Earnings Call Transcript - CLSK Q2 2026

Operator, 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 provides instructions to participants. Harry, you may begin your conference.

Harry Sudock, Head of Investor Relations

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, Chairman and Chief Executive Officer (CEO)

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 spent years operating dynamic, energy-intensive infrastructure. At a high level, there are four 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 is 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 types 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 for sports, holiday events and local business patronage. When you put down roots in the American Heartland, you join 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 timeline and are running a parallel commercialization process. Brazoria has 600 megawatts in two 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 capability 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-connected 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 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, President and Chief Financial Officer (CFO)

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 decreased 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. Our 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 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-backed 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 and 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 activities 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 our 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 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, Head of Investor Relations

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

Operator provides instructions to participants. And your first question comes from Nick Giles with B. Riley Securities.

Nick Giles, Analyst, B. Riley Securities

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 Bitcoin? And how much expansion potential do those sites have?

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

Thanks for the question, Nick. This is Matt. Within our portfolio, we have a number of sites. What we talked about was when we started 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 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, Analyst, B. Riley Securities

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, President and Chief Financial Officer (CFO)

Yes. This is Gary. I'll take that question. So we've deployed a couple of hundred million dollars, which includes the miners. 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 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. 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. 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, because we want to make sure that we have a lease signed before we deploy significant amounts of capital.

Operator, Operator

Your next question comes from the line of Greg Lewis with BTIG.

Gregory Lewis, Analyst, BTIG

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 one 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 are looking for? Any kind of comments around elaborating on that, I think, would be helpful.

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

Yes, Greg, good to hear from you. Sounds like you're fighting the same cold I am. I'll tell you how it got to be really interesting. Obviously, we're experienced in land and power. But when it comes to fit-up and fit-out of the data centers, there are third-party vendors that provide a significant portion of that build-out. 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 and decreases the on-site work. In meeting with some of those vendors, they obviously have strong working relationships with the hyperscale tenants. As a result, they understand the demand for capacity. 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. We have some low-latency opportunities and then larger scale opportunities. What we found is that there are single tenants that require many of those characteristics. While it's still early, I can tell you those conversations have been very fruitful.

Gregory Lewis, Analyst, BTIG

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 build out the HPC business, but also accentuate the Bitcoin mining business. As of May 2026, how is the company thinking about incremental power acquisitions?

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

That's a fantastic question. I want to be really clear about the way we disclose this. The 1.8 gigawatts is power that's approved, contracted and available. When we sat down earlier this year with our leadership team, and we designed OKRs and a strategy for accountability going forward, we talked about the pipeline. In our pipeline, there's greater than 5 gigawatts of capacity beyond what we shared on the call. However, that's speculative or potential. We are 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. I would say that the future is pretty bright for us. Everyone has seen the political headwinds on data centers. One of the positive comments 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. That avoids the scale-related political headwinds others have seen. Being disciplined and taking a bite-sized approach with communities and utilities that actually want us there, based on prior experience with us, has created some interesting tailwinds.

Operator, Operator

Your next question comes from the line of John Todaro with Needham & Company.

John Todaro, Analyst, Needham & Company

Congrats on the quarter. Matt, you mentioned earlier progressing with a lead prospective tenant. Is that still in advanced conversations framed before as an investment-grade hyperscale? Or has that evolved at all? And I have a follow-up.

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

I think that's pretty consistent. We're not providing another update beyond that, but our conversations remain consistent with the first tenants that we've been talking to.

John Todaro, Analyst, Needham & Company

Okay. Understood. And then maybe for Gary on Bitcoin mining and the hashrate: as more of the site portfolio goes toward HPC, what range should we think about for hashrate moving forward?

