CleanSpark, Inc. (CLSK) Earnings Call Transcript & Summary

Ladies and gentlemen, thank you for standing by. My name is Colby, and I'll be your conference operator today. At this time, I'd like to welcome you to the CleanSpark's Fiscal Full Year 2025 earnings results. [Operator Instructions]. Harry, you may begin your conference.

Thanks, Colby, and thank you for joining us today to review the fourth quarter and full fiscal year 2025 financial results for CleanSpark. We encourage you to review our earnings results press release, which was issued today and is available on our website. Our 10-K will be filed shortly. A webcast replay and transcript of today's call will be added to our website once available. On the call today, I am joined by Matt 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 a 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.

Thanks, Harry. Good afternoon, everyone, and thank you for joining us. I'm so excited to have stepped back into the role of CEO of CleanSpark this past August after serving as Executive Chairman for the past 5 years. In my first 100 days, the team has been relentlessly cementing our current leadership position in Bitcoin mining, while simultaneously positioning us to evolve our portfolio. We've also set a strategic direction for CleanSpark going forward as a digital infrastructure platform serving a wide range of compute opportunities. These opportunities include, but are not limited to, generative AI workloads, grid balancing through Bitcoin mining and high-performance computing broadly. I've also had the opportunity to meet with many of you listening to today's call. Your enthusiasm for the future of CleanSpark's business means the world to us and we're excited to execute our strategic plan and extend our track record of operational excellence into AI factories. As the company has matured, I'm inspired by a world-class team and operating business. Our strong balance sheet and most excitingly, our growing power and land portfolio across the U.S. and the optionality it represents. Together, all of these elements are evolving into a diversified compute platform to serve the needs of the next digital age. I've taken stock of what we built. And I want to share with you just how well prepared the company is for this moment in time. While Bitcoin mining remains foundational to our business, we recognize that our expertise in securing power, developing infrastructure and deploying at scale uniquely positions us to support the fast-growing demand for AI compute, a blended approach to growing and monetizing our portfolio serves to diversify revenue, enhance margin and build long-term shareholder value. 2025 was a year Clean Spark achieved escape velocity, reaching 50 exahash per second in operational hash rate with 100% U.S.-based infrastructure and run by our operations and technology teams. We delivered record revenues and demonstrated capital stewardship by not issuing a single share through an equity offering throughout this calendar year, all without slowing down our growth. I'm proud to share a few financial highlights from our 2025 fiscal year. We achieved record revenues of $766 million our gross margin was 55%. Now that's a 1% decrease year-over-year. This small decrease is actually impressive due to this being the first full year post having when the Bitcoin Block rewards were reduced by 50%. Our Bitcoin Treasury grew by nearly 62% to over 13,000 generated entirely from our wholly owned operated hash rate. This puts us in a fundamentally different position relative to treasury companies purchasing spot Bitcoin since we mine it at greater than a 55% gross margin, and we're actively monetizing our holdings. We now have a sustainable self-funded mining business, thanks to our industry-leading mining team. and they're backed by an innovative digital asset management operation that's generating meaningful premiums and leveraging our treasury balance as a truly productive asset. We're in the process of deploying the 19,000 S21 XP immersion units that have an industry-leading 13.5 jewels per terahash. It's beginning this quarter, and we expect that process to be complete in calendar Q1 of '26. Now while this time line is a bit longer than we had initially contemplated, our priority was a comprehensive portfolio review to ensure that we would not consume any AI applicable megawatts with this deployment. We have always had an infrastructure first thesis we avoided the asset-light strategies of past cycles, and we prioritized control of power and infrastructure given the fundamental scarcity we're now seeing borne out in the market. Scaling our mining business required securing and developing a world-class power and land portfolio and growing significant supply chain, engineering, construction and operational capabilities. all highly relevant as we evolve into AI data center development. Today, we have more than 1 gigawatt of power under contract live in our data centers and infrastructure. Additionally, we have nearly 300 megawatts in Texas, fully contracted and scheduled to begin energization in early 2027, coupled with a multi-gigawatt pipeline of additional near-term opportunities. Importantly, many of these locations are excellent