Good afternoon, everyone, and welcome to the HIVE Q1 2025 Earnings Call and Webcast. At this time, all participants are in a listen-only mode. [Operator Instructions] And now at this time, I'd like to turn things over to Holly Schoenfeldt, Director of Marketing. Holly, please go ahead.

Hello, everyone, and welcome to today's webcast reviewing HIVE Digital Technologies' financial results for the quarter ended June 30, 2024. On Slide number 2, I would like to briefly note disclosures. Except for statements of historical fact, this presentation contains forward-looking information within the meaning of the applicable Canadian and U.S. securities regulations. These forward-looking statements are based on expectations, estimates, and assumptions as of the date of this presentation. On the next slide, I'm pleased to introduce today's presenters, Frank Holmes, Executive Chairman; Aydin Kilic, President and CEO; and Darcy Daubaras, Chief Financial Officer. On the next slide, I would now like to hand the presentation over to Mr. Frank Holmes for a macro recap of the quarter. Frank?

Thank you, Holly. And it's great to see how dynamic we are and that we're functioning in all these different jurisdictions from Vancouver to speaking for this conference. It will be Aydin actually from the Canaccord, small cap, mid-cap, biggest conference on the East Coast in Boston. I'm in San Antonio, Texas, and we function in nine time zones in Bitcoin mining and running and managing data centers. So I think it's really quite amazing how fluid this is. But as we jump into the next slide, the DNA of volatility is understanding the risk. And I think this is a very important visual for investors to recognize all these external forces that do impact the daily volatility. And one standard deviation for Bitcoin is plus or minus 2% is normal. But over 10 days, it's 8%, which is, you can see, four times greater over 10 days than the S&P 500.\ Nvidia is 9%, and MicroStrategy is 21%, and HIVE Digital Technology is 23%. On a daily basis, it is a non-event for MicroStrategy or HIVE Digital to go up or down 6% in a day. And that's predominantly because of external forces such as Bitcoin prices, geopolitics, geoeconomic events, which recently happened in Japan. But during all of this volatility in the next slide, I like to point out that HIVE is operating in nine time zones, four languages, soon to be five. We have locations in Switzerland and functioning facilities in Sweden, Iceland and Canada. We have offices in Bermuda and in Texas, and I'm calling in from San Antonio. And we are expanding into Paraguay in South America. Next slide. HIVE Digital is proud to be bolting ahead, in particular, sourcing green energy, which has been a big challenge to get sized, like 100 megawatts, which…

Great. Thank you very much, Frank. This part of the presentation, I'll be taking you through a snapshot of the period, as Frank had mentioned, looking at the most recently completed quarter and some financial indicators. We are providing certain non-IFRS measures in our presentation today, and the company believes that these measures, while not a substitute for measures of performance compared in accordance with IFRS, provide investors with an improved ability to evaluate the underlying performance of the company. These measures do not have any standardized meanings prescribed under IFRS and therefore may not be comparable to other issuers. Further details can be found in our management discussion analysis for the three months ended June 30, 2024. Moving on to the next page, I'd like to remind our stakeholders that our earnings are comprised of our operational earnings or cash flow, plus our investment earnings, which includes realized and unrealized earnings, which often includes non-cash charges. Taking a look at the next page, mark-to-market is an accounting practice that involves adjusting the value of an asset to reflect its value as determined by current market conditions. The market value is determined based on what a company would get for the asset if it was sold at that point in time. Mark-to-market losses or gains are paper losses or gains generated through an accounting entry rather than the actual sale of the security. The swings in digital assets impact our paper profits and losses each quarter. So our Bitcoin digital assets do generate unrealized gains and losses each quarter. It is important that investors understand the differences in operating earnings or losses in addition to mark-to-market paper gains and losses each quarter. Speaking about non-cash charges, those are the write-downs or accounting expenses that does not involve cash payment. Items…

