Earnings Call Transcript - IREN Q2 2025

Operator, Operator

Thank you for standing by, and welcome to the IREN Second Quarter Fiscal Year 2025 Results Conference Call. At this time, all participants are in listen-only mode. After the speakers' presentation, there will be a question-and-answer session. As a reminder, today's program is being recorded. And now I'd like to introduce your host for today's program, Lincoln Tan, Director of Investor Relations. Please go ahead, sir.

Lincoln Tan, Director of Investor Relations

Good afternoon to those joining us from North America, and good morning to our listeners in Australia. Welcome to IREN's second quarter FY '25 results presentation. My name is Lincoln Tan, Director of Investor Relations, and I'm pleased to be joined today by Daniel Roberts, Co-Founder and Co-CEO; Belinda Nucifora, CFO; and Kent Draper, Chief Commercial Officer. Before we begin, please note that this call is being webcast live accompanied by a presentation. I'd also like to remind you that some statements made during this call may constitute forward-looking statements. These are based on certain assumptions and are subject to risks that could cause actual results to differ materially from our expectations. Listeners should not place undue reliance on these statements. Please refer to the disclaimer on Slide 2 of the accompanying presentation for further details. With that, let's get started. Over to you, Dan.

Daniel Roberts, Co-Founder and Co-CEO

Thanks, Lincoln. Good afternoon, everyone, and thank you for dialing in to our Q2 FY '25 results presentation. So straight into the highlights, a record result in Q2, underpinned by some pretty rapid growth over the last 6 months and low-cost Bitcoin production. So the investments we've made in scale and efficiency over the past 12 months, as you can see, are starting to flow through to our earnings, recording $18.9 million of NPAT last quarter. We expect this continued earnings momentum in tandem with that growth profile, noting that the 31 exahash of energized capacity was only achieved at the end of the second quarter, and you can see that the average operating hash rate for the prior quarter was substantially lower than that. So as we continue ramping up to 50 exahash by the middle of this year, we expect to see those numbers continue to grow in parallel. Final point on this is to note that, yes, Everi's miners strategy and financials are slightly different, but these financials are not impacted by any Bitcoin revaluations. We don't hold Bitcoin on our balance sheet, so there are no fluctuations in the value of those up, down, or sideways. Consequently, our results reflect the pure profitability of our underlying operations. On the growth front, three key initiatives, which we'd like to highlight today. Firstly, we're excited to announce Horizon 1. Horizon 1 is our new 75-megawatt liquid-cooled AI data center to be delivered at Childress this year. 75 megawatts of capacity translates to around 50 megawatts of IT load. And what we've seen, particularly coming into the start of this year, it's increasingly clear that there is a strong market opportunity for us to capitalize on the shortage of liquid-cooled data center capacity. As many of you would know, the liquid-cooled Blackwell GPUs from NVIDIA are in the process of coming to market and the required rack densities are continuing to escalate with this generation and subsequent generations of GPUs. So we are future-proofing this deployment. We're designing and building for densities of up to 200 kilowatts per rack, well beyond the 120, 130-kilowatt rack density required just for the Blackwell GPUs. So we're excited. There appears to be a real scarcity of liquid-cooled data center capacity, and we feel that we're in a relatively unique market position to deliver this type of capacity in a relatively short time frame. Secondly, we're excited to announce the development of Sweetwater 2, a new 600-megawatt site that will further strengthen our position in West Texas, situated around 28 miles from the 1.4 gigawatt Sweetwater 1 project and about 39 miles from Abilene. This expansion would create a 2-gigawatt data center hub at Sweetwater with design work already underway for a direct fiber loop between the two sites. We've secured over 500 acres of land and we're in the process of finalizing a 600-megawatt grid connection agreement with a target energization date of around 2028. Now one important caveat: we always talk about ones and zeros with grid connection agreements. We do not have one yet. So at the moment, this is a hypothetical project. However, we've chosen to disclose the details today for two key reasons. One, this project is in late-stage development, and we're getting closer to the execution of binding agreements. But secondly, the 2-gigawatt data center hub it potentially creates with Sweetwater 1 is material to the market opportunity we are seeing in West Texas. So while this is still a project under development, Sweetwater 2 would provide yet another organic growth pathway for our AI and mining business. However, more importantly, it expands upon the potential value of the single site AI opportunity at Sweetwater, something that we'll come to a little bit further on in the presentation. Finally, let's not forget about the rapid expansion of our Bitcoin mining business. Our operations from 31 exahash today to 52 exahash this year. We're dedicating 75 megawatts of Childress Phase 6 to Horizon 1. And as a result, this adjusts our prior expansion plans from Bitcoin mining from 57 down to 52 exahash. But to reiterate, we remain on track to hit that 50 exahash target by the middle of this year. The 500-plus site team at Childress continues to be well positioned to deliver 50 megawatts every month of data centers. In terms of the corporate and funding update, we're pleased to confirm that U.S. domestic issuer status and U.S. GAAP reporting will commence from the second half of 2025. Our balance sheet remains robust. We closed our oversubscribed convertible note at the end of 2024 and recently put in place a $1 billion ATM facility to provide flexibility to fund growth initiatives. However, market conditions, as we've all seen, are very dynamic, and we are always looking for ways to optimize our capital structure, and we are continuing to evaluate alternative funding options. Finally, in terms of investor distributions, as you can see from this result, there