CleanSpark, Inc.

Operator: Good afternoon. My name is Krista, and I will be your conference operator today. At this time, I would like to welcome everyone to CleanSpark's Second Quarter 2026 Financial Results Conference Call. [Operator Instructions] Harry, you may begin your conference. Harry Sudock: Thanks, Krista, and thank you for joining us today to review the second quarter 2026 Financial Results for CleanSpark. We encourage you to review our earnings results press release, which was filed today and is available on our website. A webcast replay and transcript of today's call will be added to our website as well once available. On the call today, I'm joined by Matthew Schultz, our Chairman and Chief Executive Officer; and Gary Vecchiarelli, our President and Chief Financial Officer. Some of the statements we make today will be forward looking based on our best view of the world and our business as we see them today. The statements and information provided remain subject to the risk factors disclosed in our 10-K. We will also discuss certain non-GAAP financial measures concerning our performance during today's call. You can find the reconciliation of non-GAAP financial measures in our press release, which is available on our website. And with that, it's my pleasure to introduce Matt Schultz. Matthew Schultz: Thank you, Harry, and thank you, everyone, for joining us this afternoon. This quarter represents continued meaningful progress in CleanSpark's evolution into a digital infrastructure and data center development company. One that utilizes our heritage as energy natives, builds on the strength of our mining operations and ultimately expands the set of opportunities our portfolio can support. I'm going to take a few minutes to share what we are seeing in the market, how that informs our higher-level strategy and then provide an update on Sandersville and the unique opportunity it represents as well as our broader portfolio. We are in the midst of a technology wave similar to the personal computer, the Internet, mobile phones and cloud computing. Except in this case, it has a larger potential total addressable market and an outsized impact on the infrastructure layer required to power it. AI is different because it is compute-denominated and compute is a function of access to energy and data center infrastructure. We've all watched the hyperscalers guide to higher CapEx this year and the central question was if those investments would have a return profile sufficient to justify them, The revenues reported for the AI labs and the cloud service providers are proving this now in real time. The commercial landscape for AI is converging across hyperscalers, chip manufacturers, Neoclouds and the AI labs themselves. Demand for compute continues to grow, but real-world constraints challenge their ability to secure what is required to continue scaling. Power and infrastructure are at the heart of the supply squeeze. Grids are rapidly adjusting in conjunction with their largest customers to meet the moment and deliver capacity for data centers, while maintaining service reliability and price stability for households and businesses. That dynamic aligns directly with how we built CleanSpark. We spend years operating dynamic, energy-intensive infrastructure. At a high level, there are 4 key activities that enabled our evolution into a large-scale digital infrastructure business. These are in various stages of completion, but cover the full scope of our activities. First is land and power. This has been the core of the business for years. We have cultivated a range of key relationships across the country that propelled our growth to 1.8 gigawatts of currently contracted capacity and we'll continue driving fundamental value as we add high-quality assets and projects to our portfolio. We are always striving to enhance or expand existing sites. I can probably share that we added 25 megawatts of contracted capacity to one of our Metro Atlanta locations just last month, making the existing footprint more attractive for HPC utilization. In keeping with our conservative and transparent approach, all of the megawatts are fully contracted and approved, and they do not include our multi-gigawatt growth pipeline or potential expansions at our existing facilities that we're pursuing. Second is commercialization. We're focused on long duration leases with high-quality tenants as our priority. But as the landscape for compute evolves, we will always stay nimble and aggressive. One important shift we are seeing, prospective tenants engage with us on a portfolio basis rather than just a single site. This is reflective of their demand for capacity and of our large diverse set of assets. Third is financing. Gary will share more detail on our capital strategy, but the markets are constructive, and we have a range of attractive options across the entire project life cycle. And finally, construction delivery. We've spent a significant amount of time building out the internal talent and key relationships required to deliver projects on time and on budget. We are setting up the supply chain to create repeatable processes, allowing us to rapidly scale up and scale out. Importantly, this includes working with suppliers that have manufacturing and fabrication processes that can reduce on-site labor by up to 70% by moving production out of the field and into the factory. This business transformation is the largest endeavor CleanSpark has ever embarked on. It pulls on the threads that made us a market leader in energy development and management and also the discipline and operational excellence that propelled us to become the largest domestic producer of hashrate. And now we have the pieces in place to execute on building the AI factories that are required for the intelligence age. As we actioned our go-to-market efforts for the portfolio, Sandersville was the natural starting point given that all 250 megawatts are live. We have a rock solid community relationship. And in January, we closed on an additional 122-acre parcel necessary to support full greenfield data center build. In marketing the site, we received a range of indications of interest with several coming from high credit quality tenants. Among those, we are progressing with a lead prospective tenant. We understand their engineering requirements and their basis of design. In parallel, we have been negotiating the commercial relationship and the associated suite of contracts. While the process is complex, we're encouraged by the progress and confident that we can offer a compelling solution to their significant data center needs. Ultimately, we know how valuable Sandersville is in this environment, and we're committed to providing the right shareholder value through its monetization while also building