candidates for AI campuses while others are best positioned for Bitcoin mining, load balancing and securing the grid. Our objective is clear. to deliver each megawatt to its optimal use case. We have always had an internal philosophy of people first. As we look to expand our business, that was true in the earliest days of microgrid development. It was true as we grew into a Bitcoin mining company. It was clearly a winning strategy when we hired Taylor Monnig to lead us to the forefront of immersion cooling. And most recently, it remains true as we added Jeff Thomas to lead our AI data center initiatives following his successful tenure as President at Humane. We've accomplished 3 key initial steps in our business evolution thus far with Jeff on board. The first thing is we reviewed our diverse portfolio to identify the most productive use of every single megawatt Second, we secured a 285-megawatt site in Texas with the explicit intent of building an AI factory for a high-quality tenant. And three, we are aligning and expanding our internal team in conjunction with market-leading partners, to deliver projects on time and on budget that meet the exacting needs of offtake customers. When we took a close look at our facilities, it became clear that our 250-megawatt site in Sandersville, Georgia, provides an immediate opportunity to host a large-scale tenant. Other sites surrounding the Atlanta Hartsfield Airport, totaling over 100 megawatts with ready access to fiber are already in extremely high demand. In Texas, the site we recently acquired just outside of Houston will be the location of our first exclusive exclusively purpose-built AI factory. We hold 271 contiguous acres of land located on a regional fiber backbone and have executed 285 megawatts in long-term power supply agreements that have already been fully approved by ERCOT. Better still, the site is located near several high-capacity natural gas pipelines, which are being evaluated for industrial scale behind the meter generation opportunities. This purchase positions us to deliver scalable, resilient and energy-efficient capacity to meet demand from AI, cloud and enterprise workload and represents a key step in our long-term strategy to leverage our vertically integrated infrastructure first model. While this may be our first purpose-built facility, it certainly won't be our last. The entire team is focused on first securing tenants for [ Santillan ] Houston, which will then drive efforts to take the projects from commercialization to commissioning. Long-term tenants represent a superior risk-adjusted profile return profile, pardon me, for these assets rather than direct GPU exposure initially. Similar to past industrial revolutions, AI represents a new ecosystem. Power companies, chip companies, hyperscalers, infrastructure technology providers and others are all collaborating, and we're in direct discussions at every level to deliver maximum value for our customers and our shareholders. Jeff has been building the full life cycle playbook for AI campus development and operations that best serve this ecosystem. Together, his growing team is already vetting potential tenants building high-quality site commercialization plans for our pipeline and defining our project delivery road map. As part of those efforts, we entered into a memorandum of understanding with Submer, a global pioneer in liquid cooled and prefabricated data center solutions. Its end-to-end capabilities spanning from liquid cooling systems and mechanical, electrical and plumbing modules to full facility builds set new benchmarks in energy efficiency, density and sustainability, making them an ideal partner for CleanSpark's growth strategy. This relationship is our first step in taking elements of the construction process away from the data center and putting them into the factory with approved reference architecture designs to support a broad range of tenant requirements. Together, we're working on an infrastructure platform that integrates power generation, data center development and AI service delivery. Under this framework, CleanSpark focuses on selecting developing, building and operating AI-focused campuses while Summer will offer its technology and expertise as a strategic vendor in delivering sustainable modular data center systems. Meanwhile, we completed our largest financing ever with a $1.15 billion upsized 0% convertible note. Gary, our President and CFO, will discuss the finer details and numbers momentarily. But before I pass it over to him, there are some elements I'd like to highlight. The terms are even better than our prior rates in December 2024, with the same 0% interest rate, a higher 27.5% conversion premium and a 6.25-year term. This financing provides the resources to expand our power and land portfolio, exceed our first AI deployments and continue investing in strategic growth opportunities. And as part of this transaction, we bought back $460 million worth of our own stock, more than a 10% reduction in outstanding shares. We've once again bet on ourselves and we will succeed the Clean Spark way. With that, I'll hand it over to Gary to take you through the financial results, both for the quarter and the full year. Over to you, Gary.

