Blockmate Ventures: Building Exposure to AI Infrastructure

0:00 Good morning, good afternoon, good evening, wherever you are in the world today. With me is Dominic Carrozza, founder and chairman at Balkan Adventures, who will discuss with us how the company is advancing multiple exciting AI infrastructure projects. Dominic, thanks for being here today. How's it going? Yep, fantastic. Thanks for having me on, Jay. Glad to have you on board as well. We got a presentation first by Dominic and plenty of questions that were pre-submitted to work through. But Before we get into the presentation, I want to take care of a bit of quick housekeeping for everyone today. So the chat and the Q&A function at the bottom of your screen is open and active throughout the session. If you have questions as you go in, feel free to drop them in. We've already, like I mentioned, we've already received a few ahead of time via email, but we'll take as many live questions as time permits.

0:50 If we don't manage to reach yours during the session, don't worry, the BlockMate team will receive all unanswered questions so they can follow up with you directly using the contact details that you provided during registration. If you look at the bottom of your screen as well, there's also a request meeting. If you'd like to book a direct conversation with the Block.me team, you can hit that button and the team will get you connected after we wrap up. Lastly, Today's session is being recorded, so a full replay will be available likely later today. And you'll receive a link sent directly to your inbox, and it will also be published on SIX.com as well as on YouTube. So now I'll pass the floor over to Dominic to kick the event off with his presentation.

1:32 Thanks, Jay, and g'day everyone. What I'd like to do today is give you an overview of BlockMate. I'll spend about 15 to 20 minutes going through the presentation, and then I'd like to try and keep this as interactive as possible. So for those that have got questions, feel free to ask. Maybe to start, a couple of the big thematics that we're seeing in the marketplace are AI. I think it's clear that AI is now really starting to grow and take hold and many companies around the world are starting to take advantage of AI and all of its benefits.

2:18 But at the same time, it's still very nascent. It's a bit like where the internet was in the late 90s. And the other big thematic is energy. And when I say energy, I'm talking about power or the lack of power globally in order to be able to power all things AI. And they're 2 things that I'm going to address in today's presentation. In terms of background, I'll maybe start with that and then jump into the presentation. I've been involved in technology since the mid 90s, listed a number of companies both on the Australian stock exchanges as well as the Toronto stock exchanges and my last company that I founded called Banxa. We sold that to Hong Kong listed OSL group for 62 million US earlier this year.

3:07 So in effect, BlockMate's got a number of people involved who have been there and done it before, as they say. So what do we do? We're a company that owns and operates AI infrastructure. Companies pay us in order to be able to use our infrastructure to run their AI. And how we make money is by renting out the GPUs. And I'm going to talk more about that. And I want to get sort of too geeky and sort of too nerdy, but I'll just try and sort of make it really simple. People pay us to use our infrastructure. And at the same time, we're also developing a very big asset in the US in Wyoming, which is a 200 megawatt AI data center.

3:56 And the 3 core assets that the business currently owns is a NVIDIA B200 AGPU node, which we actually just announced yesterday. Our Wyoming AI data center development, and then on a fully diluted basis, we own 50% of HiVelo, which is our consumer facing product. So in effect, we're all about AI infrastructure and making it available and generating income on that. So where are we today? With our announcement yesterday, revenues are now just starting to flow. We've ordered out our 1st NVIDIA B200. It's expected early Q1, and we expect revenues to flow a week after installation.

4:46 Wyoming, which I'm gonna talk more about as well, is a land opposite a substation. It's less than half a kilometre, about 500 metres away from a substation. I'll talk about why That's important. And in terms of insiders, we did a capital raise early this year and just under a third was subscribed by the directors. Once again, showing absolute alignment with shareholders. And as you'll see through the presentations, there's a number of very strong catalysts in my view in order to basically move and create value in the company for all of us as shareholders. We know that AI is growing, both in terms of training and inference, And more and more people and companies are now starting to and wanting to use AI for their companies.