a relationship with this tenant that can extend far beyond Georgia. Our approach to counterparty selection is disciplined and prioritizes long-term risk-adjusted equity value creation. We are thinking about the potential multi-decade relationships and how to best deliver over those type of time horizons. At the heart of everything we've done in Sandersville is a commitment to win-win outcomes. That meant a power arrangement that protected residents on reliability and affordability while securing the volume and pricing we needed. It meant adding substantially to the local tax base. It meant hiring full-time staff and contractors locally and it meant showing up, use sports, holiday events and local business patronage. When you put down roots in the American Heartland, you joined those communities roll up your sleeves and you contribute. Our community focus is why the acquisition of the additional acreage was seamless. The local economic development authority knows what CleanSpark has built over several years, and they have confidence in what comes next. We are working to replicate this model everywhere we operate. It is the right way to build infrastructure in this country because it creates structural advantages that protect and accelerate our projects for the long term. Looking beyond Sandersville, the same principles apply across the portfolio. We are building a platform, not a single strategy. On our last call, I described the formation of what we see at the Houston area infrastructure hub. Sealy and Brazoria together represent nearly 900 megawatts of current potential utility capacity, strategically selected to support multiphase AI campus deployments. Sealy has 285 megawatts approved with just over 200 megawatts scheduled to come to energize in the first half of 2027. Substation construction is already underway. We have strong visibility into the energization time line and are running a parallel commercialization process. Brazoria has 600 megawatts in 2 phases. ERCOT approval is already in hand for the first 300 megawatts, a meaningful milestone that reflects the scale of the opportunity and the coordination required to advance projects in the market. The second 300 megawatts is progressing through review, and we look forward to growing across the region. I also want to highlight a capacity -- a capability, excuse me, that has continued to differentiate us in tenant conversations. Our ability to expand within established grid relationships. Historically, we increased at Sandersville and Washington. More recently, we added 25 megawatts to our Metro Atlanta footprint. When a prospective tenant asks what the growth path looks like, we can show them a real track record of unlocking additional scale. We see significant expansion opportunities at several sites throughout our portfolio. Across the broader land and power portfolio, we hold 1.8 gigawatts of currently contracted capacity. Not every site will transition to HPC, and that is not the goal. The goal is optionality, aligning the right assets with the right opportunities as demand evolves, with discipline around capital allocation and shareholder returns always at the center of the analysis. Access to grid connect power at scale remains scarce, and we believe it will remain so. The ability to find contract and develop that power is what we spent years building, and it is -- that is exactly what is most necessary to meet the market's relentless demand. As always, none of this is possible without our world-class teams working tirelessly to push the business forward. Their grit and their talent continues to inspire me every day. And with that, I'll turn it over to Gary to walk through the numbers. Gary? Gary Vecchiarelli: Thank you, Matt, and good afternoon, everyone. Before diving into the numbers, I'm going to briefly cover the role that mining continues to play in our operations, and how we see its evolving role in our strategy. Mining remains foundational to our business. It generates the cash flow that allows us to develop our platform deliberately. It provides operational flexibility, and it continues to give us a strategic advantage when competing for power in grid-constrained environments. Our thesis has remained the same for years now, energy production is not coming online fast enough and our infrastructure first approach to building our almost 2-gigawatt portfolio is reflective of that strategy. As we expand into AI and HPC, we are building on mining, not moving away from it. Both businesses share the same foundation, power, land and operations. Mining funds the platform, AI monetizes it. Together, they create a more balanced, durable business. And as we evolve, mining is the engine that funds our future growth. Turning to the numbers. The average bitcoin price in Q2 was approximately $76,000, which was a 24% difference from the prior quarter where the average bitcoin price was $100,000. For the quarter, our revenue decrease compared to the immediately preceding first quarter by approximately $45 million or 25%, directly attributable to the decrease in Bitcoin price. During the quarter, we mined 1,799 Bitcoin, which was only 22 Bitcoin less than the prior quarter, indicating network hashrate growth has flattened, while our operations team has maintained its industry-leading uptime. Despite the lower revenues, we maintained a healthy gross margin of over 40% for the quarter compared to 47% for the previous quarter. Power prices were more favorable this quarter at $0.052 per kilowatt hour compared to $0.056. Best-in-class team continues to execute and deliver regardless of the market climate. This quarter, we recognized a net loss of approximately $378 million, which was flat compared to the net loss in the prior quarter of the same amount. The current quarter's net loss, it is important to note that it includes unfavorable noncash charges of approximately $263 million related to GAAP mark-to-market adjustments on Bitcoin balances. Our adjusted EBITDA was negative $241 million compared to negative $295 million last quarter. which is indicative of the significant drop from the Bitcoin highs of approximately $126,000 in early Q1. Turning our attention to the performance of the second quarter versus the same quarter last year. Revenues declined approximately $45 million or 25% to $136 million. Our Bitcoin production for the current quarter decreased approximately 7% year-over-year due to difficulty but we saw lower power prices of $0.052 per kilowatt hour compared to $0.06 in the same period last year, which helped margins in this lower Bitcoin price environment. Net income decreased year-over-year by approximately $240 million, which was almost entirely due to noncash mark-to-market adjustments. I will also point out that the average Bitcoin price in the same