is significant operating free cash flow being generated. However, from a capital management perspective, we've made a decision to accelerate our investment into the strategic initiatives that I've outlined, and as a result, are deferring the consideration of potential investor distributions. So into a little bit more detail about Horizon 1. As mentioned, we are planning to deliver a 75-megawatt liquid-cooled AI data center within Childress Phase 6, targeting completion in the second half of this year. That will support a 50-megawatt IT load and include the installation of backup generators and UPS systems for power redundancy. Building this as part of Childress Phase 6, you can see on the map in the blue overlay. It is designed to support 200-kilowatt rack density, well beyond what's required for the NVIDIA Blackwell GPUs and consequently with an eye to be able to accommodate future increases in rack density architecture. We've been working with a leading global EPC firm to deliver this project and leveraging our existing data center footprint with an estimated build cost of $6 million to $7 million per IT megawatt. So why? Why are we doing this? Well, liquid cooling is a key priority for large-scale customers. And what we've observed is a very limited availability of liquid-cooled data center capacity to come online in 2025. Meanwhile, the liquid-cooled NVIDIA Blackwell GPUs are expected to ship in more meaningful quantities in the second half of this year. IREN is uniquely positioned to address this growing market demand, leveraging our immediate access to power and our existing infrastructure on site. 50 megawatts of IT load is a size that appeals to both smaller as well as larger customers and also caters to NVIDIA's reference architecture, allowing a single cluster of up to 16,000 Blackwell GPUs. We have a unique opportunity to move fast in a market that is changing quickly. Horizon 1 will enhance delivery certainty for customers, helping them cut through the chicken and egg dilemma by showing them more concrete plans around design and delivery. Initially, we're focused on Childress and building out a multi-tenant site there. We're engaged in meaningful two-way dialogue with a variety of prospective customers, noting that, again, Childress will be prioritized for multi-tenant colocation in respective AI data centers, preserving the Sweetwater data center hub and Sweetwater 1 site for more single-site opportunities, which I'll come to later in the presentation. So uniquely, we have this opportunity to continue to scale this approach across our entire portfolio, something which we are actively looking at. Noting, of course, that the critical piece, access to the grid connection, the power, and the land is already secured. With grid connections already secured, we're seeing it's four to seven years to bring these projects from origination into the real world. The data center buildings are already in place. All we need to do is continue to install liquid cooling and power redundancy to scale up this side of the business into the market demand. So the 2-gigawatt Sweetwater data center hub. We're very excited to announce our new 600-megawatt Sweetwater 2 development. We're in the process of finalizing the grid connection agreement. As mentioned earlier, it's situated on over 500 acres of land, 28 miles from Sweetwater and approximately 39 miles from Abilene. Design work has already commenced on a direct fiber loop between the two data center campuses to support network connectivity and create this 2-gigawatt data center hub. We expect it to energize around 2028. And again, I'll caveat this; it's not signed yet, so there's absolutely still risk around this development. But the development process has been in the works for some time. We feel it's getting closer to a potential outcome. Importantly, in the context of our strategy, how we're thinking about Sweetwater 1 and the single site, this 2-gigawatt data center hub it potentially creates with Sweetwater 1 is material to the market opportunity we are seeing in West Texas. In terms of the existing 1.4-gigawatt Sweetwater 1 project, it's full steam ahead towards energization in 14 months. Procurement is well underway to support energization of the full 1.4-gigawatt bulk substation and additional primary substations all by April next year. As mentioned earlier, we're working with a leading EPC contractor with experience on hyperscale deployments. General site and civil works to support construction are commencing shortly. Clearly, with the recent market announcements, there's an increasing level of focus and interest on West Texas as a data center market. Accordingly, this is a very exciting opportunity for IREN to build one of the largest data center sites and maybe hubs in the world. I'll touch a little bit more on how we're thinking about this site strategically later on in the presentation. Large-scale, low-cost Bitcoin mining. So two key pieces I'd like to reiterate today. One is the scale of our operations, having 31 exahash installed. It was on time as we said it would be. We're on track for 50 exahash in five months and then 52 exahash shortly thereafter. This delivers economies of scale today and will continue as we expand. We have a demonstrated track record around operations, our uptime, and ability to deliver projects on time. The second point I'd like to make is that we are a low-cost producer. It supports strong unit economics and ultimately, earnings for shareholders. Since announcing our plans in May last year to expand, all the investments we have been making are starting to flow through to operating efficiency and earnings. We have best-in-class efficiency at 15 joules per terahash. Our spot pricing strategy has delivered actual power prices of $0.03 a kilowatt-hour, where the majority of our operations are today at Childress. Finally, you've seen this start to show in our quarterly earnings, but also in our monthly updates. Our hardware profit margin is over 75% now as we've scaled over that 31 exahash mark. So the expansion to 52 exahash. As we've discussed, allocating 75 megawatts of Childress Phase 6 to support Horizon 1 adjusts our prior mining expansion plan from 57 to 52 exahash. To reiterate again, our 50 exahash target remains on track to be delivered in full in the next five months. Continuing to build 50 megawatts per month of data centers at Childress. The efficiencies that we have with large