Thank you, Matt. I'd like to start by reviewing the numbers for the entire 12-month fiscal period, which was a landmark year for CleanSpark. Our revenue grew more than 100% year-over-year to [ $766.3 million ] with almost 8,000 Bitcoin produced. The major driver of this increase was due to a combination for our growth in exahash and Bitcoin price. Our full year gross margin was 55%, which we're particularly proud of, given that this was the first full year post having. These margins remained relatively in line with the prior year which is attributed to the significant increases in efficiency our fleet had over the last 12 months. Also contributing to our gross margin consistency is our average marginal cost per bitcoin, which was slightly below $43,000 for the fiscal year, while our average revenue per bitcoin was approximately $98,000. Our margins and cost per bitcoin represents the strength of our infrastructure quality, our world-class teams and commitment to managing our business to profitability and margin rather than any single operating metric. Our high margins translated to an adjusted EBITDA of over $800 million, which I must point out, does not adjust for certain noncash items such as the mark-to-market on fair value of Bitcoin. When normalized by excluding our gain on the fair value of Bitcoin, the adjusted EBITDA from operations would be approximately $305 million, which represents a net margin of approximately 40%. Additionally, the combination of increases in margins and fair value of the 13,000-plus Bitcoin we have on the balance sheet contributed to a significant positive net income of about $365 million. Looking at the most recent quarter-over-quarter performance, we also saw significant gains between the third and fourth quarters. Our revenue increased by approximately $25 million or 13% in Q4 versus Q3 and our margins increased 2 points to 56.5%. It is important to note that we achieved 50 exahash in June. And while that remained our operational high for the fourth quarter, we still experienced increases in revenues and margins because of favorable mining economics during the quarter. Our high uptime also allowed us to capture periods of significant appreciation in Bitcoin price. In the fourth quarter, we recognized a slight net loss compared to the third quarter. This was due to a much larger gain on fair value of Bitcoin during the third quarter and noncash tax adjustments recorded at our fiscal year-end. Our adjusted EBITDA margins also saw similar changes which is inclusive of the noncash mark-to-market adjustment on fair value of Bitcoin. However, when adjusting any noncash mark-to-market effect, our normalized adjusted EBITDA was $97 million for the fourth quarter, a 25% increase over the $78 million normalized in the third quarter. This translates to margins of 43% and 39%, respectively. Going forward, we do expect that our professional fees payroll and G&A line items will increase as we execute on our AI strategy. Additionally, I will point out that the AI data center business comes with stable cash flows and high margins, both of which will help CleanSpark. Through the peaks and valleys of Bitcoin mining economics. Our escape velocity translates to operating leverage. We have developed scaled data center infrastructure that is delivering revenue and margin necessary to self-sustain and further support incremental investment in AI data center capabilities as we evolve into our power, land and compute platform. Turning our attention to the balance sheet. I want to point out that we are one of the first, if not the only company which has a scaled cash flowing business that is also using Bitcoin as a productive capital asset. The utilization of our Bitcoin stack resides and the team we refer to as digital asset management or DAM. The fourth quarter was the first full quarter of dam activity, and we are extremely excited to share in more detail the steps we have taken in our crawl phase. Two initial strategies rolled out by DAM are our spots and yield strategies, both utilize covered calls, but spots is designed to optimize for the cash needs of the business. while yield is designed to generate go-forward risk adjusted output from our treasury holdings. Given that we are monetizing a significant portion of our monthly Bitcoin production, the spots strategy delivers a tactical uplift to cash generated on a weekly, monthly and quarterly basis. This program functions smoothly because of the consistent output from our world-class operations and strong uptime. We are able to utilize this approach because of the investment we have made in making DAM a true institutional grade platform. It began with a comprehensive RFP for a range of products that you have heard us discuss on prior calls and executing these option overlays requires a disciplined approach to risk management. Rather than selling bitcoin through the Spark, we utilize at or near the money covered calls to generate both option premium and realized proceeds. If and when we ultimately get called away on these contracts. Our yield strategy utilizes covered calls as well, but instead of high delta short duration, we shift Delta and extend or ladder turn to reduce the likelihood of exercise. Under our yield program, we saw an annualized yield of approximately 12% on a blended basis. In addition, as we scale our strategy and increase the volume, we believe there is room to incrementally increase the annualized yield and cash generated, potentially significantly. While the fourth quarter represents a period when we're still in the crawl phase of the strategy, we were nonetheless able to generate a total of $9.3 million in premiums. To illustrate what that represents, our average spot Bitcoin sales price for the quarter was $111,721. However, when considering additional premiums generated for Bitcoin, of $4,184, the all-in effective cash generated for Bitcoin was almost $116,000, a material uplift. One of the early wins for the DAM team was the successful monetization of