quarter last year was higher at approximately $94,000 and our HODL balance increased by almost 1,700 Bitcoin year-over-year, which further amplified the mark-to-market adjustment. As of the March 31 balance sheet date, our liquidity remains strong with almost $1.2 billion of liquidity. So we had $260 million in cash and 13,561 Bitcoin worth $925 million. It is worth noting that as Bitcoin prices have started to recover since quarter end. The value of our HODL alone sits at approximately $1.1 billion as of today Additionally, we have the entire $400 million capacity on our Bitcoin back lines of credit available to us. The strength of our balance sheet is a key feature for CleanSpark as we have sufficient capital to acquire land power while also preparing our sites for long-term tenancy. We've always taken a disciplined approach to capital stewardship. As demonstrated by the significant reduction in our share count over the last 18 months. This evolution into AI and HPC infrastructure development is exciting because it opens the door to long-term predictable and high-margin cash flows along with access to capital at much lower costs than we have seen historically in the mining business. You have heard us use the word optionality in our prior calls, and that approach has not changed. We have designed our capital strategy to be flexible in order to take advantage of real-time opportunities in the marketplace. Multiple instruments are available for us to finance either CapEx for high credit quality tenant AI site builds or acquisition of new land and power sites. We believe we have a fundamental second-mover advantage. For example, over the last year, pricing and terms have become significantly more favorable for data center landlords. Recent financings have been priced constructively with investor demand far in excess of the offerings. Recent deals have been oversubscribed as much as 5 to 6x. Additionally, several of these deals have priced at slightly over 6%. This is a large reason why counterparty selection is so important and why we are taking a disciplined approach to commercialization. Next, I'd like to provide an update on our digital asset management activities, where we continue to lead and innovate in the monetization of our Bitcoin HODL balance. While Bitcoin and markets broadly experienced elevated volatility and a significant drawdown during the quarter, we were still able to drive net positive cash returns of approximately $4 million, bringing the total cash generated from DAM equities to $17.2 million for the fiscal year-to-date. I would note that these numbers are being generated while we are only activating less than 40% for Bitcoin and DM strategies. While overall cash returns were lower this quarter, we view this as validation that our approach has durability across different market environments. The considerations here: Foremost, we generated a yield on our Bitcoin balance in a down market. Second, our active risk management and approach to trade execution allowed us to be nimble, flexible and aggressive in managing downside risk. I would also note that the investments we are making in people, process and technology and the DAM team will further enhance our capabilities and drive returns going forward. And lastly, the data and experience that we are capturing each day are enabling continued innovation across our institutional grade desk. We have proven our DAM strategy to produce meaningful cash flow to supplement our mining operations. During this last quarter, we saw elevated volatility in a much lower Bitcoin in price, which, while challenging, proved that our institutional grade 24/7 desk was able to both manage risk and monetize the volatility. The actions taken this past quarter will further help us refine our trade execution and risk management. And since quarter end, we have seen a resumption of returns, similar to those we experienced in the first quarter. With that, I will hand it back to Harry to lead us into Q&A. Harry Sudock: Thanks, Gary. We'll now open the floor to questions from the analyst community. Operator, please provide instructions and manage the Q&A for the Q&A session. Operator: [Operator Instructions] And your first question comes from Nick Giles with B. Riley Securities. Nick Giles: Just thinking about your portfolio, Sandersville, Brazoria, Sealy, those seem to be the obvious targets, but when you zoom out, which other assets could be converted that you once thought might only be for a Bitcoin. And how much expansion potential do those sites have? Matthew Schultz: Thanks for the question, Nick. This is Matt. Within our portfolio, we have a number of sites. And what we talked about was when we started Washington -- or excuse me, College Park. We had 5 megawatts of capacity, we expanded that to 50. When we started Washington, we had 36, it's now 86. We've done that. We've duplicated that process across the portfolio. So we have some phenomenal assets. Washington probably is the next highest probability in the pecking order. It's got 86 megawatts energized capacity right now but we're in the process of completing a line study with significant expansion possibility. Additionally, we acquired 60 megawatts of capacity in Jackson, Tennessee. That also looks very promising for AI development. And last but not least, we have 110 megawatts in Cheyenne, Wyoming, and we're fence-line neighbors with another hyperscaler. So there's tremendous opportunities to convert that from an interruptible load to a firm load to accommodate the needs of potential data center clients. Nick Giles: Super helpful, Matt. I appreciate that. Maybe just as a follow-up. Can you just speak to how much capital you've deployed to date at Sandersville? And then what's really the threshold. The amount of capital that you're willing to deploy at any given site before lease signing? Gary Vecchiarelli: Yes. This is Gary. I'll take that question. So we've deployed a couple of hundred million dollars, which includes the miners. And while we may be subject to some impairment like some of our peers are, once we convert to AI, that's simply going to be a book adjustment. We feel pretty confident that we'll be able to move those some of that infrastructure and miners elsewhere. But until AI is up and energized, we're going to be able to monetize those megawatts at that point. But right now, the amount that we're actually investing in the site is minimal. As you know, we acquired some additional acreage adjacent to the current Sandersville site. And really what we're doing is we're clearing those trees and moving some dirt but we're talking millions, not tens of millions of dollars that we're doing because we want to make sure that we have a lease sign before we deploy significant amounts of capital. Operator: Your next in comes