5:40 And the model and the focus that we have is where we own the AI equipment, the GPUs and the video GPUs, and companies pay to rent all or part of those particular GPUs. And we obviously get paid for that. And our first iteration is that We've teamed up with a group called Scanit, who's our operating partner. They're gonna be hosting, monitoring, and maintaining that particular GPU. That way we've kind of outsourced that first part of it, just because it made sense for us to do so. We'll eventually take that in-house over time. And in effect, what we're doing is we've got infrastructure, we're renting it out and we're generating income.

6:26 And you may ask, well, why don't people use Claude or OpenAI or, you know, ChatGPT? And sure, many companies, millions of companies around the world will do that. But for companies that have got sensitive data or proprietary data, they're not gonna use those open models. They're gonna run their own AI models. They're gonna own the confidentiality and basically control who has access to it. And this is really where we see the next big wave. So it's around the confidentiality aspect of data. And then secondly, it's also about data sovereignty. You know, basically countries like in Europe or places in Europe and Australia, sovereign data is becoming increasingly more important. And in fact, from a European perspective, it's actually legally required to keep certain data on shore within Europe.

7:21 So double click, we've just announced in terms of the acquisition of our 1st NVIDIA B200 node. The objective here is to really prove up the demand, the utilization, the rental price, and I'll talk more about the financials around that as well, and then look at which channels actually perform. In terms of the costs, it's 635k US, which is basically a turnkey price. We're going to be working with Scannit, as I've mentioned. We're gonna be hosting in a US data center for our first major node. And then we expect delivery Q1. And what I should say is, we've actually already got customer demand for this particular node. So we're actually starting on the front foot and frankly if I had a stronger balance sheet, we'd be buying many, many more because we do know that we've got customer demand.

8:17 In terms of the numbers, the financials, the way the investors should look at this is you've got some hardware, it's just over 600K US, there are 8 GPUs and the way that it's sold is you rent it on a GPU hour basis. That's the best way to look at this. And so if you take this particular node, we're talking just over 70,000 hours over a year. And then from a pricing perspective, right now, the median price is around $8 US per GPU hour. Now, if you look at our financial models, and if you look at the break even here, we're working on the basis that we expect payback in around 2 years for the hardware.

9:08 And noting that the hardware has a minimum of a 5 year lifespan. And in fact, a lot of the Nvidia hardware is still running 8 or 9 years later. It's very hardy and there's still revenue to flow even many, many years down the track. And we've obviously taken a much lower GPU per hour valuation or cost when it comes to working out our particular ROI. So I'll leave that with investors to model that themselves. So what are we doing? We're starting with 1 node. Starting in Q1, we're going to take that data. We're going to take those particular learnings. And then from there, we're going to add further nodes, continue to build it out. And I'll sort of talk about Hibello, which is our consumer facing site in a moment.

9:58 And then longer term, our objective is vertical integration. So right now we've got a GPU, NVIDIA GPU, we're gonna host it in third party data center. But over time as, and I'm gonna talk about Wyoming in a moment, once we get that permitted and built, we're gonna be able to house our own GPUs in our own data centre and ultimately own a larger part of that value stack. So in effect, we're earning more GP, more gross profit from that particular opportunity.

10:31 Now, we've also acquired land in Wyoming. As I mentioned, it's about half a kilometre, about 500 metres away from a substation. And you need to have your data centres close to a power station or a substation, because running high power lines is very expensive. And so right now we are going through a permitting or planning process. We're working with the local city on Wyoming. We expect to have some clarity over the next quarter or so. And in terms of the capacity, we're talking about 200 megawatts is we know that that's actually available and it's possible on the site. And so what we may do post-permitting is do a deal with a Hyperscaler or NeoCloud for half of it and then use half of it ourselves.

