The GPU Mortgage: How USD.ai Finances AI's $10 Trillion Build-Out

David, Troy, welcome to StableDash live. How are we doing today, man? Doing great. How about yourself? We good. Excited to chat with you. We we've been Zach and I were talking as we were kicking off the show. USDAI has been an asset, a framework, a team that we've been following for probably a year now, and we really just admire the approach that you seem to be taking around stablecoins and how they fit into emerging emerging industries.

So we're excited to unpack what you're building and get to know you a little bit. Thanks for taking the time. Absolutely. Thanks for having me on.

So I I think this is an exciting time. Right? We see Anthropic raise, you know, this this massive $65,000,000,000 round. You guys have had an amazing year. I I think you've you've kind of floated around that that 500,000,000 TVL mark. You've gone through a TGE, right, within the past week or two. Kind of walk us through, like, what was the beginning thesis for USDAI? What's this past year been like? And where do you see the space starting to go? We originally started like, we actually started 2021.

It's almost five years now. Generally, around the early DeFi period because most of DeFi was around money markets and, like, extremely liquid assets. Like, you could turn over the entire market cap of Bitcoin in twenty five days because the volume's so high. That doesn't usually work for most assets in finance, whether it's houses, companies, or stuff like that, where it's very illiquid assets. And it becomes like an entire sector.

That's the sales and trading, and then there's investment banking where it's focusing on things that just aren't very illiquid, all the way from asset backed securities down to just even, like, traditional investment banks servicing various niche services. And we kind of built Meta Street, which was our v one, more focusing around lending. Like, how do you provide loans for things that just don't have a price that don't have a mark to market, which is how often is it traded?

And then we we were kind of, like, doing a lot of pivots to find a sector we wanted to tackle. Yeah. We tackled watch lending. We tackled, like, Tesla charging lending. Remember that. Interesting.

Yeah. So we remember having breakfast, like, a lunch at, like, my first consensus with a with a BD guy from Metastree and when you were hit really deep into, like, watch lending.

We we were looking for a market that didn't have a lending market. It was, like, kind of the first heuristic. Like, it's really hard to get loans for illiquid assets unless the debt is tradable. It's kinda like raft staked ETH. It's my go to example. Before ETH staking Yeah. You had a sector called LSDFi where it was about, like, hey. Liquid staking derivatives. But before that, you when you staked ETH, it was stuck. You you couldn't trade it. You couldn't do anything with it.

It was only with the invention of a tradable debt instrument, the raft like this, that it it it flourished. Right? You had so much more TBL. You had so much more capacity. TBL exploded after that because you can borrow against that little bit, and it becomes far more flexible. You can you can imagine us, like, a bit like a wrap date deep, but for data centers. We're effectively taking all this money that's stuck, that is literally immobile inside a data center, and making it tradable.

It is not possible to do that right now, but this is the only way they can scale The US housing industry, this largest asset market in the world. Right? It's, like, $4,050,000,000,000,000 dollars. But this was only feasible because of Fannie Mae. They kind of took all the stock money and made it tradable, and then you borrow against it, and then you do more with it because you can use it in different functions of the financial markets.

What we realized was watches, Tesla chargers, and the markets we were tackling didn't have a strong debt market. You couldn't borrow against it because a mortgage implies the existence of a mortgage backed security. You can actually trade the debt. That doesn't that didn't exist for GPUs because we ourselves try to get a loan against the GPU we bought because it it was very basic math. We were renting out Yeah. We're renting out GPUs, $3 an hour.

But you multiply that by the eight thousand seven hundred hours a year, and then you realize, wait. I'm paying 80% of the value of the chip in rent in a year. And I'm like, I should just buy I should just buy the chip. You know? Like, why why would I pay rent? Well, people don't buy chips because, obviously, operational processes, but more importantly, it's just really expensive. You're not raising venture funds, like diluting yourself by 80% to buy chips that are depreciating.

You just have to rent it out and just, you know, refresh those cycles. It's also just, like, not feasible. You can't even afford the chips because there's no mortgages. That was our realization. We bought the chips, we're like, let's let's refi this. I refied my Honda Civic. I refied my my my kitchen equipment for a small business, but you can't refi the most important commodity in the world, the GPU. That's the basis of all $3,000,000,000,000 IPOs that are coming up in the next few months.

You can't get a loan against Wow. Yeah. It it makes no sense. You can get a loan if you're a large company, and you can do it against the credit of your of your large, large mega corp, but you can't get a loan against a single GPU. It's not feasible. Doesn't exist. Why doesn't it exist when you can get a loan against a house or a Honda Civic? Well, it doesn't exist because the debt is too hard to trade. Why is it too hard to trade? It it seems pretty obvious. Equipment finance, ABS instruments exist.