from the line of Greg Lewis with BTIG. Gregory Lewis: Matt, I was hoping you could elaborate more on the comment you made around I guess, some of the conversations are around multi-site -- an agreement or potential multi-site deployment with maybe 1 tenant. Just kind of as we think about that, is that like broad strokes in terms of the first deployment maybe to the next deployment, is there kind of a target range of power. These potential tenants they're looking for? Any kind of comments around elaborating on that, I think, would be helpful. Matthew Schultz: Yes, Greg, good to hear from you. Sounds like you're fighting the same cold I am. So I'll tell you how it got to be really interesting. Obviously, we're -- we consider ourselves to be pretty experienced in land and power. But when it comes to fit up and fit out of the data centers, obviously, there are third-party vendors that provide a significant portion of that build-out. And what we talked about in our comments was that 60% to 70% of the data center build that we're contemplating takes place in a factory. So it short-circuits the time to market, but it also decreases the on-site, the specific build. Well, in meeting with some of those vendors, they obviously have strong working relationships with the hyperscale tenants. And as a result, they understand the demand for capacity. So it's through some of those relationships that the introductions have been made to potentially deploy a portfolio approach where we can meet the needs of a single tenant that has requirements across a diversity of geography. So we have some low latency opportunities and then we have larger scale opportunities. And what we found is that there are single tenants that require many of those characteristics. So while it's still early, I can tell you those conversations have been very fruitful. Gregory Lewis: Okay. Great. And then my other question was around future Power acquisition. And clearly, you already have a lot of power that we can deploy to kind of build out the HPC business, but also accentuate the Bitcoin mining business. But as we sit here and I guess, what May 2026, I guess how is the company thinking about incremental power acquisitions? Matthew Schultz: So that's a fantastic question. And I want to be really clear about the way we disclose this. The 1.8 gigawatts is power that's approved and contracted and available. when we sat down earlier this year with our leadership team, and we designed OKRs, kind of a strategy for accountability going forward. we talked about the pipeline. And in our pipeline, there's greater than 5 gigawatts of capacity beyond what we shared on the call. However, that's speculative or potential. So I think there are a lot of companies that disclose maybe capacity that they've had discussions about. We've had many of those same discussions. We're in meaningful conversations about significant additional power but we're very careful and the number that we disclose is restricted only to what is already contracted. So I would say that the future is pretty bright for us. Everybody has seen the political headwinds on data centers. And so one of the positive comments that we received as we've put these sites available to the market is that we have manageable amounts of power and meaningful amounts of land in different jurisdictions. So it doesn't create like what we've seen with Mr. Wonderful's proposed site of 9 gigawatts and whatever the case may be. It's caught some tremendous political headwinds just based on scale. So being disciplined and have a bite-sized approach with communities and utilities that actually want us there, based on prior experience with us, it's created some interesting tailwinds. Operator: Your next question comes from the line of John Todaro with Needham & Company. John Todaro: Congrats on the quarter. Matt, you mentioned earlier, progressing with a lead perspective, tenant. Just wondering if we could get a little bit more on it. Is it still kind of the advanced conversations that was framed before as an IG hyperscale? Or is that the right way to think about it? Or has that evolved at all? And then I have a follow-up. Matthew Schultz: I think that's pretty consistent. We're -- we don't have another update beyond that, but our conversations remain consistent with the first tenants that we've been talking on. John Todaro: Okay. Understood. And then I guess, just a housekeeping item, maybe for Gary, on Bitcoin mining and the hashrate. Just kind of as we think about more site portfolio going towards HPC, just any kind of range we should think about on a hashrate moving forward? Gary Vecchiarelli: Yes. Great question. So as I mentioned in my comments, look this -- Bitcoin mining is really our functional currency going forward, and that's what's going to pay the bills until we get a stabilized lease. And I think that that's one of the benefits of having a very large fleet at scale with one of the most efficient fleets at least of the public miners or in North America. And we did see through our -- the last contracted commitment with BITMAIN. And so we have some immersion miners that have landed that we're going to be deploying at various sites. And the great thing there is that, that's going to drop our efficiency even more from the 16 joules terahash that were currently posting, which will increase margins. And the best part is these are going to be in emergent pods that we can then up and move as part of our AI strategy. So you'll see the hashrate start to tick up. I mean, our Bitcoin production has broken 23 just recently. So obviously, some of that's going to depend on Bitcoin difficulties, network difficulty and Bitcoin price, but you'll probably start to see this trend to 55 through the end of the year. Operator: Your next question comes from the line of Brett Knoblauch with Cantor Fitzgerald. Brett Knoblauch: Matt, maybe just on the tenant conversations and the maybe evolution of this portfolio approach. Is that the same tenant who was looking at Sandersville, who is also looking at maybe some of the other sites? And does this portfolio approach potential maybe push back when the first lease signing could be? Or do you think it doesn't change the timing from your lens? Matthew Schultz: No, I don't -- Brett, first of all, thanks for the question. It was good to connect with you at the Bitcoin Conference. So it doesn't change the timing whatsoever. The assets are -- they have value individually and independent of one another. I think for us, first prize is having the right credit quality tenant that has interest across the portfolio. So it doesn't change timing. I feel really good about where we are in the cadence we've had to date. And I think the portfolio approach is more frosting on the cake. Brett Knoblauch: Understood. And then, Gary, on