costless Bitcoin repurchase options received as part of a Bitcoin minor procurement contract from the third quarter. This was an excellent example of how our investment in the asset management function can help us to complete the arc of opportunities driven by our world-class mining operations. While our mining operations drove large and preferential terms to obtain mining rigs, DAM was able to monetize that option, which would have otherwise expired worthless, driving $7 million of additional cash to the balance sheet. Due to the performance of DAM to date, we have increased the volume of transactions subsequent to our fiscal year-end. In October alone, we traded more contracts than the total number of contracts traded during the entire fourth quarter. Additionally, we generated over $5 million in cash premiums for the month of October alone. The last leg of our current strategy in bolster writing puts. The put transactions we entered into are cash secured primarily using the premiums previously generated under the spots and yield programs. While this cash corpus is still growing, we saw analyzed returns at 8% on the put strategy. These 3 strategies do 2 things: First, they integrate in our operating business with the enhanced sale of production; and second, creating capital flywheel as they relate to our balance sheet. I would also like to add that the results we are seeing in DAM do not necessarily translate directly to telling the story via U.S. GAAP accounting. While all pieces are reflected across the income statement and balance sheet, there are certain punitive treatments of noncash mark-to-market valuations at contract expiry. What we think is important about these tables that once again, CleanSpark is at the cutting edge of real non-hyperbolic strategies, paired with full market-leading transparency. These tables can be found in the Management's Discussion and Analysis section of our Form 10-K. I want to note that U.S. GAAP rules separate the accounting for cover call exercises into 2 different line items for what is in substance a single transaction. These 2 line items on the income statement are loss on derivative contracts and gain on fair value of billing. This is important because there are 2 sides of the same transaction. For example, the difference between the spot price at expiry and the strike price is shown as a loss on derivative contracts. While the corresponding markup in Bitcoin value to the spot price is recorded separately as a gain on fair value of Bitcoin, offsetting that noncash loss with a noncash gain. Taken together, they reflect economic gain outcome of our covered call program, which continue to generate attractive risk-adjusted returns. I also want to point out that the binding option was effectively costly to us. However, GAAP required us to bifurcate a portion of the ASIC contract to the option value, even though the contract didn't explicitly state a value. That value of $6.8 million was recorded at contract inception in the third quarter. As the option ultimately expired out of the money, had we not taken steps to monetize the option, we would have had a noncash write-off of that $6.8 million or instead, we generated almost $7 million of cash on that option, which under GAAP considered it to be a net gain of approximately $200,000 even though we ended up with 7 million more cash -- bank at the end of the day. The overall takeaway is that the digital asset management strategy has met and, in fact, exceeded our expectations this and become a second source of cash generation to the business. We are looking to increase -- the team to allow for greater volume and more complex derivative trades which we believe will not only grow the total cash generated from premiums, but also maintain attractive yields. On a final note, I'd like to take some time discussing our capital strategy. Our focus is on building a capital stack, which minimizes dilution. This starts with the sale of monthly Bitcoin production to cover a monthly OpEx. We also have Bitcoin back lines of credit with a total capacity of $400 million, we will continue to use the lines of credit opportunistically in the marketplace for accretive acquisitions. And as we previously mentioned, we issued a $1.15 billion convertible note with a coupon of 0% and a conversion premium of 27.5%. Proceeds from this transaction were used for several purposes. First, we bought back [ $260 million ] of our stock which represents a reduction in our outstanding shares of 10.9%. The stock buyback not only helped facilitate the convert, and we saw this as a bet on ourselves as we are evaluating increasing given the opportunities in front of us. Second, we used over $200 million from hash rate to pay off our lines of credit. It's important to note that we have access to the full $400 million line available to draw down any time on terms we continue to believe our market leading. The remaining net proceeds from the transaction will be used to do what we have a prudent track record of doing, and that is hunting for power and land. The acquisitions of Power and land, such as the most recently announced transaction in -- Texas are expected to be primary used for our AI data center strategy. While we are in the early innings of our AI data center journey, the market is moving quickly and so is CleanSpark. Our conversations with offtakers are ongoing, and it is not a matter of if but when we will have our first customer. Details regarding financing of our data centers will be coming in future periods. However, I will tell you this. There is an abundant amount of capital at a much lower cost of capital than previously available to our mining business. Our venture in AI data centers will open new pools of capital, allowing us to benefit from the significant levered rates of return the market is providing. To close out another strong and defining quarter for CleanSpark and to discuss how these results position us for what's next, let's return to our Chairman and CEO, Bet Schultz.

