The Real AI Gold Rush: Tokenizing the Shovels, Not the Hype

Transcript

Boom. Brilliant. 1 1. Wonderful. Hello everyone. Thanks for taking the time to come to the talk. Welcome to ECC. I promise I won't be holding you here too long. Um, I know lunch time is coming, so uh, before you go and indulge in your oysters and champagne, we'll, um, discuss the AI gold rush and, um, how we can tokenize the exposure to the real, um, revenue generating assets. So, to give you a context of AI revolution and what's happening, um, in the world of AI now, I'm sure that all of you, um, are familiar with what, um,

with what what AI is and, um, um, you use these products on a daily basis. But, to give you, um, context in terms of how large the industrial build out, um, is needed to, um, actually make all of this work. Um, the data centers and facilities under construction and these are the figures from September 2025 going in all the way to 2027 is 23 gigawatts. I'm sure that some of you guys, uh, um, immediately aware of how large this is. For the ones that, um, you need to do some mental math now and, uh, think about your grade 10 physics, I put together a few, uh, comparables.

That's an equivalent to 23 million microwaves running all at the same time and heating up your supper. Or, since we're in France now, that's the, um, energy that's emitted by 3 billion packs of cigarettes smoked all at once, continuously, forever. Um, so what's the hype about? Um, I'm sure you guys read about AI in any of the news, um, that you follow basically on a daily basis. Um, so just wanted to go over a few products that, um, you may have come across and, um, let's see whether this hype is actually warranted.

Um, here you have Copilot or any other, um, coding products for that matter. Um, I'm sure that some of you guys, um, have used it if there is any engineers in the room. Um, I haven't yet spoken to a single engineer that, um, told me that they're not using AI at all. Um, they're all using it in in some capacity. Um, I guess this is also the, um, main product with product market fit, um, in the AI, um, market, um, at the moment if we don't consider, kind of, the retail facing stuff that is just your chatbots and, um, um, take a picture and name an animal type of thing. Um, this is the product that will actually add uh, value in terms of creating GDP and,

um, improving productivity for, um, the smartest and brightest on the planet. Um, now let's go deeper. Um, I'm sure you guys have seen, um, a bunch of people stocking up on Mac minis and I think you know where I'm going with this. Um, Apple results must be brilliant for their Mac mini, uh, Mac mini sales, um, for the last couple of quarters, um, so let's see how they do. Um, of course, I'm talking about Open Floor and any of the any of the agents, um, for that matter. The people prefer to run, um, on the, um, separate devices or uh, virtual servers. Um, have any of you guys used agents, um, yourself? Show

hands. Okay, nice. Um, did you guys use it to book your trip to ECC? Help you with tickets, places to go, anything like that? No? Not yet? Well, that's the promise anyway, that that it will be doing all of these things for you. It will be, um, sorting your calendar, will be, um, buying you plane tickets, um, can also erase your mailbox as well if you're not careful. Um, then we have Modal which is essentially a network of these agents running around and talking to each other doing God knows what, maybe trading some shitcoins as well.

Um, it's quite a matter, by the way. And this is a stat, um, I want to, pause on a little bit and, um, just, um, have a think uh, about, uh, this number. So, 88% of all YC startups in Summer Batch of 2025 are using, um, and building something related to AI. Um, the other 12, um, is basically deep tech or anything, um, anything like that. So, 88% of B2B and consumer facing, um, software and, uh, companies that are being built in YC are, AI adjacent or AI native. Um, something to think about again, um,

crazy crazy numbers. I think this probably surpasses crypto in 2021. So, why should we care? Um, outside of all of the main, um, things about, um, that, you know, it's all going to make us so much more wealthy, it will do so many things for us and, um, we're all going to be basically collecting universal income from these agents doing all the work for us. Um, I guess the more immediate concern is, um, how do we actually get a piece of the pie? Um, how do we get exposure to this, um, industry and this asset class?

The easiest way to do it is, of course, your public equities. Something you can go and buy on, um, on your mobile phone now or phone your broker if anyone still does that. Um, so here I put together a few examples which I believe to be, um, the closest matches, uh, to earning on um AI um bubble, AI revolution, however you want to call it. Um, in my opinion, these are the four companies that encapsulate it perfectly. Uh, first one, Nvidia. Of course, everyone knows. Most valuable company on the planet now. Um, if you look at a chart, that's um, past year's worth of performance, it's, uh, up and to the right.