It doesn't exist because it depreciates too fast. If it takes you a year to issue these kind of instruments, but it takes but the the half life and stuff, so like a year and half, live for three years, then it takes a year to issue, but then you're already halfway through your life cycle of that instrument. It wouldn't trade. Yeah. That's why you you haven't seen an instrument with 10,000 loans put together to make it tradable. It's usually against a single corporation. That's the only way to it. Oh, yeah.

CoreWeave has 10,000 loans to make that tradable. But you're not trading the risk of or a single asset. You're trading it's like trading only Coinbase's ETH. That's not the same thing as, like, a bunch of different validators and, like, kind of abstracted REST. You're just doing a single credit analysis. So a debt market hasn't really appeared in the traditional markets. As a result, you can't get a loan for a chip. All I'm saying is there's no tradable debt instrument for GPUs.

This means that there is no interest rate, which means that you can't get a loan against a GPU. So what what's been the most challenging part about building this ecosystem?

Has it been the underwriting? Has it been sourcing liquidity? Has it been going to these data centers and getting them to buy in, you know, to this instrument, or were they that they were just hungry, and they were just like, hey. Thank you for bringing this product to market because we've been looking for it. Yeah. We we actually had the time.

Maybe I'll post it one of one of our tweets. We had a guy tell told us he loved us so much. He, like, he's falling in love with us because everything he said was everything he ever watched. He was like, dude, I I I am I am absolutely in love with you guys. You saw something that we we literally have been looking for for, like, almost a year and a half. So the moment they hear about this product, it's it's it's just it's just so obvious. Right? Like, you get a loan against a jet, but you can't get it against the GPU.

And these are operating businesses. They have, like, real customers. People are paying for them. There's not enough compute in the world. Anthropic doesn't have enough compute. So they wanna scale at the speed of CoreWeave, but they can't. They're buying it with all cash. Imagine The US housing industry. If everybody had to buy a house with cash, it wouldn't really exist. You don't even think about how expensive a house is. You just think about the mortgage rate. So listen for 2,000,000, but what's my mortgage rate?

That's that's that's the entirety of of house. Every asset is tied to this debt market because you only think about how expensive is it to me on a monthly basis. And we've kind of invented that as a category. What we've realized is our first customers were Neo Clouds. Right? These were people who specialize in building out chips in, like, in data centers. But now we're realizing every single AI company we've talked to is starting to explore buying chips. I mean, this this wasn't unusual.

You've seen OpenAI do it, but now you're starting to see smaller and smaller participants, these token factories, like the base tens or the or the fireworks, which are the hot startups of today. Even, like, the the rental companies, like the SF Computes and a few other examples I'll name, but they're all potential conversations and very advanced conversations of we should just buy the chip.

We should just vertically stack because why am I paying for rent when I can own when I'm paying the same in a more even less, actually, in a mortgage than I am in rent? I can actually own the chip at the very end of the day. Do you wanna rent for the rest of your life, do you wanna own a house? We're inventing something that didn't exist, which is the GPU mortgage. If you solve solve the small little problem of lending against a single chip, that's when you can lend against the largest batches.

And that's what we're starting to see too. You'll see an announcement later this week for our largest loan yet. Every every month, we've had our largest loan ever done, and it's getting bigger. Publicly trading companies are starting to borrow against us because it's something that well, everyone else a little surprising. But it it We haven't hit you with a money cannon yet. Well, maybe we owe you a money cannon. There we go. Oh, boy. Yeah. Yeah. It's kinda like the sausage. It's chip. It's just chip.

It's a compute hardware interest rate protocol.

Oh, we don't have that one. Let's see if we have So interested to see so interested to see this later this week. One, this paradigm is fascinating. I know just to make it a little personal. I was in my grandpa's kitchen few days ago. My grandpa ran a mattress company in St. Louis for his whole life, and it was pretty successful. He is retiring on it now, and he's, like, you know, day trading stocks and watching CNBC all day, and that's that's him.

And he was telling me about the business and why he was able to accumulate wealth through it, and it had nothing to do with the mattresses that the business was focused on. He wasn't really in the mattress comp business. He was in the real estate business. He effectively used his expansion of these stores to physically acquire actual real estate.

And that is the only reason why he was, you know, now successful was the the ability for the mattress company to give him the ability to accumulate real estate through debt and through, like, this exact process that you guys are talking about, but he's not like, I love the the similarities that entrepreneurs and businesses have probably felt in the physical real estate world for

since that market the debt market has existed there. That has been a huge way that businesses have operated. And, like, Donald, you're right. Yeah. So it's it's that same with Teen. Right? So would you equate it to a similar type of thing for chips? Like, how would you bring relate those two worlds together? Would you are you creating effectively the same mental model? The beauty of this is that blockchain specifically solved this problem better than anything else. Like, why why do this on chain?