Thank you for that excellent summary. It has been a solid quarter and we're going to start off actually by looking at our production of Bitcoin over the last 12 months. So we have mined almost 2,600 Bitcoin with green energy over the last 12 months. As Frank and Darcy both mentioned, this is a first fiscal quarter for all of the public miners where we are post-halving. And as you can see here, we've delivered solid and consistent production since the halving 119 Bitcoin in May, June and 116 in July. And this is a time when we've seen difficulty really maintain high levels in the 80 trillion to 85 trillion range. And how we continue to have consistent production is we had new machines coming in from our S21 and S21 Pro orders to reach our target 5.5 exahash install. Now, I do want to note, and we're going to start talking about this more, we are in a bearish cycle of crypto mining. It's obviously post-halving. We're seeing difficulty actually out of this last epoch, all time high to 90 trillion. And so, what that means is we're in a bear cycle where you really want to manage profitability and have the best in economics. So, we're actually now operating about 5.2 exahash where we've improved the efficiency of some of our older machines, by firmware optimization. And there's a very good reason why we do this. And that's on the next slide. Profit. Another quarter's with positive gross mining margins. As you can see, this is $32 million of revenue this quarter with $11 million of gross mining margins. So add another one to the chart. I think next quarter we should have a slide that shows, Hall of Fame jersey hanging from the rafters. So…

Thank you, Mr. Kilic. [Operator Instructions] We'll go first this afternoon to Mike Colonnese at H.C. Wainwright. Mike, please go ahead.

Hi, good afternoon, guys. And congrats on signing the new site in Paraguay. Great to see that. A couple from me. First, I'm curious, which site will you guys be converting over to HPC from Bitcoin mining? And what will be your strategy for the 20 megawatts? Do you think you'll purchase and deploy your own GPUs as you've done in the past? Or will you focus more on a co-location model here?

Hi Mike, it's Aydin here. I'll take that one. I just want to make sure. Can you hear me?

Yes.

Okay. Excellent. I'm just dialing in from Boston. So the sites we are building for New Brunswick and then one of our sites in Boden. So these are, I mean, all of our facilities are data centers. New Brunswick in particular and Boden. I mean, Boden used to be a GPU mining facility. And our facility in New Brunswick is, what I like to call a military grade data center. So, you know, you've got you've got a structure there. You've got great bones. And of course, all the power distribution from high voltage down to 415 volts is already been done. So it's a matter of adding UPSs, chillers, generators, et cetera. Everything you need to get to Tier 3 uptime. And the nice thing about the retrofit is a lot of the long lead time items are on the high voltage side. And so a Greenfield just in broad strokes might take two years. Retrofit is more than nine to 12-month range. And so the CapEx on a dollar per megawatt basis on a retrofit, you're looking at more in the $5 million to $8 million per megawatt range. Whereas in a Greenfield, you're more in the $10 million to $12 million per megawatt range. So that's that.

Great. Great. Now, appreciate the color, Aydin. And as far as the go-to-market strategy there, do you think you'll fill that capacity up with self-owned GPUs as you've done previously? Or do you think you'll take more of, again, that co-location model approach? And how should we think about the unit economics there?

All right. Yes. So to put it into context, like one, like, using the H100 as an example in broad strokes, about $30 million a megawatt to populate. Right. So it's very CapEx intensive. And the other ratio that's important to consider is that when you're building Tier 3, you have PV, power utilization ratio. And so, if you have 30 megawatts of power, you're going to end up with about 20 megawatts of IT load or compute. Whereas in crypto mining, if you have 30 megawatts of power, you're going to be using maybe 29 megawatts for mining. There'll be very little ancillary overhead. So with 20 megawatts, right, and $30 million, that would be approximately $600 million. Again, in broad strokes, there's obviously Blackwell architecture coming out next year. There's H200s coming out in Q4. I'm just talking about giving a high level understanding of the economics of it. Now, it is, of course, attractive to purchase and operate your own GPU cloud, because that's how you realize those $2.50 per kilowatt hour revenues when you're renting out H100s. On the colo side, we're seeing that rates could be anywhere from $0.25 to $0.50 a kilowatt hour. So there are two different business models akin to mining crypto, where if you offer hosting, margins are thinner, they're more stable. It's more of a long-term ROI. Whereas if you're self-mining, there's more CapEx involved. But you can realize potentially quicker ROIs based on crypto mining economics. So I think it would be a hybrid and we would want to bring the capacity on. We have been generating revenue. We hit $10 million of annualized revenue this quarter from leasing our own GPU. So, we do - plan to expand that. But 20 megawatts is a lot of CapEx to fill up with GPU. So we're evaluating the most secretive path forward right now.

Got it. Thank you for taking my questions, Aydin.

I just wanted to add, it's Frank here, if you can hear me. In Paraguay, we're about an hour and a half to two hours from the capital of driving, Valenzuela. And it's a nice, nice, beautiful green town. And we're so impressed with the infrastructure and the ease of being able to drive there. We took a chopper and visited all the other sites. And a lot of them, are just too far away for our initial. We want to be able to drive, and make it very functional and easy to build.