Um, exactly what we like. Um, the next one is Microsoft. Obviously not as AI native as Nvidia, but still a, um, massive partner for OpenAI, the main partner, and, um, very much involved in the, um, AI space. Um, here you can see that it was up and to the right for some time and then it's gone back down to, um basically um uh, beginning of last year's. Um, now we have CoreWeave for the ones that are not familiar. This is a neo cloud which is essentially a, um, cloud company that focuses on serving clients only in the AI space. Um, so they provide the compute service for companies wanting to train their models and use it for

inference. And we have Lambda which is another neo cloud very similar to CoreWeave. Um, but um basically serving serving very similar clientele. Um, you can see that, you know, performance ranges quite significant. Um, even for the two neo clouds at the bottom. Um, this comes down to, um, stocks behaving like stocks. You have, um, different markets. You have different management boards. You have different products, different decisions. Um, the one thing that unites all of these four examples is the volatility. It's something that you can't, um, necessarily escape, um, in the public markets unless you go down the derivatives, uh, route and try and structure some

product. But, basically, as a retail investor or a long only fund, uh, you're exposed to, um, uncap upside, but also no downside protection. Um, in terms of private market, uh, we have Anthropic that sits at 350 billion valuation. We have OpenAI that sits at 750 billion. And we have Google that, um, helps them along the way. Uh, basically, these two names, Anthropic is a parent company, uh, of Claude for the ones that are not familiar. And um, unless you are a, um, very well reputable fund, um, a large asset manager, uh, or uh, know someone uh, that has exposure to the aforementioned, you have

no way of getting into these rounds. Um, unless you can call up Sam Altman, um, you know, it's basically closed to you. Um, of course, you can, um, go and, uh, try and, uh, take some of the derivative exposure. Uh, Robinhood has tried to democratize that through um, tokenization of, uh, private equity. Uh, but this is still very niche and, um, you're not really sure what you're getting in return when you buy these products. So, this brings me to the point that behind all of these stories, all of the hype, all of these wonderful companies, there is one thing that makes it all possible and that's the GPUs themselves. That's the

metal, uh, the chips that run all this inference and generates all of the, uh, cats eating cakes and, um, making sure that your, uh, cloud bot, um, can actually run sustainably. Um, chips and shovels, we call it. Um, to give you some context about the chips right now, as of March 2026, um, rental prices are going up. H100 is, um, the Nvidia's Nvidia's high performance chip for the ones that are not familiar. Um, the, um, shortage of newest chips is also a reality. Uh, B200 and GB200 are the um, latest chips from Nvidia for the ones that are not familiar again. Um, and, um, some of the tech companies are

actually offering tokens, um, as a way to, um, boost engineers' productivity. And these are not the tokens that we're all um familiar with and, uh, um, you know, uh, have been buying on, uh, on Phantom, uh, for the last couple of years. But, these are the tokens that are essentially your your outputs, um, of the of the model. Um, so if you're an engineer earning roughly, um, 300 to 500 grand a year, um, you can maybe, um, get half of that on top as, um, as a token budget for yourself if you work at one of the largest, um, tech companies.

So um what's the way for us to get exposure to the actual chips, uh, to this magnificent machine that is powering all of the hype and all of the, um, surging AI use and AI investment. RWA's. Um, for the ones that have stuck around from, um, the previous chats, um, RWA's is, um, nascent um although growing significantly, um, segment of the crypto market. All of the figures that I'm going to show you now are basically up from March 2025 by 70 to 100%. Um, and total RWA credit, um, sits at roughly 6 billion. I exclude,

um, any of the pre-originated real estate um capacity that was then put on chain, um, simply because the number is so large and the origination happened off chain. Um, so this is kind of a more, um, accurate measure than if we, um, then drill it down to private credit. We're sitting at, um, 2 and 1/2 billion. The growth is roughly 100% as I mentioned, but still if you consider the um, amounts of stable coins that are circulating, um, across all of the chains, the number is quite insignificant at the moment. And this brings me down to the GPU backed RWA, which is what we do here at Gauntlet, and this sits at, uh, basically under a billion at the moment.