the Bitcoin side, would you guys be looking at maybe co-locating Bitcoin mining with AI? Is that something that you've discussed with tenants? Or do you think every site is either going to be bifurcated between -- its just going to be AI or it's just when Bitcoin mining? Gary Vecchiarelli: Yes. Look, I think -- we think we're innovators in the space when it comes to pairing both AI with Bitcoin mining. And in my comments, I'm not sure if you caught what I said exactly, but I think that the Bighorn mining is actually going to allow us to land new land and power opportunities because we'll be able to monetize that energy near to 100% of that base load. So we are having conversations, and I think you can expect different flavors of AI and Bitcoin mining. One of the ones that we're really looking forward to is builds right now are 18-plus months right now for AI or HPC. And if the site is energized, we can pop down some mining pods and actually monetize that energy while we're waiting for the AI site to be built. So again, there's different flavors that this will come. And I think that because it's rather innovative and new to the market, it's going to take time and education for all parties, but it's part of our strategy, and we think it's pretty viable. Matthew Schultz: And maybe just to add a little flavor on that, Brett. We have a patent from our Bitcoin mining days. And the patent basically says that the control module has the ability to make a decision if and when to distribute power to a requesting module. So back when we were in the microgrid business, we had a controller that would allocate power as requested based on a number of criteria. So the idea is that with many of these data center loads, the peak PUE, as an example, on Sandersville, the discussion surround that, the peak PUE for the tenant would be in that 1.4 range. The average annual POE would be around 1.15. So that leaves a significant amount of megawatts available 93-plus percent of the year. So the challenge then becomes for the utility to have that power available to meet the [indiscernible] requirement for a hyperscale tenant, what do they do with those non-monetized megawatts? So the conversations that we've had contemplate utilizing Bitcoin mining to make sure they meet the minimum utilization thresholds to ensure the pricing that's so important for the hyperscale tenants but also having the ability to mine Bitcoin profitably. And even in some instances potentially subsidized using that spare capacity. So it's early in the discussion yet but it really creates a win-win scenario. So the utility doesn't have on-demand available yet underutilized power. They still have the ability to monetize that to create the revenues to invest in future CapEx. So it's a hybrid approach, but we're really excited about what the potential looks like. Operator: Your next question comes from the line of Paul Golding with Macquarie Capital. Paul Golding: Just wanted to ask around the incremental capacity under review at Brazoria 300 megawatts. I was just hoping you could give some detail around how that review process is going and maybe as a foretelling of how markets may look over time in terms of incremental power acquisition you may be able to do? Is the market tightening here? Is the review process fairly stringent and scrutinizing this capacity. How should we think about this review process and the current sort of pipeline of interconnect capacity and the availability of powered land or, I should say, lack thereof impacting power land banking going forward in your view for your portfolio? Harry Sudock: Thanks Paul. It's Harry. So I think that the most important piece in all of this is that our process by which we achieve contracted power, approved power starts during our site evaluation and M&A process. We have a team that has done this very, very many times across a diversity of markets. And I think we've been able to build a reputation for being an incredibly strong deal partner in the acquisition of powered land assets. So the ability to work with the approval bodies, whether that's a cluster study or something else, that all begins early on in the acquisition process actually. It's a function of our diligence. It's a function of the way that we engage cross functionally throughout the full stack of not only the utility but also the regulatory body as well as the land owner as well as the adjacent parcels, should we need additional capacity as well. So how does that relate exactly to Brazoria is that we came into that process with a very clear understanding of the Greater Houston market because if you remember, we closed on our Sealy asset just 60 days prior to that. And so we've already been in market with ERCOT, we've already been in market with CenterPoint. We understood the nuances, the zoning, the planning, et cetera. And so we came in from a position of tremendous strength of that conversation. It's why we were able to strike such a constructive deal there. It's also why we were able to close on the 300 at the full approval status with the second 300 in a deeply progressed posture. And so ultimately, our sense of it from CenterPoint from ERCOT is that, we're living in that 0 territory for the additional component, but it also signals this track record that we have, which is that we enter a market, we land there, and then we rapidly expand there. It was the same playbook in Georgia, obviously, a very different power market, same in Tennessee, about as different from ERCOT as you can get with a fully federally regulated utility there. And now you're seeing us do that in ERCOT in the Greater Houston area. And we didn't just trip on land in the Greater Houston area for no reason. It was about the transmission availability there as well as the range of utility sophistication that we're excited to grow with. Paul Golding: Harry, I was also hoping that you might be able to comment on maybe incremental acreage you would need to acquire at any other sites as you think about the spec that's unfolding for Sandersville, you did the incremental acreage acquisition there. Are there -- are you looking at your portfolio right now and thinking about that spec as a blueprint for additional land you may need to acquire at other sites. And is that filtering out some of the sites from an HPC viability perspective? Harry Sudock: Thanks, Paul, and great question. I want to tackle it in pieces by region. So in Texas, the land parcels that we've acquired as part of the Power acquisition are sufficient to build the AI campuses that we've contemplated. That's the benefit of going in on a purpose-built kind of acquisition trajectory as we did in that market. I think Sandersville looks different because we were able to land 250 megawatts of mining on a 50-acre parcel, whereas the data