single-site expansions, 500 people on the site effectively copying and pasting every single building. As you can see on the right-hand side, the schematic diagram reflects our data center construction status. To the extent you're interested, you can continue to track our progress in our monthly updates. Our mining economics remain strong. We've seen the material step-up in earnings and cash flow generation as we scale up. And on the left-hand side of the page, you can see some illustrative economics that reflect where we are today and where we expect to go. So the escalating power shortage seems to be the only word we can use given what we're seeing in the market. Clearly, it's been a very interesting couple of weeks with Stargate and DeepSeek. However, if we look through some of this volatility, nothing has changed for us. If anything, we've seen an uptick in observed interest in cloud and colocation services. AI doesn't live in a vacuum. Humans crave more and more, particularly if it gets cheaper over time. And I think the Wikipedia page for Jevons Paradox has never seen so many hits over the last two weeks. This slide illustrates two market dynamics that are relatively well-documented: the first is hyperscale CapEx continuously being revised upwards. The second is the power scarcity dynamic looking at a 36-gigawatt shortage in data center capacity across the U.S. With over 2 gigawatts of secured grid-connected power and land, this creates a compelling opportunity for IREN. This two gigawatts, 2.3 to be precise, is not hypothetical potential megawatts. This is power that is either flowing today or is secured by binding interconnection agreements with utilities to be delivered by April next year. This slide essentially summarizes the macro backdrop of the opportunity for us here at IREN. Picks and shovels for the digital age. So just to change tack for a moment and share a bit about our history, why we are in this position, and ultimately set the scene for my final slide where I'll discuss our strategic focus. Will and I have been setting this platform and gearing up for this market backdrop since day one. There's been no pivot. This has been a deliberate strategy. It can be traced back to our earliest investor presentation in 2020, a vision that has remained unchanged. The value then and now is in access to large-scale renewable energy required to power the future of supercomputing, whether that be Bitcoin mining or AI, two super trends that are front of mind right now, as well as whatever comes tomorrow. We have delivered on our strategy to bootstrap a large-scale and profitable business with Bitcoin mining while continuing to leverage that asset base into new, higher-value, and increasingly interesting use cases such as AI. Our team's depth of experience in energy infrastructure and data centers has allowed us to organically develop a leading data center platform from the ground up. Our ability to continue to find and incubate greenfield opportunities and then execute upon them is a very important differentiating factor and an enormous opportunity for us in this power-constrained and constantly evolving market. Finally, we've covered a lot over the past 20 minutes, so to try to tie it all together and highlight how Will and I are thinking about our strategic priorities. The market context, as we've covered, presents significant opportunities for IREN. The escalating demand for power and liquid data center capacity along with the increasing institutional adoption of Bitcoin. If we start with our existing data centers in British Columbia, they continue to mine Bitcoin profitably and generate positive free cash flow. But how are we thinking strategically about British Columbia? Well, firstly, we continue to grow our AI cloud services business in a sensible way, currently operating side-by-side with Bitcoin mining. As many of you would have seen, in our Prince George data center, we have Bitcoin mining ASICs securing the Bitcoin network operating right next to the latest generation NVIDIA H100 and H200 GPUs, providing cloud services to AI clients. This is something that we will continue to explore, and it's been pleasing to see even an uptick in observed demand post DeepSeek. In addition to that, we're continuing to explore AI colocation opportunities at these sites. Again, that strategy of bootstrapping sites with Bitcoin mining and then, as higher and better value use cases come along, looking to insert them into our infrastructure base. At Childress, it is fair to say it's a very exciting data center campus today with a hive of activity. We're growing Bitcoin mining capacity to 52 exahash this year while continuing to generate significant operating cash flows and profitability from these operations. However, we are also really excited about delivering Horizon 1, our 75-megawatt liquid-cooled data center to be brought online this year. The way we think about Childress strategically is as follows: Firstly, once again, it is underwritten by Bitcoin mining, as was the plan five years ago, and as is the plan going forward, then utilize the site for higher-value use cases as they emerge. We believe liquid cooling AI data centers is likely to be one of those and are accelerating our investment into this space. Importantly, from a strategic standpoint, we view Childress as a focus for multi-tenant colocation, preserving Sweetwater, which I'll come to in a second, for a potential single-site deal and tenant. Sweetwater, the Sweetwater data center hub. We have terminated the site sale process with Morgan Stanley. We're now working with a range of different advisers, brokers, and partners on a broader range of opportunities where almost all of those opportunities now involve us retaining long-term ownership of the site. Given the scale of the site and the market backdrop of hyperscalers chasing 1 gigawatt plus single-site campuses, we are prioritizing Sweetwater 1 as well as the emerging Sweetwater data center hub for a whole site single-tenant colocation opportunity. Sweetwater 2 and the potential creation of this 2-gigawatt data center hub in West Texas only goes to further strengthen the strategic opportunity in this area. In the meantime, as we continue to explore those conversations, we will retain flexibility to once again bootstrap this site with Bitcoin mining, as we've done with the rest of our portfolio since day one. Belinda, over to you to cover the financials. Thank you.