11:28 And just to give investors a bit of a sense of the growing market. So for a powered permitted site in the US, today if you wanted to buy a site like this, effectively a greenfield site with power and with a permit, it's currently selling for between 1 to 1.5 million US per megawatt. I'll repeat those numbers again. They're selling right now for between 1 to 1.5 million US per megawatt. And So on the basis that we have 200 megawatts capacity and power on our site, we're talking about an enterprise value just for this site of between $200 million to $300 million.

12:14 It's actually quite extraordinary, which is why, once again, we acquired the land and are going through the engineering and the planning and the process to do that, because the prize at the other end is very, very large. I should note that our Wyoming property is in a place called Evanston. It's just over an hour from Salt Lake City, and it's in an industrial area. So all the other places around there are basically industrial, you know, unlike some of the other data centers that I know are having some problems, because they're also located in more residential type areas. As I mentioned, we also own 50% stake in Hivelo, which is gonna be our go-to market, our customer facing platform.

13:01 So we own that 50% on a fully diluted basis. Hivello a couple of months ago launched Bitcoin cloud mining, you know, effectively taking infrastructure and making it simple for people to buy the service, you know, using their credit card using stablecoin. And what I should say, and I'm super happy to say this, we launched just over a month ago and the product sold out. It was actually amazing response from customers out there who wanted to engage in Bitcoin cloud mining. So we're gonna use the same process and adopt the same strategy with GPUs. So you will, as a customer, will be able to go to HiValo.

13:46 You'll be able to go through that customer experience and then rent GPU time on our servers for X dollars per hour per month, like the model I explained earlier. So really, Hive LO is our direct to customer channel. So it means that Blockmate doesn't need to go and build that because we already have that within Hivalo. And Hivalo obviously is also there to make a margin, which obviously is benefit to us because we're a major shareholder of it, of roughly 15 to 20% of what's being sold through the high value platform. And this once again, just really reinforces the opportunity for vertical integration. You're owning the land and power, then owning the GPUs, the NVIDIA GPUs, and then owning the customer relationships.

14:40 So effectively, we got more bites of the cherry, as they say, which means more gross profit for us as a company. In terms of the leadership team, I've already mentioned with regards to myself and the exit and the experience that I've had. Our CEO is Justin Rosenberg. He's a chartered accountant, very strong experience in the startup space, both across Australian Stock Exchange and Toronto Stock Exchange listed companies. And then we have Corbyn and George, our non-executive directors. And then Fred, who I'll mention, who has got a lot of data center experience and his claim to fame was running the data centers for the Vancouver 2010 Olympics. So as you know, a lot of load and a lot of pressure and stress running that particular infrastructure and we're glad to have him on board.

15:33 So in terms of the milestones, in terms of the last 6 months or so, we gave the market an update with regards to Wyoming and we continue to update investors with regards to the engineering and surveying that we're doing. And then obviously the planning process in July, we completed 1 million dollar raise and 29% of that was subscribed by the directors. So we're absolutely aligned with all the shareholders. And then in August, we launched high value Bitcoin cloud mining. And then we just a couple of days ago announced the bonding agreement with Scannet for the NVIDIA B200 nodes. Looking forward, we expect delivery early Q1 and then first revenues flowing thereafter.

16:23 We expect some clarity and we do believe we've got a line of sight with regards to Wyoming permits, with regards to building up that particular data centre. And then from there, I'll sort of talk more about some of the catalysts in some future slides. Just to give you a sense of, you know, we're really at the early stages. And this is, I'm going to say, not too dissimilar with banks. And when we started Banksa, it was an idea. It was basically 0, 0 value there. And then over time, we ended up building it and sold it for 62 million US, which is about 80, 85 million Canadian. So right now we're at the starting gates. We've got a market cap of under $5 million, both Canadian as well as US. And just to give you a bit of a feel for some other companies in this space, there's Carrier Connect, Shire, $50 million, Canadian Valuation, and then BitZero and DigiX, which both have a couple of hundred million market cap.