Thanks, Gary. As I listen to those results, I can't help but think back to the earliest days of this company and the journey we've all been on together. Our fundamental thesis on being infrastructure focused and people first has served us incredibly well. They are 2 of the reasons we have such a meaningful opportunity in front of us today to grow into an infrastructure and compute platform that maximizes the value of every megawatt. The task in front of us is clear. We're working to secure tenants at our 2 initial flagship AI-ready locations while simultaneously expanding our land and power footprint to meet the market's insatiable demand. These efforts are made possible by our strength as a scaled Bitcoin Miner, our capital markets rigor and critically, our company's cultural focus on operational excellence. This past summer, our operations team coined -- Amato, be the standard. I have the pleasure of having them present to me what that phrase meant all of them. And I commit to you that in each of our endeavors, you can count on Clean Spark to continue to be the standard. I want to take a moment to thank our entire team for their tireless work. I'm beyond grateful to our shareholders for their trust and I truly appreciate all of you for joining us today. With that, I'll hand it back to Harry to lead us into Q&A.

Thanks, Matt. We will now open up the floor to questions from the analyst community. Operator, please provide instructions and manage the queue for the Q&A session.

[Operator Instructions]. Your first question comes from the line of Brian Dobson with Clear Street.

Just a quick question. There's been a considerable amount volatility in the stocks as of late. Perhaps you could take this opportunity to -- so give us a little bit of color on the types of conversations you're having with potential clients and your outlook for demand in the HPC AI space over the course of the next 2 years.

Yes, absolutely. Brian, thanks for calling in, and thank you for the question. I can tell you that we've had extensive conversations. Now I posted on my social media. Our whole team was invited to Northern California to spend some time with the team at NVIDIA. From that meeting, we've had subsequent follow-ups. And I can tell you that there is I don't want to say a bidding war, but strong multiple layer inquiries about Sandersville specifically, and we're starting to gain additional traction on the Sealy Texas site. So we feel like the demand is there. Obviously, there have been some delivery challenges and credit risk on some of the other peers that maybe haven't been able to perform to the expectations. But we're based on the fact that we're running a company with nearly an $800 million annual run rate at 55% gross margins, we have the cash necessary to get us to that next level. So we actually feel very optimistic about it.

Yes, outstanding. And as you're thinking about various campuses, what do you think about hearing Bitcoin mining with HPC campuses to provide, call it, power usage versatility or do you think that they'll be separate to start?

That's a really thoughtful question. We were invited by Jack Dorsey and his team to go to Dalton, Georgia and spend some time as they launch their new domestic manufactured ASIC, the Proto rig that's built by block. And it was a fascinating event but leaving the event, the CEO of the utility there in Georgia, grabbed Gary, Harry and myself and ask us to go to lunch. And he shared that there's about 120 hours a year that really causes problems for the utility. And he said, historically, they love Bitcoin miners because of the interruptible load. Now we've experienced providing that service in load balancing in many of our jurisdictions. I mean you've heard the stories about redirecting power in Georgia to a hospital when the hurricane hit or whatever the case may be. But the takeaway from the utility was they're interested and the fear that came from them was because Core Scientific is a big consumer of power there in Dalton, and it's historically been a flexible load. And the concern is that extra 120 hours a year when they need somebody to be able to give back. So what they specifically the request from us was to consider blending AI, HPC and Bitcoin mining. So a component of those loads remain interruptible. So we see it as a dual-pronged strategy. And I think you'll see a lot of our sites will serve both loads.