Belinda Nucifora, CFO

Thank you, Dan. So good afternoon to those in Sydney - sorry, actually, it's morning here. So good morning to those in Sydney. It's actually a beautiful day, and good afternoon to those in North America. Thank you for joining us for our Q2 FY '25 earnings update. As Dan earlier mentioned, this was a record result. We delivered Bitcoin mining revenue of $113.5 million and operating cash flows of $53.7 million, and overall net profit after tax of $18.9 million. Adjusted EBITDA for the quarter increased by $60 million to $62.6 million, with the average operating hashrate increasing from 12.2 exahash to 22.6 exahash, and we mined 1,347 Bitcoin at an average realized price of $84.3k. So I just noticed my camera may be out, so I'll try and adjust that if possible. Okay. I think that may be back on. Okay. Net electricity costs remained relatively flat for the quarter at $28.9 million, with the increased megawatt usage at Childress offset by lower cost per megawatt due to the transition to a spot pricing strategy in August 2024. As such, the average net electricity cost per Bitcoin mined decreased from $35.4k versus $21.4k. Other costs of $25.1 million increased by $3.7 million, primarily due to the additional purchase of renewable energy certificates at Childress for the increased megawatt usage as well as construction insurance costs at this site. The cost base reflects a business today that is delivering significant growth and projecting continued expansion over the coming years. So moving on to cash flows. Closing cash at bank at 31st of December 2024 was $427.3 million, with receipts from Bitcoin mining activities of $113.6 million and AI cloud services of $3.5 million. Increase in electricity payments of negative $3.5 million reflects continued expansion at Childress with commission capacity increasing from 200 megawatts to 350 megawatts during the quarter. A decrease in net cash used in investing activities of $217 million was due to a decrease in mining and hardware, primarily due to significant milestone payments made in the previous quarter. There was an increase in net cash from financing activities of $372.3 million, with $311.7 million net proceeds from the convertible notes and a $63.7 million increase in net ATM proceeds during the quarter. Since the balance date, a further $50.4 million of net ATM proceeds have been received, and the total current number of ordinary outstanding shares is approximately $219 million. Now moving on to the balance sheet. During the quarter, total assets increased by approximately $500 million to $1.9 billion as of 31 December 2024, providing a strong balance sheet to support our future growth opportunities. On 6th of December 2024, we issued a $440 million convertible note with an annual interest rate of 3.25%, which is due to mature on June 15, 2023, unless earlier purchased, redeemed, or converted. Concurrently to the convertible note, we entered into a cap call transaction of $44.9 million and a prepaid forward of $73.7 million. As IREN currently reports under International Financial Reporting Standards, the convertible note embedded derivatives and financial assets have been brought to account at their fair value at inception and revalued at the reporting date at fair value through profit and loss. As IREN transitions to U.S. GAAP reporting from 1 July 2025, the accounting for the convertible notes will be reassessed, as well as the cap call transaction and prepaid forward in line with the applicable U.S. GAAP standards. Total equity increased to $1.3 billion with the sale of 25.3 billion shares during the quarter ending December 31, 2024. I think now we're turning back to Lincoln for the start of Q&A.