halls and the adjacent required mechanical, electrical and plumbing are going to be a much larger physical footprint. So that's why we took down that additional acreage. I think what's great about the way -- and you'll hear this from us many, many times is that community relationships are business tool as well as doing the right thing. And so because we've built the community relationships that we have, the ability to acquire additional land where necessary is something that is very kind of fertile ground for us and very low friction. But ultimately, we're going to evaluate this on a site-by-site basis. I think we have sites today where there is sufficient land. I think there are others where we're going to look to get a little bit more elbow room in order to take down a full HPC deployment. But ultimately, what's important is that we have the real estate shops and the development shops to enable to be -- in order to be able to do that in a seamless way. Matthew Schultz: And Paul, this is Matt. Just to give you a little bit of color on how that works. When we met with the Economic Development Director in Sandersville and told them what we were contemplating doing. Not only did they assist us in securing the 122 additional acres. It's not a fence line neighbor parcel. It's drive or 9-iron away. So they then assisted us in negotiation a right of way for the easement to pull the power across. So we still have the ability to mine Bitcoin up until the day we cut it over and then we can mine it in conjunction. But the reason that, that's important is this was the city taking the initiative to ensure that our requirements were met, and they were excited to do so based on our history. And that's consistent across our portfolio. We've -- there have been times that we show up at a little town in Tennessee or a little town in Georgia, and wearing a CleanSpark golf shirt, I can't pay for a cup of coffee because folks are happy about us being there, and they're excited with what we've created. So it's a wildly differentiated approach to what you're seeing in the media. Operator: Your next question comes from the line of Brian Dobson with Clear Street. Brian Dobson: So earlier, you're like in the AI data center build out to that of personal computing. I guess we're still in the very early stages. Maybe can you just articulate your vision for CleanSpark in, say, 2030 a few years down the road once you went up to start out on this or rather embark on the build-out? Harry Sudock: Brian, I think what's important to understand about where AI build-outs are happening and I really differentiate it from sort of the dark fiber wave of the Internet is that, that was a opportunistic speculative build out where the fiber was laid because you were digging the trench one time, it made more sense to delay 50 strands, not 5 strands. And ultimately, the users at the other end of those strands just weren't there yet. Not because the Internet wasn't going to become a valuable tool, but because there was a distribution curve that the Internet had to go through to get the economic purchase power into enough hands to make it the globally transformative technological platform than it is today. AI is building on the shoulders of giants because we're all walking around with laptops and mobile phones, we're all walking around with connectivity in our houses and in our cars. And so because the distribution rails were built 30 years ago, AI is able to proliferate across the user base at a much, much more accelerated fashion than with prior computing waves that were fundamentally hardware-enabled. And so how does that read through to our portfolio and our vision for the business is that, every megawatt that can be economically pushed into an intelligence posture is likely going to be over time. And so -- what does that mean in the short term? It means that we don't have enough electrons. And so the electrons that are available to be built out into infrastructure environments that are going to be very rapidly and are going to be done through a very aggressively bid process by which those tokens can get into the hands that they can do the most economic good for the most people. And so that means, 10, 20, 50 megawatt sites out of the gate, if you look back to 2022 and '23, it meant 250-plus megawatts sites, if you look '22 to today, and it's going to look like gigawatt campuses if you look out into the future. And so what's important to note is that just because the gigawatt campuses today are providing tremendous economic value. The 250s are still producing. The 50s are still producing. The 10s are still producing across the entire spectrum of these deployed token factory environments. And so what's the next phase for this? It's going to be continued proliferation and distribution of intelligence as a fundamental economic good. And so our job in the midst of all of this is to make sure that the communities that didn't benefit from the proliferation of the smartphone and the Internet are not going to get left behind in this AI technology wave. These are the places where infrastructure is going to fundamentally change the lives of hundreds of thousands of people. And if we can be the developer who can bring and distribute that, we're going to, we're going to do that on behalf of the places where we work and on behalf of the shareholders who supported us. Matthew Schultz: And Brian, I want to apologize for Harry's lack of enthusiasm there. One thing to note, there's a study called Rethinking Load Growth that Duke University published. And what it discussed is the fact that as of today, right now, in the 22 largest power grids in the nation, there's between 76 and 125 gigawatts of headroom, that is unlocked and available if it's able to be curtailed between 0.5% and 1.5% of the time. So we hear all this about, the grid is overtaxed, it's overcapacity, it's overstressed, it's overstrained. But in reality, there's 100 gigawatts of headroom that if you can curtail 115 hours a year, you really unlock that capacity. So what we found is by working in some of these smaller jurisdictions, and by discussing the ability to add Bitcoin mining, which is a rapidly interruptible load to that factor, it really changes the way that these discussions open up. So we're excited about not only the gigawatt campuses and the 250-megawatt campuses that we're working toward deployment. We're also very excited about the 20s and 50s and 60s that we have. In fact, to not go too far down the rabbit hole, we've had conversations with tenants that have specifically asked how much can I buy 60 megawatts or above between now and 2027. So the demand is everything you have as fast as you can get it. And for us, it's become more of a sorting process to ensure that financial capability of the offtake and