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

Congrats on the strong fiscal year here. First one for me on the HCC side. curious, what are some of the key development milestones that investors should be on a lookout for as it relates to the HPC strategy. It sounds like the near-term focus will be on deployments at your Texas and Sandersville sites. So if you get some more color there.

Yes, you nailed it, Mike. I can tell you I had a conversation not with a neo cloud or anybody like that. But actually, with the Senior Director of site development for a global hyperscaler last night on my way leaving here. And what he shared with us is their 2026 forecasts are so constrained that they're looking at alternative types of builds just to facilitate the needs for '26. The takeaway from the conversation was Sandersville and Sealy because both of them can be energized, Sandersville is live and active right now powering 11 exahash Bitcoin miners. But it could switch to a 200-megawatt critical IT load and be online in a reasonable period of time. And I think kind of a cool thing that maybe has gone unnoticed and that is this MOU summer. We don't historically -- I mean if you look at CleanSpark's pass, we don't announce MOU or LOI or anything that isn't definitive or concrete. It was really important to ink that with summer because of the way that they approach the business. Submer has approved reference architecture for AMD for NVIDIA and they build the entire MEP solution. So mechanical, electrical and plumbing with all the fiber runs, they take that out of the field, put it into the factory. So a company like CleanSpark builds the powered grayshell, we contract with Submer in the MEP solution for -- specifically to the reference architecture of the end user requirements. So speed to market is really, really critical right now. And having that modular approach, I think, is going to be a massive differentiator for us.

That's helpful color, Matt. I appreciate that. And then the second one is on the Bitcoin mining side. I know you mentioned some of these near-term deployments you guys are looking to install in the first quarter. Just reminded us of what your near-term expansion plans will look like for the Bitcoin mining business and existing site expansions versus any sort of new development opportunities on the greenfield side or mergers and acquisitions at this stage?

Yes. So I think what you're going to see is a migration of our Bitcoin mining away from areas that are closer to major metropolitan areas that are maybe more sensitive to utility rates and into more remote locations. There are a number of utilities that have either recently passed or are discussing the blockchain specific tariffs. To my point on the last question, that interruptible component of the load is in such demand that they give us favorable rates whether it's TBA or Wyoming or any of a number of different jurisdictions to have the offtake that allows us to flex and to assist the utility. So I think what you'll see is locations like Sandersville, locations like the Metro Atlanta sites in North Cross in College Park, et cetera. those will probably be prioritized for HPC AI because of the quick access to fiber, the low latency loads that they can serve, et cetera. And then the Bitcoin mining from those facilities will likely migrate out to some of the other locations. So to answer your question directly about scale, we're at 50 exahash per second right now. We have 6 exahash of the S21 XP Immersion miners. We had slotted out a deployment strategy. Now we use modular emergent cool data centers for the vast majority of those. When we secured the 100 megawatts in Wyoming, we actually beat out a hyperscaler because of the fact that Wyoming wanted to energize those megawatts today and not in 3 years. So we have the infrastructure purchase, delivered on hand, ready to roll to deploy these in short order. So I think what you'll see is between now and towards the end of Q1 '26, calendar Q1, you'll see that additional 6 exahash come online. Above that, what you'll see is as we do fleet upgrades, not in the $250 million capacity that we've historically done, but in a more disciplined, more thoughtful manner to ensure that we're protecting our share of the hash rate and supporting what we believe to be a national security issue, and that is ensuring that there's Bitcoin mining hash rate domestically. So we're going to take a real balanced approach at that. But you'll see us continue to grow. And really, the differentiator is just in the fact that we have right now one of the most efficient fleets in the world. And with the deployment of this 6 exahash of 13.5 jewel machines will have hands down the most efficient fleet around.