Lincoln Tan, Director of Investor Relations

And we'll go with our first question. Our first question comes from the line of Joseph Vafi from Canaccord. Your question, please.

Joseph Vafi, Analyst

Hey, guys. Good afternoon to those in the U.S. and good morning to you. Really great update, a lot of exciting things. Maybe we just start here on the update on Horizon 1. It sounds like a pretty exciting project, energizing or being ready later this year. I was wondering if you could kind of talk about CapEx and where that stands at Horizon 1 and kind of what you've learned, especially with this rack density that you're building there? And then I have a quick follow-up.

Daniel Roberts, Co-Founder and Co-CEO

Kent, would you like to take this one?

Kent Draper, Chief Commercial Officer

Yeah. Happy to jump in there. So we had some guidance that Dan mentioned within the presentation itself around the CapEx levels that we're seeing. Importantly, on our side, we are able to utilize a lot of the existing data center architecture, which means that we believe we're able to deliver this liquid-cooled capacity at a very effective cost per megawatt, but still with all of the key redundancy and other features that ultimate end-users of that capacity would expect to see. We've also noted, and we've discussed it in prior calls as well, that we have already been ordering many of the long lead items associated with that build-out as well. So as we do with all of our build-out, whether it's on the Bitcoin mining side or on the AI data center side, we are making sure that we are well ahead of the curve in terms of ordering those long lead items.

Joseph Vafi, Analyst

Great. That's great color, and thanks for that update, Ken. Maybe we talk a little bit on Sweetwater. Clearly, a very ambitious project with the update that we heard here about adding Sweetwater 2 to Sweetwater 1 and the fiber loop interconnect. And I know, Dan, you also mentioned that the agreement you had with Morgan Stanley is now kind of moving to the next phase where you're in different kinds of discussions. 2 gigawatts, quite an ambitious big data center, which would probably be used for AI. And just - I know you've had probably a lot of discussions with a lot of players in the space. At a high level, this would be like massive. And so just trying to get a feel for demand out there amongst hyperscalers or other players around 2 gigawatts and what you're hearing kind of at a high level. Thanks a lot.

Daniel Roberts, Co-Founder and Co-CEO

Yeah. No, thanks, Joe. Look, we were called far less polite terms than ambitious six years ago when we first suggested that the future of data centers may not all be in metropolitan areas because maybe there wouldn't be enough power there. So I think you fast forward to today, suggesting that a 2-gigawatt data center hub may be viable in the context of the market backdrop, all the announcements that have been made, including people with objectives far greater than 2 gigawatts. I don't think it's ambitious. I think it's entirely realistic and reflective of where the market is today. We have been surprised coming into 2025, just over that Christmas holiday period, at the step-up in intensity in this sector. I can't think of many, if any, hyperscalers that we're not talking to that aren't interested in 1-gigawatt plus campuses that can be delivered in the next few years. The market absolutely seems to be there. We're in a lot of active conversations. But again, I'm going to temper all of this because ones and zeros. You either do a deal or you don't. We've got the ultimate backstop of building out bootstrapping with Bitcoin mining and preserving that single site opportunity where the value of having clusters all contained geographically on one campus seems to be quite powerful from these hyperscaler perspectives. So it's interesting. I wouldn't call it ambitious. I can call it entirely realistic. Yes, there's risk around Sweetwater 2 and getting the final signatures and executing that and turning it into a real project rather than hypothetical, but we're working hard on that. And yeah, I'm excited.

Joseph Vafi, Analyst

That's great. That's a great update. It's a great path forward. And thanks for the update.

Lincoln Tan, Director of Investor Relations

Thank you. And our next question comes from the line of Greg Lewis from BTIG. Your question, please.

Greg Lewis, Analyst

Yeah. Hey, thank you. And good afternoon or good morning. And thanks for taking my questions. I guess my first one, Daniel, was around the update at Sweetwater with the incremental 600 megawatts. A couple of questions around that. And I guess the first one is you're looking to potentially reenergize in 2028 - or energize in 2028. When did we have to start getting in the queue to even be able to try to get on that track to 2020, realizing we don't have the approvals yet and there's still work to do. But I'm just kind of curious as you see that to even be able to kind of talk about the 600 megawatts? And then kind of has that changed, i.e., if we were to kind of get in the queue today, what would that look like?