Well, bank have banks have tried to create this data. Like I said before, they're doing the fastest process, but they need so much data. They're, I mean, they're doing a bunch of other things. We only do GPUs. We're the only real we're really the only participant that only lands GPUs. And it's also why we're the only participant in the entire market that will lend consistently under $300,000,000, even down to 300 k. Nobody else will even talk to you to ask for under under 50 mil loan. The door's that way. Right?

But, you know, we're not the only ones that realize securitization systems in traditional markets aren't very efficient. You look at, like, Figure. Right? They they've kind of figured this out too. Like, HELOC's obviously, a market that was, like, not very well penetrated, and they obviously, like, kinda took took market share. But you really kinda have to specialize in one specific sector, but most importantly, build the technology to expedite that process so it's a single multi call. Like, it's it's one click.

Right? Like, every time a loan's created and originated, it's it's immediately put into SUSDI. There's no holding period. There's no batching. There's no asynchronous, like, maturities, which is how traditional markets work. SUSDI is a perpetual instrument, much like a hyperliquid and and perps. Right? Like, the invention of perps kind of created this new new basis is that you can always have it have things be continuous. You can add things to it based on this perpetual blockchain in a sense.

That process of having rule sets and adding to it continuously is the beauty of well, what we are able to invent. This is why this is something that banks will never catch up to because it's an it's an infinite machine, without a doubt. And that's why that the liquidity is so deep and so much it it really is a 100 x improvement. And that's why our securitization, we we don't do. Right? It's just a a a debt receipt This solves this problem and then shortens that period.

So a three year loan is not that bad when the actual time it takes to get a loan is, like, under two weeks, under one week. And that's how you keep up with that ratio. A thirty year mortgage is fine. It's I mean, thirty year mortgage works if it takes a month to underwrite or, like, know, six months or a year. Right? That's that that ratio is is kept. Doesn't work if it's a three year long mortgage, and it takes one year to underwrite. It works if it's three

three years, and then it takes a few weeks. Mhmm. Then you can actually keep up with the systems. David, can you walk us through do you you're kind of building this market as you as you go. I'd imagine there are some key buckets of people, partners, clients. Like, how would you simplify how the market should be thinking about who u s t I USDAI is engaging with to to create this type of market? You it sounds like you have a big announcement coming out later this week.

What are these typical counterparties that you are working with and partnering with to bring this type of marketplace to life or this this new paradigm to life? Yeah. I'm on two platforms now. I'm on Twitter for DeFi, and I'm LinkedIn for Clouds.

So that's that's my audiences. But we we we're going to we're going to AI conferences globally than we are DeFi conferences, but they're both equally important. Frankly, this is very much a DeFi product. Like, the borrowers are just incredibly hap we found the PMF on terms of founding an audience that would wanna pay money for yield. And the yield is very simple. Like, where does the yield come from? It comes from the $10,000,000,000,000 of spend we'll do for next five years all on compute hardware.

It comes from the delta between how much they'll pay over the next three years. They call it total contract value. They'll pay you $250,000,000. How much is the bill how much does it cost to build the the actual chips and install them and do everything? It's gonna gonna cost you a 125 to 150. So that's the difference, and that's the yield. So it's a 100,000,000 in that exam. $2.50 versus $1.50. And then you're just trying to take as much of that as possible in the interest rate.

If it's, like, 10%, then obviously, just slightly that that delta compresses, and that's where the yield comes from. Yeah. People are saying, like, the yields are low. I mean, the yields are low because somebody has to pay for you. There there needs to be demand for money. And in terms of our partners, it's like, how do we generate this this engine? We've kind of established it. We've kind of proven the early PMF.

This is kind of why we teach Eid because we wanna launch early into our PMF process, which is what we just discovered. And we were saying this since the beginning of the year. Like, we first 100 mils are gonna be done, and we actually achieved it. We announced it in January, and now we hit a 100. We're about to accelerate that even further.

We've kind of now shifted focus away from the origination because it's such a strong origination pipeline, and now we're moving toward Did I see you at, like, a 2,000,000,000 in pipeline right now, I think? Yeah. More now. It's it's it's only increasing. Private credit is under retreat. Banks are are tapped out. I mean, you can't borrow more from Oracle. The the the revenue to pay down is is maxed out. The credit cards are tapped.