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

Congrats on a strong finish to the year. I wanted to ask with the 13,000 Bitcoin on the balance sheet around $1.2 billion at fiscal year-end. And with the recent financings that you've done, how should we think about the total operation to build this powdered land bank as you think about the opportunity to bring tenants in for HPC or to simultaneously grow your Bitcoin mining fleet as you were just discussing? And then I have a follow-up. Thanks for the question, Paul.

Around the Bitcoin stack really hasn't changed, right? To give you context, we consciously have stacked Bitcoin quite rapidly over an 18-month period, and that brought us about 13,000 Bitcoin on the balance sheet. And we believe that we're one of the only companies using it in the strategic ways that it should be used as a capital asset. So I think going forward, what you can count on from us is a few things. One, we'll continue to monetize the Bitcoin stack through yield strategies to generate some cash. Two, we'll continue to borrow against it to be opportunistic to draw down on cash to make sure that we're nimble in the marketplace and take advantage of accretive acquisitions. And we've always said that we're not ideological about the bitcoin balance. We're very strategic. And so if there comes a point in time where we needed to or we felt that the right thing to do is to part with a Bitcoin balance through sales, we most certainly would do that, and we were open to do that. Because, again, we've built this entire company and given the financial wherewithal on optionality. But I'll tell you that with those sales comes to punitive tax treatment because we have MIMO at such a low basis. we'll have to pay -- will be a cash paying taxpayer on those items. So we take those into account when we're looking at the stack. But overall, we'll continue to use this as a form of nondilutive capital.

Got it. And then turning to the MOU with Submer and at the explanation you were just giving on how you might break out the shell development versus the MEP componentry. How should we think about the potential economic impact of that? If you can give any color, just thinking about how pricing on some colocation deals involve yield on cost and, of course, still to suit can involve more capital but with a partner just looking for any additional color you could provide?

Yes. Great question, Paul. Thank you for joining. So the summer relationship really was born out of a prior relationship between Humane and Submer. And Jeff had a working relationship with Patrick, the CEO of summer. And we've also got summer infrastructure deployed in our Bitcoin mining side. So we're very comfortable with them. And I can tell you that the quality of the product that they deliver, it's much simpler in the Bitcoin mining side than some of the other modular immersion cool type companies. But because we haven't done a deployment domestically and they're just spinning up a manufacturing facility in Houston, I don't want to comment too much on what the cost per megawatt is, but I can tell you, in general terms, that the cost to build out a megawatt of mining infrastructure is about $1 million to do the same for AI and HPC according to the reference architecture required by the major chip manufacturers, it's closer to $10 million. We also know that the mechanical, electrical and plumbing, the MEP solution, is a pretty extensive, pretty robust build-out because you've got all those trades working inside a facility at the same time. Building these in a factory increases the speed to market by an order of magnitude and the initial representations are that it saves us anywhere from 10% to 15% over a stick built in the field deployment. So we believe there's cost savings and speed to market that give us a very unique competitive advantage.

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

I guess, Gary or Matt, I was hoping you could talk a little bit more about the Texas facility and just kind of -- you mentioned that it starts to energize in 2027. Is that energization? Is there steps along the way? And then longer term, as we think about that site, is there the ability to expand at that site or potentially grow with the customer?