Gary Vecchiarelli, President and Chief Financial Officer (CFO)

Great question. 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. That's one of the benefits of having a very large fleet at scale with one of the most efficient fleets among public miners in North America. We did see through our last contracted commitment with Bitmain that some immersion miners have landed and we're going to be deploying them at various sites. That will drop our energy per terahash efficiency even more from the 16 joules per terahash we're currently posting, which will increase margins. These are going to be in emergent pods that we can then move as part of our AI strategy. You'll see the hashrate start to tick up. Our Bitcoin production has broken 2,300 just recently, and some of that will depend on Bitcoin difficulty and Bitcoin price, but you'll probably start to see a trend toward 5,500 through the end of the year.

Operator, Operator

Your next question comes from the line of Brett Knoblauch with Cantor Fitzgerald.

Brett Knoblauch, Analyst, Cantor Fitzgerald

Matt, maybe just on the tenant conversations and the 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 potentially push back when the first lease signing could be? Or do you think it doesn't change the timing from your lens?

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

No, Brett, thanks for the question. It doesn't change the timing whatsoever. The assets have value individually and independent of one another. First prize for us is having the right credit quality tenant that has interest across the portfolio. The portfolio approach is more frosting on the cake and doesn't change timing. I feel really good about where we are in the cadence we've had to date.

Brett Knoblauch, Analyst, Cantor Fitzgerald

Understood. And then, Gary, on the Bitcoin side, would you be looking at co-locating Bitcoin mining with AI? Is that something you've discussed with tenants? Or do you think every site is either going to be bifurcated between AI and Bitcoin mining?

Gary Vecchiarelli, President and Chief Financial Officer (CFO)

We think we're innovators in the space when it comes to pairing both AI with Bitcoin mining. The new immersion mining allows us to land new land and power opportunities because we'll be able to monetize that energy near to 100% of that base load. We are having conversations, and you can expect different flavors of AI and Bitcoin mining. One approach is that AI or HPC builds are currently 18-plus months. If the site is energized, we can deploy some mining pods and monetize that energy while we're waiting for the AI site to be built. There are different flavors, and because it's innovative and new to the market, it will take time and education for all parties, but it's part of our strategy and we think it's viable.

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

And maybe just to add a little flavor on that, Brett. We have a patent from our Bitcoin mining days. The patent describes a control module that can decide when to distribute power to a requesting module. Back when we were in the microgrid business, we had a controller that would allocate power as requested based on a number of criteria. With many of these data center loads, the peak PUE, as an example on Sandersville, would be in that 1.4 range while the average annual PUE would be around 1.15. That leaves a significant amount of megawatts available over 93% of the year. The challenge for the utility is having that power available to meet the on-demand requirement for a hyperscale tenant. The conversations 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 hyperscale tenants, while also having the ability to mine Bitcoin profitably and potentially subsidize using that spare capacity. It creates a win-win scenario: the utility can monetize underutilized power to create revenues for future CapEx, and tenants can access competitively priced capacity. It's early, but very promising.

Operator, Operator

Your next question comes from the line of Paul Golding with Macquarie Capital.

Paul Golding, Analyst, Macquarie Capital

Just wanted to ask around the incremental capacity under review at Brazoria — the 300 megawatts. Could you give some detail around how that review process is going and maybe how markets may look over time in terms of incremental power acquisition? 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 pipeline of interconnect capacity and the availability of powered land impacting power land banking going forward in your view for your portfolio?

Harry Sudock, Head of Investor Relations

Thanks, Paul. The most important piece is that our process to achieve contracted, approved power starts during our site evaluation and M&A process. We have a team that has done this many times across diverse markets. We've built a reputation for being a strong deal partner in the acquisition of powered land assets. The ability to work with approval bodies, whether that's a cluster study or something else, begins early in the acquisition process. It's a function of our diligence and how we engage cross-functionally with the utility, the regulatory body, the land owner and adjacent parcels if additional capacity is needed. Related to Brazoria, we entered that process with a clear understanding of the Greater Houston market because we closed on our Sealy asset about 60 days prior. We were already in market with ERCOT and CenterPoint and understood the nuances. That's why we were able to strike such a constructive deal and close on the first 300 megawatts at full approval status with the second 300 in deep progress. Our sense from CenterPoint and ERCOT is that this signals the track record we have: we enter a market, land there, and rapidly expand. It was the same playbook in Georgia and Tennessee. Now you're seeing us do that in ERCOT in the Greater Houston area, and we didn't land there by accident. It was about transmission availability and utility sophistication that we're excited to grow with.