It's because of this fundamental issue in the very beginning of my call, which is you can't get liquidity on this money that's stuck in a data center. You have to, like, pull it out somehow, and that's for derivatives. Right? That's that's rough take deed, but for e staking. If we didn't have rough take deed, all the ETH would still be in e staking. Would actually be better for the East price, but this is dollars in terms of that just could be more productive. They wanna lever loop data centers.

That's that's that's literally what we're trying to enable. So if you if you can solve this problem, then you can lever loop a positive carry because you know this is gonna make money for next ten years

as as long as it's structured correctly. And yeah. Can we dive into Chip a little bit? Right? Because I think, you know, you're on crypto Twitter. You're seeing all this talk on hype. You've seen Matt Matt Hogan from Bitwise saying that hype is as easiest conversation with financial advisers. I think we're seeing this renewed interest into tokens as the bear markets kind of simmered down. We had Yano on from Blockworks talking about the token transparency alliance.

Help people understand you know, I think a lot of people just even ask, like, why a token? You mentioned TGing early on in your product market fit life cycle. Help us understand where Chip fits into this and and how we should be thinking about your token.

Yeah. But I think the foundation has been pretty quiet about utility because we're doing quite a bit in terms of rejecting the existing conversations. People say a buyback and burn is the only mechanism you have, but then it it's a little questionable. Right? Like, where exactly Yeah. How does the growth come from? Does does the growth come from just, like, volunteers in terms of a a company that's making it work?

Maybe if you're, you know, already worth $300,000,000 like Jeff and Labrador Liquid, then you can afford that charity. Yeah. But end of the day, like, you you don't wanna do a buyback and burn, in my opinion, in of itself. I think buyback and burn is one of the most important mechanisms should establish, but I I think it's the way that we've accepted it now, a lot of companies don't aren't having value accrual from that system. I think some are starting to evolve that formula.

I think VVV is a good example where there is still an equity company, but they're also doing buyback and burn based on certain performances. And that has started to work well. We will be announcing ourselves along with the USDA Foundation our own plan that isn't buyback and burn. We call it commit and compound where you will see a decreasing amount of supply over time, maybe even more rapidly than a simple revenue. Because that that is the core issue with buyback and burn. It is x revenue. Right?

Like and I'm all about multiplication, as you can tell. Love it all. But 1x revenue is just not that great. I mean, it it worked great if it's a very fast platform business, and that's really what Hype and VBB, like, marketplaces where it's, like, more about revenue churn, but we're not an overnight, like, revenue source. We like, our revenue is split across three years for every loan that we do.

And that's powerful in of itself because you can DCF that back forward, and you can multiply how much revenue you can generate based on the sustainable cash flow, based on that and amplifying that so you can do even more amplified revenue a multiple of revenue. It's not just a one x. Because right now, that is the one x efficiency in terms of the token design.

So we'll we'll be sharing more, but I I think the the design space for, obviously, the balance and tension of token to revenue or token to equity and equity to me is just revenue. Right? That that's what it represents. I think there is a a formula to be solved, which we've been spending a lot of time on. It's been almost my entire corporate objective more than anything else. Because we're starting there's two questions in the start. One is how do you make money, and then two is how do you share it?

They're both equally hard problems, but they're solved sequentially. So we're the second stage of the company where we're trying to figure out how do we how do we bring everyone on board, how do we have the participants and stakeholders rewarded, as well as speculators and investors, as well as, yeah, the the various amounts. I think clarity is starting to help, but clarity is a bit of a distraction because because, you know, a token is not equity, and equity is not token, but they're both very powerful mechanisms.

Any case, I think people will be at least at the very least, intellectually curious about what we're we're planning to do Yeah. On the equity side as well as what the foundation plan through the token.

But I can't reveal yet, but this is I really wanna put a big Part two. We'll have you back on once you announce it, and and we'll do a proper proper deep dive because I think just, like I said, personally, the questions that we get all the time is like, hey. You guys cover stablecoins. Like, at the end of the day, people wanna engage with financial media because they want an edge. Right? They wanna make For sure. They want they want yield. They want something. And for us to just be like, yeah.

Well, stablecoins move money faster. It's kinda like, where where where can we point them next? So I think that's something we're excited to bring our audience more of.

Our our quarterly annualized revenue is about $30,000,000. And so it's it's it's starting to show PMF. The question is how do we amplify it so we're not trading at 1x revenue? How do we make it so its multiples higher and the valid capture is better represented to the users? Here we go. Yeah. And double counting does it doesn't work. I think that's one common thing. People try to double count equity in tokens, but it's one way or the other. And that bridge, as long as it's well defined, I think, can be a new meta.