Daniel Roberts, Co-Founder and Co-CEO

1,000%. It's actually extraordinary how hard this is. To give you some context, we received a draft connection agreement for that site following completion of the studies in July last year. So where are we now? February? How many months is that trying to get signatures on the document? It is so different than how it was 12, 18 months ago. I don't know how all these megawatts are going to be developed going forward. We've obviously got our multi-gigawatt development pipeline, including sites in Texas as well as globally and broader in North America. The congestion, the difficulty in actually getting these projects over the line and crossing that zero and one barrier to having a project that you can actually build out on, it's actually extraordinary the way we're seeing it at the moment. So we'll continue to push. We almost didn't disclose Sweetwater 2 today, but we are getting close. We feel like we're getting close and it is material to how we think about Sweetwater 1 in the context of the broader market dynamics we are seeing in West Texas. But absolutely, there's risk around any megawatts that aren't contracted, and we are living and breathing that every single day. I don't know if Kent has anything he'd like to add to that?

Kent Draper, Chief Commercial Officer

Yeah. The only other thing I would add, as Dan mentioned, the process has taken a long time for that project, and those were connection requests that we submitted over a year ago now. For projects that have been submitted in more recent months, we know that there has been a massive increase in the number of applications to utilities and to ERCOT. Those utilities in ERCOT haven't substantially expanded their teams and the number of people that are looking at those connection requests. So anything submitted in the very recent months is going to take incrementally longer than what we saw for that Sweetwater 2 site. So it is becoming increasingly difficult.

Greg Lewis, Analyst

Okay. Great. Thank you for that. And my other one was around just kind of looking for color and it's realized things are evolving on a daily basis. But you mentioned that $6 million to $7 million per megawatt. It's our understanding that a lot of the grid and the power equipment that's going to be needed will be sourced from Mexico or definitely inside NAFTA. And so as we think about that $6 million to $7 million, is there any way to kind of frame out how much of it's imported without getting into specific price numbers, maybe how much of it is imported? And yeah, I guess that's kind of curious how that could be changing at least over the next few quarters.

Kent Draper, Chief Commercial Officer

Yeah, I'm happy to jump in there, Dan. So yeah, it is obviously a very dynamic environment at the moment, changing on an almost daily basis. Yes, it's something that we continue to monitor going forward. There is the potential for higher construction costs. We do have a very diversified supply base. So not everything today is coming in from the countries that have been announced as potential targets for increased tariffs. So we do feel like we're in a good position to handle it. And of course, anything that gets implemented is likely to hit all other providers as well. I think it would be felt across the industry rather than something specific to us. But yeah, we have been working on the diversification of our supply base for a long time to make sure that we always have alternate providers in place. While we didn't envision these specific tariffs, things like that were very much with multiple different scenarios in mind.

Greg Lewis, Analyst

Okay. Thank you very much for the time. Have a great day.

Lincoln Tan, Director of Investor Relations

Thank you. And our next question comes from the line of Darren Aftahi from ROTH. Your question, please.

Darren Aftahi, Analyst

Yeah. Hey guys. Thanks for taking my questions. Nice job on the progress. Just two, if I may. I know in the past, you guys have always talked about looking at ROIs from the lens of the cost to build an exahash with Bitcoin and the ascribed value the market gives you. I'm just sort of curious what calculus you went through with Horizon 1 in terms of reducing the 5 exahash and deciding to build the 75-megawatt HPC endeavor. Was that more of just always planned and now you're formally announcing it? Or did something change in what you're seeing?

Daniel Roberts, Co-Founder and Co-CEO

Yeah. Hi, Darren. I appreciate all your support. Look, we deliberately didn't put numbers in the presentation because none of us know. This is unique; liquid-cooled data center capacity just doesn't exist at scale, and there's no real market for it. What we do know and can see is traditional colocation rates for capacity that isn't necessarily liquid-cooled for 75 megawatts. I think when you run those numbers against the backdrop of CapEx of $6 million to $7 million, the numbers potentially look quite compelling. But we're not here to estimate where we'll get to. We're in a series of very different customer conversations around utilizing that capacity. I think the decision goes more broadly, and it's more strategic. It's saying if you step back, you've got Bitcoin; we know where that's at. It's a great business, but we are so uniquely positioned from a strategic perspective to capitalize on this AI thematic and build out something that is relatively unique this year. Obviously, if AI fizzles out and becomes nothing, then the facility, we might have to repurpose it for Bitcoin mining or another use case. But I think the conviction we're seeing in AI at a macro level and the lack of ability of the current market and those mini micro data centers in metropolitan areas to service these types of capacities, it just seems a really obvious opportunity for us where I would expect we will generate strong returns, not just from the 75 megawatts but what it unlocks strategically for our platform and how it helps us in the conversations we're having around Sweetwater and potential monetization pathways there. Because all of a sudden, we've now got a liquid-cooled AI data center design signed up by global engineering firms that we're building; it creates - it's real. It's happening. So I think the strategic value to the platform beyond the discrete economics from 75 megawatts of Horizon 1 is incredibly powerful.