Paul Golding, Analyst, Macquarie Capital

Harry, could you comment on incremental acreage you would need to acquire at other sites as you think about the spec that's unfolding for Sandersville? Are you looking at your portfolio and thinking about that spec as a blueprint for additional land you may need to acquire at other sites? Is that filtering out some sites from an HPC viability perspective?

Harry Sudock, Head of Investor Relations

Great question. I'll tackle it by region. In Texas, the land parcels we've acquired as part of the power acquisition are sufficient to build the AI campuses we've contemplated. That's the benefit of purpose-built acquisitions in that market. Sandersville looks different because we were able to land 250 megawatts of mining on a 50-acre parcel, whereas data halls and adjacent mechanical, electrical and plumbing will require a much larger physical footprint, which is why we took down the additional acreage. Community relationships are both a business tool and the right thing to do. Because we've built those relationships, acquiring additional land where necessary is low friction. Ultimately, we'll evaluate this on a site-by-site basis. We have sites today where there is sufficient land and others where we may need more elbow room to take down a full HPC deployment. We have the real estate and development teams to do that seamlessly.

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

Paul, to give you color: when we met with the Economic Development Director in Sandersville and told them what we were contemplating, they assisted us in securing the 122 additional acres. It's not a fence-line neighbor parcel; it's a short drive away. They assisted us in negotiating a right of way for the easement to pull the power across. We still have the ability to mine Bitcoin up until the day we cut it over and can mine in conjunction. The city took the initiative to ensure our requirements were met and were excited to do so based on our history. That's consistent across our portfolio. There have been times we show up in a small town and folks are happy about us being there. It's a differentiated approach to what you may read about elsewhere.

Operator, Operator

Your next question comes from the line of Brian Dobson with ClearStreet.

Brian Dobson, Analyst, ClearStreet

Earlier you likened the AI data center build-out to that of personal computing. We're still in early stages. Can you articulate your vision for CleanSpark in, say, 2030, a few years down the road once you embark on the build-out?

Harry Sudock, Head of Investor Relations

Brian, important to understand where AI build-outs are happening. I differentiate it from the dark fiber wave of the Internet, which was largely opportunistic and speculative. AI is building on distribution rails that were built over the last 30 years—laptops, mobile phones and connectivity exist today—so AI can proliferate much faster than prior computing waves that were fundamentally hardware-enabled. Every megawatt that can be economically pushed into an intelligence posture is likely to be over time. In the short term, we don't have enough electrons. Electrons that are available to be built into infrastructure environments will be deployed rapidly through aggressively bid processes. That means 10, 20, 50 megawatt sites out of the gate, 250-plus megawatt sites from 2022 to today, and gigawatt campuses in the future. The 250-megawatt and 50-megawatt sites still produce value across the spectrum of deployed compute environments. The next phase is continued proliferation and distribution of intelligence as an economic good. Our job is to ensure communities that didn't benefit from previous technology waves are not left behind and to be the developer to bring and distribute that infrastructure to those places, on behalf of the communities and our shareholders.