17:23 In my view, the reason that I've presented those like that is that's where I believe we can go as an organisation.

17:32 In terms of some of the key catalysts, I've spoken about what we've done over the last 6 to 12 months. Really moving forward as we get more monthly operating data from our Nvidia B200, We'll then make some decisions around effectively acquiring more and the objective in my view will be acquire more. There'll be announcements around Hivello and launching the GPU marketplace, as well as then with Wyoming, whether it's a joint venture, a long-term lease agreement or some kind of customer contract, there's a lot of positive news planned in 2027. In terms of the capital structure, so we have roughly 171 million shares on issue and 232 shares on issue on a fully diluted basis with insiders, management, the directors owning just under 20% of the company. And one of the things that I'm going to note is on the balance sheet, for those of you that have read the recently released financial report, you'll see roughly a $7 million liability on the balance sheet. I want to note that that's basically a Hivelo safe or sometimes called a convertible note, which sits as a liability on the Blockmate balance sheet, because right now we have to, as in Blockmate has to consolidate Hivello, but they are the debts of Hivello, and we believe by the end of this year will be converted into Hivello shares, and

19:02 then that Hivello debt will basically be extinguished. And just to be clear, it is not a blockmate debt. It's just an accounting nuance. In terms of the capital markets, as you know, we're listed both on the TSX under code MATE and on the OTCQB code MATEF. So maybe at this time, we can open it up to some further questions. I'm not sure, Jay, if you've got some questions you wanna read out, but I've also got some questions that some investors have sent me as well. Yep, so thanks for that again, Dominic. So there were a few kind of questions sent in the email. So let's start with 1. You know, this investor mentioned, you know, they repeatedly hear about the high demand for enterprise AI hardware, which includes of course, the Vaco P200.

19:52 Once this first node that you've acquired is performing well, how scalable is the model from a hardware acquisition timeline perspective? Yeah, we've got some very good partners in the space that give us a clear line of sight to infrastructure and being able to actually acquire the infrastructure. We have partners with third party data centers that we're going to use in the early days in the first instance, but longer-term objective, as I mentioned with Wyoming is to actually own our data center. We are also looking at data centers in Europe, as well as in Australia. The reason is, I talked about sovereign AI

20:33 where data has to be kept onshore. That's super important. And so we're looking at some opportunities specifically in that area as well. And I'm going to say, and I should have mentioned it earlier, this is not dissimilar from a business that I built from 1999 to 2005. I built the second largest virtual web host in Australia, which we ended up selling in late 25 for just over $20 million. And it was very similar. We bought hardware infrastructure, we built a software layer on the top that made it really simple for businesses to consume. And then we went out there and marketed and generated income from it. And We were a highly, highly profitable company at that point.

21:22 What changes is we're buying CPU servers in those days, and now we're talking about GPU servers today. The infrastructure and the concept is almost identical to what I did 20 years ago. It's just that the server names have changed. And so, you know, this is one of the reasons that I'm getting back into this space in a big way is because I see significant opportunity. The way that I like to look at where we are with AI, we are in the late 90s in terms of the equivalent growth of where the internet was at that stage. And if anyone remembers the internet in the early 90s, It was very early and very primitive. And so, you know, I think we're really at the starting, you know, almost ground 0 in terms of this opportunity.

22:14 Awesome. Thank you, Dominic. Actually, continuing on the topic of partners, I think on your deck, you mentioned, or rather this investor says that you mentioned that ScanIt doesn't guarantee pricing or utilization. And also you kind of mentioned that pricing and perhaps even demand long-term can be volatile. So could you expand on some of the strategies that BlockMate and Scandit will take to protect and grow long-term cash flow?