I think I'm hoping everything that we do, as you can tell, we try to innovate the efficient frontier of, and token design is one of the is is just another product to me. And I think we've we've cracked something that nobody else has done before. I I was inspired by, like, how Venice thinks about it more than somebody like some because yeah. I mean, hype works, but I the value Who is goal it you said? Who do you look at? I was looking at SVDD and DM as well as VVV itself and how you kind of evolve over time.

But I'm also looking at how DATs are structured very poorly, by the way. I I haven't been impressed with, I think, what DAT structure, and I think that's that's a shame. And we we ourselves have to confront this because we deal with publicly traded, like like, go through a situation where it says awareness is like new Clubbook, and we're lending to a number of them. And these are the counterparties that we work with. Right? They there are institutions that will never touch tokens.

So you can't just say, like, I'm just gonna be a token forever. That might work with an extremely retail app like Hyperlink. But for us, we deal with institutions multiple times in an hour. Yeah. We have to figure out a way where they could be stakeholders too. And this is this is how NVIDIA grew. Right? They're kind of investing in everyone, and everyone invest in thumb. It's just it's it's a safety net, and that's what we're building ourselves to become and, like, execute on that strategy.

And this requires an equity vehicle. This also can be amplified with the token vehicle on the DeFi side. So we're gonna balance it out. We're gonna share it, and we'll see if people like it. David, I I have one more question for you. Learning a lot from you today. Again, thanks for coming on. Final question.

How badly are we underestimating you guys? How badly is the market generally just mispricing or misunderstanding what you guys are really building here? Because this this seems like the the general GPU conversation in the context is Seems AI. Just absolutely massive in my in my mind.

Best I mean, one of my favorite quotes is, like, nobody knows how attractive they are until you compare yourself to how attractive you think your first and second girlfriends were. I said, like, oh, yeah. You have no idea. I think the best way you can kind of estimate us is, like, see see who our investors have been and see what they say about us after they've invested us for five years and if they actually believe in what we do.

So it's hard for me to, like, kind of personally show, obviously, how beautiful I am and then how great the company is. It's yeah. Ask ask our supporters what they think about our execution ability and what we plan to do in the future and if they're gonna sell their backs. And that's a better so there's a easier way to measuring it. But I I think my goal is that GP lending is not a niche. I'll finish it here. It's like, I think people think GP lending is like a it's like a equipment financing.

It's like a small subsector, but a very large category. What we're seeing is the end of software as a service. Like, it's it's an entirely new era. You can't lend to SaaS anymore because, well, what was great cash flows is now seemingly gonna go to zero. It's not like this ten year DCF terminal value. The end of terminal value, they call it. Right? You can't lend to SaaS company. So SaaS credit is dying, like $3,000,000,000,000 of credit.

AI companies, you might be able lend to, but what are you really lending to if you're lending to an AI company, which is now all the new companies in the world? You're not lending to cash flows. You're lending to their 90% expense costs. All of the money that like, you look at OpenAI. You just go through the s ones. Where does all the money go for all these companies? It's not salary. It's going to it's going to compute. You can't lend to an AI company. You can only lend to compute.

And when it comes to compute, it's really 70% of the cost of a data center is 30%, the shell, the land, things that are securitizable, but 70% goes to the GPU. So GPU lending is not just a niche sector. It's a replacement of all credit in technology. When it comes to anything in tech that you're giving a loan to, you're not lending the cash flows. You're lending to the GPU itself because the GPU is what generates the cash flow. So GPU lending, to me, is what the only thing you'll see on Bloomberg in ten years.

Right? You'll just be trading GPU credit. You won't be trading SaaS credit, corporate credit. They're just proxies to the same thing, which is, like, how much am I gonna make money on for these for these assets? And, eventually, the replacement of of debt itself is just gonna be asset backed lending for the basis of AI. So that that is the vertical book that's playing, and it's the thesis that we believe in. And, yeah, I think That's a great ending statement. Yeah. So I I I think speechless. And nothing else.

The TAM the TAM is a $100,000,000,000,000. It's not it's not one or two or three or or 10 for the next ten years. It's all The TAM is the TAM. The TAM is the TAM. TAM is the TAM. TAM is TAM We're getting a look. Just backdrop. The new numerous. The TAM is the TAM. If we fall short in our race of of this AI build out, the dollar is gone. Right? China has won. Sovereignty is questioned. So the dollar is now printing not for oil, but it's printing for making sure that we win the AI war. So it's it's money. It's credit.

It's sovereignty. It's everything. Wow. David, thanks for the time, man. Tam is the Tam. Thanks for coming on. You.

Sure. Yeah, David. That's awesome.