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

Brian, to add, there's a study called 'Rethinking Load Growth' from Duke University. It discusses that across the 22 largest power grids in the nation, there's between 76 and 125 gigawatts of headroom that is unlocked and available if it can be curtailed between 0.5% and 1.5% of the time. So while there's talk that the grid is overtaxed, in reality there's about 100 gigawatts of headroom if you can curtail a limited number of hours. By working in smaller jurisdictions and discussing the ability to add interruptible loads like Bitcoin mining, it changes the conversation. We're excited about gigawatt campuses and 250-megawatt campuses, but also the 20s, 50s and 60s we have. We've had conversations with tenants specifically asking how much they can buy—60 megawatts or more—between now and 2027. Demand is everything you have as fast as you can get it. For us, it's become more of a sorting process to ensure the financial capability of the offtake and the financeability of the project is well within scope.

Operator, Operator

Your next question comes from the line of Mike Colonnese with H.C. Wainwright.

Michael Colonnese, Analyst, H.C. Wainwright

First, it sounds like you will be taking a unique approach to data center construction with your vendor relationships and strategy. What do you think your estimated data center deployment timelines could be once you sign your first lease and the associated expected CapEx cost per megawatt for these builds? It sounds like you might have an edge—just want more information around that.

Harry Sudock, Head of Investor Relations

Thanks, Mike. We stay conservative in our guidance. From lease signing, we're looking in the 14 to 18-month range for delivery. That's a function of first data hall versus last data hall, project size and other variables. We plan to be conservative and deliver aggressively on time and on budget initially, then over time smooth and innovate across supply chain and delivery to compress those timelines project by project—similar to the learning curve we experienced in Bitcoin mining deployment.

Michael Colonnese, Analyst, H.C. Wainwright

Is there any CapEx advantage using this more modular factory-based data center construction approach? Or is it pretty much in line with what you've seen in other builds?

Harry Sudock, Head of Investor Relations

I think these builds will be in line from a build and deploy perspective. What we're looking to do is create an offering with the best total cost of ownership for the client over multiple refresh cycles. That means delivering a total cost of ownership differential not just when they put their first GPU in, but when they refresh it repeatedly.

Michael Colonnese, Analyst, H.C. Wainwright

Very helpful. Second question: regarding GPU ownership versus colocation opportunities, would you consider pursuing GPU cloud service opportunities based on current economics and implied return profiles? If so, which assets in the portfolio make the most sense for you to own your own GPU fleet versus going the colocation route?

Harry Sudock, Head of Investor Relations

If you rewound the clock four or five years when H100s hit the market, nobody would have thought they'd be renting for the prices they are today. That's a positive sign for the GPU-as-a-Service model and overall AI compute demand. Ultimately, we want to deliver stable, high-margin and long-duration cash flows that colocation and tenant relationships provide. We won't say never to other opportunities, but we're focused on executing the projects directly in front of us and then considering other opportunities further out the curve.

Operator, Operator

Your next question comes from the line of Matthew Galinko with Maxim Group.

Matthew Galinko, Analyst, Maxim Group

How do you manage multi-site risk across your portfolio? In other words, would you be comfortable with a single tenant saturating your pipeline, or would you look to have multiple of these multi-site agreements?

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

I appreciate that question. It really depends on the credit quality of the tenant. We've had a tremendous amount of inbound inquiry. If we do a Neocloud deal, we can print a higher rent figure but it often requires a wrapper that comes at an equity cost. It includes a cost to the long-term shareholder value. So it comes down to balancing credit quality and risk with the ultimate value for our shareholders. We don't have an aversion to concentration risk if it's the right investment-grade tenant, but we'll evaluate on a case-by-case basis. We certainly don't want overexposure to a Neocloud that forces us to issue a large equity-like consideration just to get the deal financed.

Matthew Galinko, Analyst, Maxim Group

Makes sense. As a follow-up, 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? How do you envision moving those assets?

Gary Vecchiarelli, President and Chief Financial Officer (CFO)

Yes, we have older XPs and some recent S21s there in addition to the newest immersion miners. Naturally, over time we will retire some of those ASICs, but we'll monetize them while they still have useful economic life rather than waiting for them to completely die. We can be opportunistic—if Bitcoin spikes, older rigs can spike in value and we could move some. Our team evaluates in real time the best use of those mining assets and which sites to host them. It's an ongoing operational decision.