the financeability of the project is well within scope. Operator: Your next question comes from the line of Mike Colonnese with H.C. Wainwright. Michael Colonnese: First one for me. It sounds like you will be taking a pretty unique approach to data center construction with your vendor relationships and strategy here. What do you think your estimated data center deployment time lines could be once you signed your first lease and the associated expected CapEx cost per megawatt for these builds. It sounds like you guys might have a bit of an edge here. Just want to get some more information around that. Harry Sudock: Thanks, Mike. And you'll get tired of us telling you that we're conservative in everything that we do and say, but ultimately, what we think from a build and deployment perspective is that from lease signing, you're really looking in the 14 to 18-month range for delivery. That's a function of first data hall versus last data hall. It's a function of project size and some of those pieces. But ultimately, what we're looking to do is be conservative and deliver aggressively on time and on budget out of the gate and then over time, be able to smooth and innovate across supply chain and delivery such that we're able to compress those time lines project by project because like we saw with our Bitcoin mining deployment learning curve, we were able to get a lot better by the end build, not the first build. And so we expect to be able to integrate those similar type of learnings and innovations into our deployment for this as well. Michael Colonnese: Got it. And is there any CapEx advantage using this more modular sort of factory-based data center construction approach? Or is it pretty much in line with what you've seen in other builds? Harry Sudock: Yes. I think that these are going to be really in line from a build and deploy perspective. But what we're looking to be able to do is create an offering that's going to have the best total cost of ownership for the client over multiple refresh cycles. So there's a component of this that's fundamentally grounded in future proofing because we want to be able to meet their needs as effectively as possible. that means delivering that total cost of ownership differential, not just when they put their first GPU in, but when they refresh it, refresh it again or refresh it again. Michael Colonnese: Very helpful, Harry. And second one for me. When we look at GPU ownership versus colocation opportunities, would you guys still consider pursuing GPU cloud service opportunities out there based on current economics and implied return profiles. And if that's the case, which assets in the portfolio make the most sense for you to own your own GPU versus going the colocation [ one ]? Harry Sudock: Yes. Look, I think that if you rewound the clock 4 or 5 years, when first H100's hit the market, nobody would have thought that they'd be renting for the prices that they're renting at today. And so that is a hugely positive sign, not only for the GPU-as-a-Service business model, but for the overall health in the demand profile for AI compute as an overall sector. But ultimately for us, what we want to do is be able to come to market with the very high stability, high margin and long duration cash flows that colocation and tenant relationships are able to deliver. We're never going to be in the business of saying never say never. But ultimately, what we want to do is be able to put 1 foot in front of the other, look at the projects and the opportunities directly in front of us, execute aggressively against those, and then we can turn our focus to other opportunities that are further out the curve, but we have to deliver first. Operator: Your next question comes from the line of Matthew Galinko with Maxim Group. Matthew Galinko: Firstly, how do you manage, I guess, multi-site risk across your portfolio? Or I guess, in other words, would you be comfortable with the single tenant saturating your pipeline? Or would you look to have kind of multiple of these multisite agreements? Matthew Schultz: I appreciate that question. Thank you. This is Matt. I think it really depends on the credit quality of the tenant. We've -- throughout this process, we've had a tremendous amount of inbound inquiry. And so you have to consider if we do a Neocloud deal, we can print a higher rent figure. But it requires a wrapper, and oftentimes, the wrapper comes at the cost. And many times, it includes a significant scrape of equity. So as we look at this, it really just comes down to balancing the credit quality and risk with the ultimate value or the cost long term for our shareholders. So I wouldn't say that we have a an aversion to concentration risk if it's the right $1 trillion tenant. But I would say that it really is on a case-by-case basis. We certainly don't want overexposure to a Neocloud and end up having to print a tremendous amount of warrants just to get the deal financed. Matthew Galinko: Great. All right. That makes sense. And as my follow-up, I think you mentioned in the script that you'd be able to relocate some of the Sandersville fleet on conversion. Are you looking to retire some of the fleet there? Or is it a lack of space elsewhere? I'm just curious how you envision that moving? Gary Vecchiarelli: Matt, it's Gary. Yes, so we have XPs and the most recent S21 there in addition to the newest immersion miners. So I think naturally, as time goes on, we will retire some of those ASICs, but we'll just do what we've always done and rather than just wait for them to completely die, we'll look to monetize those while they still have some useful economic life. But ultimately, I think that I just refer to my comments about optionality and flexibility, like we can be very responsive in the market, right? If Bitcoin goes to $125,000 tomorrow, XPs are probably going to see a spike in price per terawatt or terahash, and we could probably move some of those and we might be opportunistic. But ultimately, our team is evaluating in real time what the best use of those mining assets in ASICs are, including which sites to have them at. And so I think it's just going to be an ongoing part of the operations that you've seen from us over the past couple of years. Operator: Your next question comes from the line of Jon Hickman with Ladenburg. Jon Hickman: I'm not sure who wants to answer this question, but I'm a little confused about the ability to provide 0.95 power for and still meets the needs of a community that is used to getting all the power they want whenever you need to turn offer on your bitcoin mining. How do you balance that out? Harry Sudock: Thanks, John. So ultimately, what's great is that we work hand in glove with the utilities and the communities as we go through a transition phase for some of the power contracts