Got it. Great to hear from you, Frank. And congrats again on the Paraguay site.

Thank you.

Thank you. We take our next question now from Lucas Pipes of B. Riley.

Thank you very much, operator. And good afternoon, everyone. This is [indiscernible] asking questions on behalf of Lucas Pipes. And my first one is on your new third quarter '25 target. How should we think about cadence of deployment here? So is it going to be skewed towards the end of 2025, let's say quarter towards 2025? Or it's fair to assume a gradual ramp up? And where things stays on financing side of this expansion. Just how much do you plan to spend on these hundred megawatt data center this year? And maybe in 2025, if you can just provide additional color here would be great? Thank you.

So, we haven't disclosed the exact capital outlay plan. However, it is extremely attractive on a dollar per megawatt basis in Paraguay. And why we chose Paraguay was threefold. The attractive dollar per megawatt build out costs for us to complete a hundred megawatt site, the low cost of power, and of course, the fact that it's green energy at scale. And so, those three unique conditions are what make Paraguay very exciting. And we will be providing the market more details on capital allocation, and build out costs, et cetera as the project moves forward.

Thank you. And regarding the cadence just, is it fair to assume just gradual ramp up from now to all the way to third quarter of 2025?

Sorry, is it fair to assume what?

How should we think about ramp up of hash rate?

Typically, in the first six months, you're doing high voltage and substation work. And you'll bring on capacity in that second half, perhaps in the third quarter of 2025, is when we would expect hash rate to start coming online.

Got you. Thanks for that. And the second one is on AI cloud progress.

I'd just like to add some color to that. Is that we will do like a New Brunswick to build a campus of buildings and then you start putting in equipment to start as fast as possible to get the cash flow. You don't wait till everything's finished and then turn on get the machines. But like I said, it'll take six months - to nine months. But as soon as a facility of these buildings are built, we will start stocking machines to plug them in.

Got you. Thank you very much. And my second one is on AI cloud. So with respect to the Paraguay expansion, will it affect your plans on purchasing latest gen GPUs for your AI expansion?

So the response I was providing Mike earlier about some great questions. Those are actually the near term AI conversion opportunities, are actually in New Brunswick and Sweden. At existing operating facilities. And so, these have a six to 12-month construction timeline on a retrofit. And so, if we were to purchase, which we've talked about, we've been very carefully going through an engineering procurement exercise for next generation GPU compute, it would go to those locations. And currently 100 megawatt allocation in Paraguay is forecasted, to more than double the Bitcoin mining capacity from 5.5 exahash to 12.1 exahash.

And I think just to add to that, sorry to add, I want to explain about this continuous process that we're spending every month and taking equipment that's for delivery immediately, and upgrading the suite, lowering our cost and improving our exahash. We have a like - and I like the best example is that I'm in South Texas, is like oil and gas and gas when you're drilling and you're fracking, you have to always be drilling. So, we are always upgrading our suite of basic chips. And I think that after we've completed our spin, it'll add with the new more efficient machines Aydin. Is it about one and a half exahash?

Yes, that's right Frank. So we have a, our global fleet average is 5.5 exahash installed with an efficiency of 24.5 joules per terahash. That constitutes, pardon me, approximately 8,030 joule per terahash machines. If we upgrade those three joule per terahash machines to S21 Pros, that would yield an additional exahash and bring our average efficiency down closer to approximately 22 joules a terahash. So that can be the upgrade within our existing facility. And by doing five, we always like, as Frank mentioned earlier, we like to have capital performing. So when we order machines, it's for immediate delivery so they can get plugged in. We have seldom, if ever, do large machine orders where they come in, six months later. We like to have rolling, if not immediate, deliveries of ASICs. So we, we always target a one year ROI. Now, of course, that's variable that stretches out a bit in a bear market, but it also shortens a lot in the bull market. So our whole business model is based on cash flow, return on invested capital. So, we see it as a success if the ASICs we purchase, or GPUs for that matter, when we run them, after you subtract your operating costs, have you made a profit? Have you made a handsome profit in the time that you've been running them? And if you even sell them, sometimes on the secondary market, you can add that back. And one golden example I'd like to point to in the bear market of 2022 is, we bought J-Pros, S19j Pro, for $10 a terahash. And even in that bear market, because we got them for such a great deal, they had fully repaid themselves off in a year and continued to free cash flow after that. So that's sort of the methodology of how we upgrade the site. We don't necessarily do massive orders. We're constantly evaluating the landscape. Yes, we talk to all the big three manufacturers, but we also talk to every broker in the industry. And whenever there's an attractive immediate delivery deal, we pull the trigger. That's why you'll notice we did 7,000 machines, another 1,000, another 1,000, another 500 to prep for the halving in the last six months.