Kent Draper, Chief Commercial Officer

I think in addition to that, Darren, what we've really seen is the demand side of things crystallizing over recent times. Historically, obviously, Blackwells were announced, and everybody had an eye towards liquid-cooled data centers. What we're seeing now is demand from a number of different customer segments, in particular, customers that have already placed orders. They absolutely need liquid-cooled capacity in order to house those GPUs. We're seeing, as I said, a range coming from various customer types, whether they're large hyperscalers, the Neo clouds, or enterprise customers as well. That has really helped crystallize our decision-making; the time is right to make that investment in this capacity.

Darren Aftahi, Analyst

That's helpful. Just one more, if I may. I know you mentioned there's some risk to getting it across the goal line. But what's the strategic benefit of you announcing Sweetwater 2? Is this more customer-focused? Or is there something else you keep in mind indulging us? Thanks.

Daniel Roberts, Co-Founder and Co-CEO

Yeah. Look, it is very close, but I also think it's material for our investors to be aware that this project exists because it provides context to our strategy with Sweetwater 1 and the single site opportunity. It is material in the context of the broader market thematic and what's happening with the chase for multi-gigawatt data center campuses, particularly in West Texas to build out these AI clusters. So yes, ordinarily, we don't disclose development details because there is risk. But given it is late stage, given its materiality connected to the rest of our business, we thought it appropriate to bring investors into the fold and share a little bit more about this specific site.

Darren Aftahi, Analyst

Thanks, guys. Best of luck.

Lincoln Tan, Director of Investor Relations

Thank you. And our next question comes from the line of Brett Knoblauch from Cantor Fitzgerald. Your question, please.

Brett Knoblauch, Analyst

Hi, guys. Thanks for taking my question, and congrats on the print. I guess the whole sector kind of sold off following DeepSeek. I think in your prepared remarks, you talked about maybe interest or demand picking up post that. Does that factor into maybe why you guys are announcing Horizon 1 after maybe just only recently kind of upping your hash guidance from 52 to 57 by the end of the year?

Kent Draper, Chief Commercial Officer

Yeah. I think as Dan mentioned earlier, that's right. And we've seen it both on the cloud and the colocation side. So it's not just one area that it's limited to. But yes, it absolutely does factor into the thinking. As I mentioned, we have seen a large uptick in demand for liquid-cooled capacity, and the timelines that people want that demand on are something that we feel very well prepared to deliver upon, and we have a competitive advantage because we don't think there are many others that will be able to deliver liquid-cooled capacity on that same timeline. Particularly in 2025, but even into the early parts of 2026, we are seeing a shortage of supply going out that far.

Brett Knoblauch, Analyst

Perfect. That's helpful. And then is one way you think of Horizon as it's almost like a model home showcasing who would be the big tenant at Sweetwater, like what you can build and design and kind of giving them confidence for doing that same thing, but on a much larger scale?

Kent Draper, Chief Commercial Officer

Yes. I think that's certainly part of the benefit of it. If you look at our executive management team and Board, we've got a very long history of successful delivery on large infrastructure projects, in the many billions of dollars. We certainly have the expertise and capability internally. We've proven that out to a large extent with the build-out of the Bitcoin mining side of the business. As I've discussed in previous calls, the fact that we're building out 50 megawatts of data centers a month is to many people that we talk to in the industry a massive surprise. We've been doing that for many months now and continue to deliver at that clip, albeit acknowledging that is a different type of infrastructure to liquid-cooled. The fact that we're able to build out Horizon 1 at the Childress site is just a further demonstration of our internal capabilities. So yeah, I think it absolutely assists with all of those conversations. Many of those are, as Dan alluded to earlier, sort of taking more of the path of us having an ongoing ownership interest in the infrastructure itself. Further demonstration of our capabilities is certainly useful.

Lincoln Tan, Director of Investor Relations

Thank you. And our next question comes from the line of Stephen Glagola from Jones Trading. Your question, please.

Stephen Glagola, Analyst

Hi, Dan. Belinda, Kent. My question relates to investor concerns around potential HPC monetization of Sweetwater 1. I think there are two large ones. One, the site's suitability for inference compute; and two, single tenants like hyperscalers' willingness to do a deal with a Bitcoin miner given sort of this perception that there is more balance sheet risk. Could you address the validity of these concerns in your view? And has your desire to retain ownership of the land potentially adversely impacted conversations with hyperscalers for a deal at the site? Thank you.