22:40 Yes, you know, absolutely. So without wanting to get into, let's call it, the nuances and the detail, part of what both Scandit has built and we're also building at Hive Allo. And I'm going to call it

22:57 a mix and match engine. So there is always demand for AI compute. It just depends at what price. And so part of this, you know, bit of sort of middle layer software that that we've, we've built, and we're also building and augmenting as well, is we will basically look match customer demand with our servers. And if there's no demand at $8, we'll then go down to $7.50, and then we'll match the demand. So it's basically like a kind of internal bidding engine where we can make sure that all the servers are running at 99% plus capacity, you know, which is really important. And if I take, you know, and I can't reveal all the nuances with relation to some of the servers that Scannad are already running.

23:49 But I think it's safe to say they're running at 95% plus, because of this particular software application. Right now, there is more demand for AI compute than what there actually is AI compute available, particularly for sovereign AI and for proprietary workloads.

24:11 And because there's demands and delays in terms of chips, and there's big delays, and there's not enough power. We've already, I'm sure, read it in the paper. There is just not enough power out there to satisfy the demand of all the data centres and infrastructure that needs to be built today.

24:34 Thank you. The next question asked about a specific slide on your October deck. So they said, or they asked, you know, on the comparable slide, you mentioned valuations rise based on 3 levers, so compute, customers and power. Longer term, as you move from the current market cap to 1 closer to on the far right side for DigiPowerX, how do you foresee that growth being financed? Is it cash flow from ops or how does that work? That's a really good question. So first and foremost, we're very concerned about dilution and as a sub $5 million company, effectively right out of the gates, we're quite limited with regards to what we can do at the blockmate corporate level, which is why, just to give you an example, this first acquisition of the NVIDIA B200 is in an SPV, a special purpose vehicle, and we're gonna be raising the debt inside that particular vehicle.

25:35 If I take Wyoming, what we'll be doing there, and there will be a Wyoming SPV, and then we'll be raising the equity and the debt inside that particular SPV. So this is a way that we can reduce or minimize the dilution at the block mate level and keep each of the opportunities within their own specific SPV. And maybe at some point, once we've got A230A billion dollar market cap, then we can kind

26:06 of roll everything back in if it makes commercial sense. But that's how we will deal with it, at least in the first instance, when we're a sub-$100 million market cap company. Okay. Thank you, Dominic. So a few more questions. The next one was on Hybella specifically. You mentioned earlier that Hybella launched back in August and you also said during your presentation that it actually sold out. So, could you kind of comment on what the customer feedback has been for it since the August launch, and perhaps how that impacts the ultimate vision for BlockMate, which is BlockMate-owned compute being supplied to Havelo customers? Yeah, so BlockMate's the infrastructural layer, and what Havelo is the customer layer. And when you're dealing with customers, you've got invoicing, you've got billing, you've got customer service, you've got all of those things that go along with managing customers, whether they're consumers, or whether they're businesses.

27:13 And so what Hive LO does, it really helps helps accelerate what we're doing at BlockMate. So BlockMate can focus on the infrastructure side, the Wyoming, managing infrastructure, the data centers, and all of that big, heavy infra, while Hivello is much more nimble and more software and customer centric, where it can go out there and actually acquire customers much, much faster than us trying to do it from the ground up. And obviously, there will be a relationship between the 2 groups.

27:45 We launched Bitcoin cloud mining just over a month ago. It sold out. And investors might ask, well, what's Bitcoin cloud mining got to do with GPUs? Now, at a product level, they're quite different. But when you go down the next level, they're very, very similar. So we're taking a Bitcoin miner, and what we're basically doing is like, think of it as fractionalizing, and we're making that available for customers. What we're doing from a GPU perspective is we've got the GPU infrastructure, and then through the Hivello platform is we're making that available to customers. So the hardware may be different, the actual process in terms of making it available to consumers out there is relatively the same. And then obviously, billing's pretty much identical, customer service is pretty much identical.

28:39 So it's just a way that we can accelerate or fast track us going to market. And as I mentioned, one of the key catalysts, there will be an announcement soon. I can't say when, but I've already alluded to it where GPUs will be available through the Hivello platform. So all I say is watch your space.