Thank you very much for your perspective. Wish you best of luck in next quarter. Thank you.

Thank you.

Thank you. We go next now to Mike Grondahl of Northland Capital Markets.

Hi guys, thanks. I just wanted to follow-up on the HPC strategy. How have demand been from customers, and what type of customers have you talked to regarding the 20 megawatts?

So the 20 megawatt is first and foremost, an infrastructure project for us. And as we work to build that out, typically larger customers like to see the project underway before you start negotiating pre-leases. So what we've been doing is focused on getting scheduling down. So customers who are wanting compute to come online will want to know when they can expect that compute to come online. And because there are so many new iterations of NVIDIA hardware, the engineering procurement exercise I alluded to earlier, you kind of have to have, it's like matching two puzzle pieces to make sure they're the right fit. So the data center capacity is going to be air cooled, it's going to be liquid cooled. Next generation stuff will likely be liquid cooled, if not all liquid cooled. And then are you going to be working with Blackwell, H200, et cetera? So we'll be able to provide more color on that as we provide more announcements. But that's just the methodology of the process that you go through when you're doing these larger build outs. I hope that's helpful.

Yes, sure. And then just secondly, what are your top three priorities for the second half of calendar '24 now?

So we hit our $10 million ARR for AI revenue this quarter, which I'm very proud of our team for realizing that milestone, which we broadcast previously. Our target for H2 this year is $20 million of ARR. So that is one target. And that's working with bringing more GPUs in our cloud business online. The other target is to have both Paraguay construction underway and the 30 megawatt conversion into 20 megawatts of HPC underway. So the completion of the latter two infrastructure projects will fall into 2025, of course. But to have those projects broken ground would be a goal that we would be very pleased with this year.

That's great. Thank you.

Great question. Thank you for asking.

I was just like that too. Oh, sorry. I was just trying to it's Frank here to add some color to what Aydin is saying, is that when people make these announcements from crypto mining, all of a sudden they're going to HPC. It's very, very complex. And it's much more complex than I ever expected. But what Aydin is saying is that we've hit $10 million. That's basically the same team that's been building that out and repurposing those chips. So that's a heck of an accomplishment, because that's almost like free cash flow what it throws off. And what we're finding is that when we go to order black wells, et cetera, the costs, the engineering is much different. It's much. You're building a brain. And it's like a bunch of neurosurgeons are brought into the room. The engineers at the beginning just don't believe how much energy is going to be consumed. How heavy is the steel infrastructure to hold up these black well servers? There's a litany of complexity that goes into these expansions. So one of the things which we said is that we have six megawatts of electricity and Boden. We own the land. We own the facility. Let's start. It's a military town. Let's start right away there. And then we said, hi, in New Brunswick, we have this additional electricity that we have ability to. We have variable and we have fixed. And let's take a portion of the fixed. And as we're in that journey of getting the engineering, et cetera, and we're discussing with Nvidia these various chips, all of a sudden it becomes that we're in conversation with Nvidia engineers that we have to rethink the structure, the building. Do you need reinforcement? Do you need this or that? So it is phenomenally fascinating, but it's much more challenging than what people think. But what's sweet about it is that you can be making $4 an hour versus $0.15 Bitcoin mining. That's the real big difference when you look at the business.

Thank you. We'll go next now to Bill Papanastasiou at Stifel.

Thank you. Good evening, gentlemen. Thanks for taking my questions. For my first one, I was hoping you'd be able to share your outlook for the Bitcoin mining landscape, and how that plays into your philosophy, to scale operations going forward, as you look to double the hash rate capacity here with Paraguay?