Um, to give you some, uh, perspective about what we actually do, um, our main line of business is tokenizing private credit and private equity exposure that we have to um, neo cloud providers. Uh, essentially, we'll be a neo cloud that will come to us and say, "Guys, I have a new contract. In order to satisfy that contract, I need to buy, say, 10 million worth of GPUs." We then go and underwrite this contract. We look at how valuable this contract is, how long it can last, how credible the borrower is, whether they've had any financing before, and how they've done with that financing.

We then give the money to that neo cloud at a roughly 75 to 80% LTV, um, and then we start getting the repayments from from this contract. We, of course, also take a first charge on the GPUs themselves. So, in case something goes wrong or the contract falls through, we still have access to the GPUs, which we can then either put to good use somewhere else or resell on the secondary market. The depreciation that the industry uses is 5 to 6 years straight line. From what we're seeing now is the GPUs that have been used for the last year or two are commanding a healthy premium on top of this straight line depreciation on the secondary market.

The way this works is it's quite simple. Um, you have the capital on the left, which could be retail, family offices, uh, private credit funds, private equity funds, high net worth individuals, anyone really that wishes to have this protected asset backed exposure to, um, the world of AI. They then deposit these stables into the Gauntlet platform. We then provide this financing to the, um, neo clouds and then they pay us back and we distribute the yield back, uh, to the token holders. Um, we are also branching out to the robotic space and equity of high growth, um, AI companies, which I won't be covering on this talk now, but we will be having a

webinar, um, in the next couple of weeks, so feel free to join and follow us on on socials for that. Um, next Oh. Okay. I wanted to show you a video. Oh, maybe it will. Let's see. So, yeah, we have, um, uh, AID, which is the main vehicle for, um, doing everything that I've just explained on the previous slides. Stands for AI Dollar. I'm going to play you a quick clip that will, um, explain how everything works and what it actually is. Oh, maybe it won't.

Um, is it going to play? Is there a way to do this? Yes. Anyway. Um, you can go and search it on YouTube. Uh, not sure why it's not playing. Um, I guess that's, um, that's a wrap. We have 5 minutes left. So, um, any of the questions, feel free to, um, ask me now. And, um, you can scan the QR code. This will take you to our website and then you can have a browser as well. Thank you.

Yes. What are the target yields that you're What are the target yields that you're offering? Great question. Um, for the, uh, private credit, um, the yields are roughly mid-teens, so they range from 12% to 18% depending on geography and the borrower in question. Uh, for private equity, um, the yield is much is usually, uh, 20 and above, but then, um, also here you're not getting the interest payments, you are sitting in a waterfall, um, of the revenue. Um, so the structuring is quite different, hence the yield is higher as well. Um, anything from 15 to 22 is, um, what we what we're positioned at.

Yeah. Um, yeah, so I just wanted to So, looking from an investor's point of view, you just stated that you you're about your yields are around 12 to 22%. Is there any systematic risk that comes with that, uh, default, uh, risk or any sort of non-commitment, uh, contingency? Great question. Um, obviously there is always a risk that a borrower will will not repay. And, um, here is where the asset backing comes into comes into play. Um, if we lend at 80% LTV, in reality we're still, um, getting a 20% discount on these assets that are then commanding a premium on the secondary market. So, for now, uh,

we're seeing that it's quite a healthy margin to the to the lending that. Um, with private equity, um, I guess it's a bit different simply because, um, even though their asset back is may may not be as easy to, um, take possession of the asset, but we have the covenants in place that will allow us to do that, um, even if we are a, um, minority shareholder there. And also for anyone interested, we're live on, uh, Base, Arbitrum, Ethereum, and BSC.

Forgot to mention that. Anyone else? Oh, yeah. Just a quick clarification. Sure. You buy the GPUs and rent them or the client? Is it the client that buys the GPUs and owns them later or is it is it you? Yeah, so if it's, um, credit model, then the client owns them and then we have a first charge on it. Same way how if you buy a house, um, you technically own it, but then the bank has a mortgage on it, um, as a first charge lender. Um, if it's, uh, private equity investment, then we are the owners of the GPUs.

So, then you can also resell them after the, uh, probably a couple years. Yeah. We're not able to do that with credit because then if if the facility is done, then obviously we release the charge and then they stay with the company. If it's equity, then yeah, we we're in possession. Anyone else? No? Okay. That's a wrap. Yeah, thank you.