that are going to go through that. I think Sandersville is really the place to start with that. The utility goes -- thinking through how the power is procured in Georgia, utility goes out to market and procures the power over duration directly for the contracted demand that we represent at that location. And so by moving into that method of procurement, they've mitigated the community risk for any pass back of that because they're essentially procuring that power on the open market, knowing exactly where it's going to get landed, which is with us. And on other side of the fence, we have a capacity factor obligation under that because we agree to consume a certain percentage of uptime underneath that power. And so because of the way that they go out to market and because of the way that we agree to consume on behalf of our tenant, there is no pass back into the community as it relates to that type of contract migration. What's great about Texas is that the regulated market in ERCOT doesn't contemplate that type of interruption or expectation of interruption in the market at large. And so we went to that market explicitly for the purpose of building large-scale AI data center campuses because we know that the power market there is more than able to support it, and the tenant community is obviously very constructive about growing their footprint there as well. So I think in Georgia, it looks a little bit different than it does in Texas, but the net result is that we're able to provide that very firm and high uptime to the tenant, and we're able to source that power in a way that is low impact to the community and the infrastructure. Jon Hickman: So the community in Sandersville that's used to this bitcoin mining operation, they're going to be okay as you turn it into HPC? Harry Sudock: They're incredibly supportive of our business evolution. Operator: your next question comes from the line of Michael Grondahl with Northland Securities. Mike Grondahl: just wanted to get an update on how demand has kind of evolved for some of your smaller sites. And kind of where does that demand sit today? Matthew Schultz: Yes, I appreciate the question, Mike. I can tell you, we had a standup call today with the Neocloud company and their ask was, show me all your sites that have 60 megawatts of power or more available today or in the short term. So the demand is going nowhere but up. In fact, during that call, they commented that they're aggressively seeking 8 gigawatts of capacity. So we're seeing the 500 megawatts or 1 gigawatt minimum thresholds really start to drop because of a couple of factors. First and foremost, a lot of those mega sites that are grid connected are years away or already under discussion or leasing. Secondly, there's a real risk attributed to behind-the-meter power generation. So many of the companies that have made behind-the-meter gas decisions are facing energization delays or utility costs that are double or triple what we're seeing for grid connected sites. Now I think that there were some constructive conversations that took place at the White House with hyperscalers agreeing to generate additional power and push it back to the grid. But I think that kind of flies in the face of that Duke University study that identifies 100 gigawatts of headroom on the grid today. And I think if you can find a way to balance that with a portfolio approach using incremental components to make up that 8 gigawatts of demand, for example, from a specific client and use interruptible loads to pair with firm loads for data centers, I think you can unlock a tremendous amount of capacity that has previously gone, I think, unnoticed. Mike Grondahl: Sounds good. Nice to see it moving down to some of those smaller sites. That's for sure. Operator: We have time for 1 more question, and that question comes from James McIlree with Chardan. James McIlree: Yes. I was hoping you could address how you're looking at managing, securing talent for engineering as well as labor, particularly over the next -- after the next 12 or 18 months, the next phase of build out, talent specifically for building out the next wave of data centers. Matthew Schultz: That's a fantastic question, and it's a very real concern. We've done some data center consulting and development with companies like McKinsey. And one of the things that was identified is the labor bottlenecks in markets like Texas. You look at the demand for plumbers and electricians over the next 36 to 48 months, it's terrifying. But I think we've done the best we can to mitigate that. And I'll give you an example. Everybody has kind of heard talk about the West Texas development for a major hyperscaler that has changed direction and ownership and had some real challenges. We've all read the stories that there are 4,000 to 6,000 guys driving on dirt roads every day and they're bringing in fresh [ pool of ] parties just to keep it roll it up. Well, our conversations with our development partners take about 70% of that out of the way. So as an example, Sandersville, there will be a 494,000 square foot building, and it would be 400 full-time employees on the construction team rather than 4,000 to 6,000. And that comes a function of building a significant component of the data center in a factory. It's duplicatable, it's consistent. It's an assembly line process. So you can meet the specific requirements from the chip manufacturer as an example. You build to suit that particular design, and then you duplicate that and you simply install it in the field. So the guys in the field are construction, plumbing, electrical only to the point that it feeds the MEP that we're bringing in, in a modular solution and decreasing some of that risk. So we have a pretty robust team so far. We've engaged with, as I mentioned, McKinsey and a couple of others, Accenture to assist us in that road map going forward. And we've had a tremendous amount of inbound inquiry because what we've done historically is hire locally. And there are a number of trades in a lot of the jurisdictions that we currently operate that aren't chasing data center builds around the country. They're looking for opportunities at home, and we're finding that we have a significant amount of talent locally that can help support what we need. Operator: And that concludes our question-and-answer session. I will now turn it back over to the presenters for closing comments. Harry Sudock: Everyone, thank you again for joining today's earnings call. We look forward to staying in touch and sharing future results with you in the coming quarters. Stay tuned for more progress and exciting achievements ahead of us at CleanSpark. Operator: Ladies and gentlemen, this does conclude today's conference call. Thank you for your participation, and you may now disconnect.