Well, we have always tried to utilize cash flow from the business to grow the business and, of course, adhere to the green energy mandate, which makes it a lot tougher. I mean, we could have - there was a lot of opportunities we've looked at. We could have gone nuclear. We could have just gone grid mix. And even within the realm of renewable energy, we did look at some interesting opportunities in the U.S. where you could have gotten $0.025 power. But then you're spending $2 million a megawatt getting on the energy production side, contributing to the CapEx and the solar farm. When you look at those ROI, they really start to stretch out very long and - it doesn't make economic sense. And so getting back to the three pillars of why I think Paraguay is such a home run, is because you have the cheap dollar per megawatt CapEx to build out. You have the attractive electricity costs. We'll provide more details on these specifics in due course. And then, of course, the green energy at scale. Now, of course, to do a hundred megawatt site, that will require capital. But this is what we believe a truly accretive use of capital. So we're excited about this project.

Appreciate the color there. And then just shifting gears to the GPU as a service business. How are you seeing or assessing the sustainability of prevailing market compute prices for the GPU hardware that HIVE has in the fleet over the next 12 to 24 months? And can you share some more color on utilization rates? Thank you.

Yes, that's a good question. Like with all technology, whether it's your home computer, your iPhone, ASIC miner for Bitcoin or an NVIDIA GP, they have economic life cycles. Right. And really, the A100, right the [Ampere] series was still in use and is still in use, and is still considered very viable. And the H100 came out, the Hopper architecture. And so, what happened was people were paying a premium for the Hopper architecture, but they would still pay a lower amount to get compute using the A100. When the Blackwell comes out, we'll have the same effect. So I would say if you use a melting ice cube analogy, the hourly rate for GPUs melts at a much slower rate than Bitcoin mining ASICs, right. And I think that, a big part of being able to be a true player in this space that, we haven't really touched on yet on this call. So your question is a good one, because it brings it to light is uptime, right. So in crypto mining, when you look at the Bitcoin per exahash, HIVE is usually in the top three consistently, not usually, but consistently. And we're talking 98%, 99% uptime. That's a home run in crypto mining. And then, some of the peers are like 95%, 90% uptime. There's no room for 90% uptime in AI, right. Like, Tier 3 is 99.998% uptime. So I think what the aspiring crypto miners that want to get into this sector. Will wrestle with is delivering that uptime for people that want to train foundational models that do fine tuning. And we've done a lot of that. I mean, we serve 120,000 GPUs in the East mining era. And we've grown our AI compute revenue, from $1 million to $10 million ARR in the last year and a half. So, we've been building that - building on our pedigree of operating GPUs. And we do a ton of stuff on the R&D side, looking fine tuning data sets, et cetera. So making sure that when we work with researchers that, they have a good keyword is user experience, because you do have a user, right? You're not just submitting hashes, to the Bitcoin network. And if your server gets disconnected, reboots and then it tries again. So I think the key is good uptime and a good user experience. Will allow you to realize those long-term contracts, because once you sign a contract for a year or two, what have you, you've got to deliver. There's SLA service level - performance requirements. And we're seeing long term contracts for H100s roughly in the $2.50 per hour and some $2.20. It's all a negotiation, right. So I hope that answers the question, Bill.

I appreciate the color.

And gentlemen, it appears we have no further questions today. Mr. Kilic, I'd like to turn things back to you, sir, for any closing comments.

Yes, I just want to thank my whole team. Frank pointed out we function across nine time zones. We are the longest standing crypto miner. We used to mine Ethereum. That business did $150,000 a day. It was a 90% profit marching business, and we never got a premium for it. Instead, we always contended with the FUD, fear, uncertainly, a doubt that the merge was coming. And then when the merge came, we pivoted, and we started mining all coins with the GPUs. And then, we started reversing some GPUs for HPC compute. This is now HIVE's second Bitcoin having event. And as Darcy pointed out, we've got $150 million of green and clean Bitcoin on the balance sheet, unlevered, unencumbered. We've not taken on any debt to purchase ASICs. So I think that we've really emerged as a best-in-class performer. But when you look at the comp tables, the enterprise value to EBITDA, we're trading at a discount. So for all the analysts that are here, thank you so much for dialing in, and asking these excellent questions. And I think that we really just focus on pound-for-pound, being the most profitable, best in economics. We realize scale is important. That's why I've been telling people we know the market doesn't care about five to six exahash. That's why we said, okay we need to go from five to 10 to 15. So stay tuned. And I think it's going to be an exciting year ahead. We always see that bull market come about nine to 12 months after the halving, if history is any indication, which it usually is. Thank you, everybody.

Thank you, Mr. Kilic. Ladies and gentlemen, that will conclude the HIVE Q1 earnings conference call. Again, thanks so much for joining us, everyone. And we wish you all a great evening. Goodbye.