29:02 Thank you. Just one last question that was pre-submitted on Hibelo. So again, the $7 million of liabilities on BlockMate's balance sheet is again attributable to Hibelo's SAFEs, But when those safes eventually convert beyond the removal of the liability, is there any other impact on BlockMe in general? No. So those safes related to Hivelo raising capital separately itself, which comes back to the earlier question about the SPVs. The way to look at high value is it was kind of like an SPV and money went directly into the SPV and the investors, external investors own, I think it's just under 50%, block made owns roughly 50%.

29:51 So that's the way that we can grow the company by having these SPVs, raising the capital. So we expect at this point that we will convert all of those safes, which is basically like a convertible note, which under from an accounting perspective without getting sort of too caught into accounting standards, whenever you've got a safe, even though it's gonna convert into equity, you still have to, it still is a liability on the balance sheet of Havelo and then conversely, Blockmate until it actually does convert. But as I mentioned, we're expecting this to occur relatively soon. You know, I'd be surprised if we don't get that converted by the end of this calendar year. Okay, thank you, Dom. So that was all the questions I had that was pre-submitted online, but I believe you had some submitted directly to yourself. Yeah, James asks, and I'm just I'm going to read these questions, James asks, what makes Wyoming project interesting and what needs to happen before it can be before it can be built? And I'll sort of maybe just go on the Wyoming slide while I'm talking about it. Just to just to take a step back.

31:02 We know that power, power is a limiting factor in terms of the growth of AI. There just isn't enough power in the world in order to be able to power it all. And so we saw this, you know, a year and a half ago, and we saw this opportunity of land opposite a substation, which is extremely, extremely difficult to get. There aren't many plots of land available opposite substations. We acquired the land and we're now going through a process of powering the land. The first step is getting the appropriate permits from the local city.

31:48 Secondly, is getting the power connected, the power lines connected with the transformers and everything else. And at that point, as I mentioned, the value and this is all in a public domain, we bought the land for 1 million dollars. If we just get a permit, we've already increased the value to 20 to 30 million. We could at that point just sell it for 20 or 30 million. But if we get it powered, that's where it's worth 1 to 1.5 million per megawatt. And We know from prior power studies, there's 200 megawatts available at that substation which can be fed into our particular site. So a powered site right now, if this was powered today, this would be worth on market between $200 million to $300 million US.

32:34 Just to be clear, there's a number of steps that we need to take in order to be able to get there, but that just shows you the magnitude of the particular opportunity, which is why I'm so excited about what we're doing at BlockMate. So that's the question there. Jonathan asked, how will you pay for these projects and what are the biggest risks for shareholders? I mentioned the SPV and how we're going to be running each of those through an SPV, at least in the early days. And the biggest risks are execution. Ultimately, in any business, it really comes down to execution. And in my view, we've put together a team who has been there and who's done it before.

33:21 Personally, I've been involved in technology and building companies similar to this in the past, which I've sold. I just sold my last company, Banksa, for $62 million US. So in terms of track record, there's a team there that's been there and has done that before. In terms of 1 more question, Sally asks, when do you expect the 1st AI compute to arrive and start earning revenue? I've already answered that, but I'll answer that again. We expect our NVIDIA B200 to arrive early Q1. Once we receive it, we expect that within a week, revenue will start to flow. Like it's actually that quick in terms of the turnaround. And I know from my hosting days 20 years ago, that's basically the way it works.

34:12 So they're all the questions that I've got at this particular point in time. Unless there's any further questions that have come through, Jay, I think we can call it a wrap. Yeah. I'm seeing in the chat there aren't any questions, so perhaps we can call it a wrap for this webinar. Dominic, again, thank you very much for presenting and answering all the pre-submitted questions. And yeah, thank you very much to all the attendees and hope you have a great rest of the day.

34:40 Thank you everyone. Looking forward to the next one. Thank you.