Daniel Roberts, Co-Founder and Co-CEO

Hi, Stephen, it's great to see you. Honestly, I believe the conversation has moved beyond those two concerns, and I'm surprised we're still discussing them. The suitability of West Texas for AI training and inference has been questioned before, and while people raised their concerns six or twelve months ago, we have consistently provided the fundamental facts and addressed the narrative that was being circulated. Fast forward to today, and you're witnessing the announcements from various trillion-dollar companies investing hundreds of billions of dollars. The debate about whether AI inference can be executed effectively from West Texas doesn't seem worth addressing anymore. Similarly, the question of whether a Bitcoin miner will partner to build out AI data center capacity feels like we're revisiting the past. Let's reflect on five years ago; we have never positioned ourselves as a Bitcoin mining company. Our data centers have been specifically designed from the beginning for multi-tenancy and various applications. We've been operating NVIDIA GPUs alongside Bitcoin miners in our data centers for the past twelve months without any issues, so I genuinely don't see the validity in these investor concerns. I believe we have thoroughly addressed them, and now we should concentrate on the next steps, which involve monetizing the portfolio.

Joe Flynn, Analyst

Hi, guys. Thanks for the question. I guess since the last business update three weeks ago, like what would you say is the biggest factor that led to the decision to build out the remaining capacity at Childress for AI HPC? Would you characterize that more as a spec build? Or do you have any like maybe soft commitments or interest to ultimately sign that capacity to customers?

Daniel Roberts, Co-Founder and Co-CEO

Yeah. Kent, would you like to talk to this one?

Kent Draper, Chief Commercial Officer

Yeah, sure. I touched on earlier, we're seeing this dynamic where the demand side is really starting to coalesce in terms of the requirement for liquid-cooled data centers. So that is a part of it. As Dan mentioned earlier, the return side of it looks very attractive given where we're able to deliver that capacity in terms of a CapEx basis. In terms of the conversations we're having, we are having multiple conversations around the capacity. I think a lot of what you see with these customer conversations is there are people talking about liquid-cooled capacity and having the ability to deliver it, but they don't have long lead items on order. They haven't initiated any plans for construction. So we think by actually catalyzing the construction and announcing this to the market and demonstrating the tangible path towards delivering this capacity, that will help with many of the customer conversations we're having. So it's not just an on-spec build. But again, we don't want to get people's expectations high, but there are many good conversations that we're having and seeing interest from various customer segments for this type of capacity.

Lincoln Tan, Director of Investor Relations

Thank you. This does conclude the question-and-answer session of today's program. I'd like to hand the program back to Dan Roberts, CEO, for any further remarks.

Daniel Roberts, Co-Founder and Co-CEO

Thank you. Just to spend two minutes addressing a couple of questions on Twitter or now X, as we've got a bit of a community there that I'd like to give a shout out to, and the level of analysis there is pretty healthy. So just to run through a couple of their questions. Agria Investments, could you provide more detail on the financing strategy for Horizon 1 ATM versus debt? As Kent mentioned, the financing strategy is flexible. The $1 billion ATM obviously underwrites certainty in our ability to deliver these, but the opportunity to layer in debt and project finance into AI data centers against customer contracts is real. Equally, we continue to look at the convertible note market. We continue to look at other forms of capital. So the expectation should absolutely not be that ATM is the sole source and focus of financing going forward, but it does provide something that we can count on to underwrite these growth ambitions. Could you please clarify the unrealized gain loss on the financial instrument in the P&L, the $12.9 million? That's just changing fair value of the convertible notes? A couple more. Why not hold some of the mined Bitcoin? We've outlined this before. We don't believe in diluting shareholders to put Bitcoin on our balance sheet. You can buy it yourselves. Will we convert to a REIT, as suggested by Wolfe and someone else? Not at this stage. We value creation more in the data center development and operations than being a landlord. How long is the sale process from start to onboarding a client into a data center? It depends on the client and the situation. Why did Morgan Stanley step out of the sale process? They didn't step out; they were terminated. We believe our opportunity is best served by dealing elsewhere. What's the best place to get food near Childress? The plaza has fantastic Mexican. So anyway, let's wrap that up. Thanks for all the support on Twitter. To wrap up our Q2 2025 earnings call. Thank you for listening in. We're really excited about the three key initiatives. One, continuing to expand our Bitcoin mining, which we're seeing flow through to robust earnings and operating cash flow. Two, the development of our Sweetwater data center project and hopefully the projects informing a potential 2-gigawatt data center hub in West Texas. The opportunity to pursue a single-site tenant for that facility, but ultimately underwrite through Bitcoin mining in the meantime. And then finally, our expansion into liquid-cooled AI data center capacity with the announcement of Horizon 1, which we're really excited to deliver on this year. So thank you everyone for dialing in. Have a good evening.