Operator, Operator

Your next question comes from the line of Jon Hickman with Ladenburg.

Jon Hickman, Analyst, Ladenburg

I'm a little confused about the ability to provide 0.95 power and still meet the needs of a community that is used to getting all the power they want whenever they need to turn on their Bitcoin mining. How do you balance that out?

Harry Sudock, Head of Investor Relations

Thanks, John. We work hand in glove with utilities and communities during contract migrations. Sandersville is a good example. In Georgia, the utility procures power on the open market over the contracted duration for the demand we represent at that location. By procuring that power, they've mitigated community risk for any pass-back because they know exactly where the power will be landed. On our side, we have a capacity factor obligation because we agree to consume a certain percentage of uptime under that power. Because of the procurement method and our consumption commitments, there is no pass-back into the community. Texas looks different because ERCOT doesn't contemplate the same expectation of interruption for the market at large, and that's why we targeted that market for large-scale AI campuses. In both cases, we're able to provide very firm and high uptime to the tenant while sourcing power in ways that are low impact to the community and infrastructure.

Jon Hickman, Analyst, Ladenburg

So the community in Sandersville that's used to this bitcoin mining operation, they're going to be okay as you transition it into HPC?

Harry Sudock, Head of Investor Relations

They're incredibly supportive of our business evolution.

Operator, Operator

Your next question comes from the line of Michael Grondahl with Northland Securities.

Mike Grondahl, Analyst, Northland Securities

Just wanted to get an update on how demand has evolved for some of your smaller sites and where that demand sits today.

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

I appreciate the question. We had a stand-up call today with a 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.' Demand is increasing. During that call, they commented they're aggressively seeking 8 gigawatts of capacity. We're seeing 500-megawatt or 1-gigawatt minimum thresholds start to drop because many mega sites are years away, under discussion or leasing. There's also risk with behind-the-meter power generation—many companies that pursued that approach are facing energization delays or utility costs that are double or triple what we're seeing for grid-connected sites. There have been constructive conversations at the White House about hyperscalers generating additional power and pushing it back to the grid, but the Duke study I mentioned earlier identifies 100 gigawatts of headroom on the grid today. Using a portfolio approach—incremental components and interruptible loads paired with firm loads—can unlock capacity previously overlooked.

Mike Grondahl, Analyst, Northland Securities

Sounds good. Nice to see demand moving down to some of those smaller sites.

Operator, Operator

We have time for one more question, and that question comes from James McIlree with Chardan.

James McIlree, Analyst, Chardan

I was hoping you could address how you're looking at managing and securing talent for engineering as well as labor, particularly over the next 12 to 18 months and into the next phase of build-out. Talent specifically for building out the next wave of data centers?

Matthew Schultz, Chairman and Chief Executive Officer (CEO)

That's a fantastic and very real concern. We've done data center consulting and development with firms like McKinsey and others. One thing identified is labor bottlenecks in markets like Texas—demand for plumbers and electricians over the next 36 to 48 months is significant. We've taken steps to mitigate that. For example, our development partners and factory-based approach eliminate about 70% of on-site labor. Sandersville will have one very large construction team of roughly 400 full-time construction employees rather than 4,000 to 6,000. It's an assembly-line process: build in the factory to specific chip manufacturer requirements, then install in the field. We have engaged McKinsey and others to assist in the road map and have had strong inbound inquiry. Historically we hire locally; many trades in our jurisdictions aren't chasing data center builds around the country but want to work at home, and we have a significant pool of local talent to support our needs.

Operator, Operator

And that concludes our question-and-answer session. I will now turn it back over to the presenters for closing comments.

Harry Sudock, Head of Investor Relations

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, Operator

Ladies and gentlemen, this does conclude today's conference call. Thank you for your participation, and you may now disconnect.