it's just like pretty money for them because they just borrow against a rate that's much lower than their revenue. um And then they can scale it much faster. um Now, for the scale of what we think this can grow to, CoreWeave just announced our launch recently. A lot of people talked about it. That was the largest financing by Blackstone. um like in Blackstone history. And that was at like 11 to 16 % APR for a size that big, which is very high in TradFi as well, um but also very high in DeFi, um accordingly. So that was $7 billion for reference. So um Starkey, it's also $50 billion. That's what they claim is going to be the size of their operation. um it's not a scale invariant, in a sense, the yields. because it generates cash flows. There's a lot of users at the end of the day that's downstream from the hardware. Whereas Vitalik complains that DeFi is a bit of an urbiros, he made a tweet about that once, where is the yield coming from? It's based on speculation. Where is the yield coming from, Deepin? Well, it's coming from users, like consumers of the, yeah, from the data fidelity all the way to the actual use cases and consumption of final products. So. I love Deepin because of that. I mean, it's actual use cases from actual humans ah and in some cases bots, right? And so, but it's actual usage. uh Since you guys are pretty well connected to other Deepin projects, are you, maybe I can ask like what Deepin infrastructure projects are you like pretty bullish on um and uh where are you seeing the growth? Yeah, the ones I think all the early ones are very definitely interesting. like, know, helium high member were definitely like pioneers in the space. um The ones that are coming up, which I think you might be more interested in hearing um the ones that we've liked to work with and or generally that I think are quite bullish are like daylight. I think daylight like the founders also exprived equity. He thinks a lot about like the um the real end goal of what Deepin is. And he already knows his N4Fi. It really comes down to the credit behind the hardware itself, because hardware is just a proxy for financing. CoreWeave, to me, as an example, it's just an arbitrage between the rental rights of Nvidia chips and the financing rates of Blackstone. And that's the business, right? As long as one's higher than the other, it works. um And that's entire business of data centers all the way to everything in infrastructure. um If you say you worked in energy, uh generally focusing on the revenues or the costs. um And then I think uh Dawn is definitely one of them. mean, the guy actually had a very successful business with a very wide distribution before. um And uh we're doing an integration with them as well of like, how do you, um Dawn or Andrina, um Dawn internet, that's on Solana. um And it's also backed by Dragonfly. But they have a number of users, and their economics makes a lot of sense. And I think it's just ripe for financing. And they've done financing before, just with traditional venues. But for them, something more native um would definitely be a very interesting offering. um 375, I also think, is a pretty interesting project. actually, on the scale of the most expensive to cheapest assets for hardware, per asset hardware for deep projects, they're in the very highest spectrum. And they know they can't sacrifice the quality. And I see this happen with lot of deeper protocols where they sacrifice the quality of the hardware for distribution. I call it the deep in trilemma, like scale, price, and quality. The lower your quality is, the less likely you have the end user actually consuming the outputs of your hardware. But you get scale, right? And then maybe you can get a list or something. But at the end of the day, what really matters is like, is the network you're sacrificing your cost of tokens for? generating something productive at the end of the tunnel. And if it isn't, it's useless. It's vaporware. If it's actually useful, then you've built a network that's greater than a centralized counterparts. That's like the goal of Deepin, right? It's like, maybe the network is more valuable. But if it's not, if you literally just sacrifice the quality, then it's not worth it. For them, they'd never sacrifice quality because they know the value of the They have these hardware pieces on billboards and traffic measures, where it's extremely important that they don't sacrifice the quality of the stuff. So that's an integration that we would love to finance down the road. We don't need permission from a lot of these different protocols to operate as well. We just need to underwrite the hardware. um We really try to not be dependent on BD, not be dependent on um businesses and operations, really looking at the hardware in and of itself. It just happens that if you are integrated with a deep protocol, the tenant issue that I talked before, it's automatically solved. You're instantly integrated into network. You can easily rent out and stuff like that because you're getting some like cash flows as a result. So um those are interesting. I like Glow, know, David works project, um know, has a long background in the crypto space himself and thought a lot about the economics and with the production credits and stuff like that. A Glow. Yeah. And there. taking a pretty innovative stance on making productive economics around solo panels without the need for tax credits, know, just in of itself. um But there's a lot of interesting deeper protocols. And as long as they are having issues with their tokens, because they realize they're giving out too much, that's for me is a great sign of somebody who's a good founder. It's like, they don't want to pay too much in order to scale the network, because they know it's really expensive. When people are giving, Yeah, when people are giving like a payback period of really like four to eight months, and you can see this in the Missari report recently, where like you see the more expensive the note is, the more that they have to give to get past this hurdle of like expensive costs per asset. This is really mimicking cost of equity and cost of debt, know, to average cost of capital. It's like a very finance one-on-one thing. It's like should they have more debt or more equity? That's a perfect point where it's cheaper to just have a combination of both. Mm-hmm. For us, it'd be cost of tokens and cost of debt. But those things, well, and I'm talking faster here, but it's just showing like there's a lot more complexity, interesting economics that we can start experimenting. An entirely new industry that sits outside of DeFi, that sits outside of Deepin, just what we call InfraFi, that becomes a very interesting game of the next era of potentially. I call it max utility instead of max memes. It's kind of like the dialectics to max memes. Like we're really focusing on productivity here. And I love that. And it really feels like crypto is maturing with the entrance of USD.AI and now supporting these decentralized physical infrastructure projects with capital and debt. I mean, that's huge. And that's really the only way we can crypto matures and crypto grows is through initiatives that you are leading. Well, David, thank you so much for taking the time to meet with us. I'll be sure to share in the show notes all of the relevant links so people become aware of usd.ai Metastreet and Permian Labs. Sure, thank you for having me, Peter. And if anybody has any questions about like how the protocol works, I know I try to squeeze a lot in over the period, but it's just the nature of structured finances that you can't simplify a lot of the things that we have. um The user experience can be, but not the technology behind it. um All that, if anybody has any questions, um myself and my team will respond in our Telegram channel, um which is found on our Twitter, which is usgai.underscoreofficial. But yeah, and our docs are pretty comprehensive. Awesome. Thanks David.