Greg, it's Gary. I want to give you the rundown on Texas because I think it's a really exciting project for us, and it's the beginning of what you've seen from us across the Georgia, Tennessee and Wyoming markets, which we take a fundamental land-and-expand approach, which is we get a foothold, and then we know that once we have that toehold in the market, the opportunity to significantly extend our footprint is available to us. The energization schedule there is that the first 200 and change megawatts scheduled to come online first half 2027. And then there's 240-megawatt tranches in '28 and '29. But what's critical is that the counterparty that we purchased, the land and the contracted power from is also among the largest substation developers in the state. And so what we were able to step into are the long lead time items and the placeholders that they had on those components, giving us a high degree of build certainty to land the power on the site. The second piece is that, that site is fully accrued. And so when we look at the amortization schedule, we've already passed all of the regulatory hurdles that would typically be associated with a project in the state. The final piece of what you asked that I want to touch on is about expansion. And the ERCOT approval status wouldn't come with the expansion on grid that we're looking to accomplish there. But one of the parts that was most attractive not only to the power contract at that particular location and the service point from the utility that's going to be delivering to us, but it's also the parcel that's there. We have significant land capabilities to be able to digest more power in the same type of AI-based center footprint. That's going to be represented by the 285. And so we're excited about the scalability. Matt touched on in his comments, the behind-the-meter gas generation opportunity, but this is where we find ourselves at our true core competency which is being an opportunistic acquirer of land and power. Not only because we're able to locate high-quality assets for our portfolio today, but also for what those assets can represent to our business going into the future.

Okay. 100%. That was super helpful. I have a great Thanksgiving and talk to you soon.

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

Congrats guys on the progress here. First question here, as it relates to the AI readiness at Sandersville, just remind us if that site had for curtailment. And then if so, really kind of how much should we earmark for HPC versus mining if you intend to kind of have both at that site?

Yes. Thanks, John. So what's important to understand about the Georgia and the -- power specifically, is that it is not subject to forced certain at that location. It's part of why we didn't just -- we didn't just trip all and land with an AI thesis around the Sandersville assets. Those were also inbound because of the highly attractive nature of the Georgia Power markets more broadly. And so we feel great about the applicability of that location to the ultimate AI campus use case and how we balance that versus the Bitcoin mining is going to be the way we do everything, which is a fundamental return on investment profile. We take a measured approach. We're data-driven. And the early indication is that every one of those megawatts highly applicable for AI as its highest and best use. But we're going to remain data-driven across the analysis period for the asset.

Great. That's helpful. And then just as it relates to credit becoming a little bit of a concern out there, especially as it relates to Neo cloud customers. Just walk us through how you are thinking about the customer profile, if there's maybe bigger parts than hyperscalers now than maybe a couple of months ago when you guys were initially thinking about it? And then also hyperscaleresult backstops, is that starting to become a necessity? Would love to get some color on that.

John, thanks for the question. I'll tell you this about the financing. And I mentioned in my prepared remarks is that there's new pools of capital that are going to be available to us at much lower cost of capital. So we feel really good about that. We know we're going to introduce at some point, secured debt into the capital act. So we're closely monitoring the deals that are going on and the debt markets. But I'll tell you, the focus really first is to get that high credit quality tenant in there to make sure that we can get the best deal possible because, as you know, levered IRR is significantly higher, the more -- the higher the loan to value is. But I'll also tell you that while we might expect to get debt at about 80%, 85% LTV over time. We have no problem also being live bit more equity to the table, maybe at a 60% to 70% LTV for the first project or 2. Yes, that will decrease our levered IRR just a bit, but it also produce cash flows in the -- arm, which would also be helpful. But ultimately, to us, it's really going to come down to execution for which we still think that the industry hasn't proven but we're going to see that over the next 12 to 18 months, particularly as we bring our land and power to market. So I don't have the specific answer for you right now, but I'd tell you that we are confident that there's a number of options for us to get financing and attract prices and get the levered IRR that this market is offering.

Your question comes from the line of Reggie Smith with JPMorgan.

Congrats on growth in the quarter and on the pivot. I just had a question on the Sandersville side as well. I'm not sure if you guys talked about what type of CapEx would be required to operate at HPC if it's ready, like kind of moving ready now. And then I'm curious, I know it's early, maybe thought about the use cases, whether it would be used for training or entrance and whether that all plays any role in the price that you'd be able to get to kind of lease that space out. So like does inference pay more or generate more revenue per megawatt than training. Any insights you can provide at least around how you're thinking about kind of self appraisal of the site and what it could be worth?

Reggie, thanks for joining, and thanks for the call. You've been to that Sandersville site, I believe, on some of our show and tell journeys. And I think what's important to note is the facility, as you saw it, would not be a conversion to HPC AI. We have a phenomenal relationship with the Economic Development Director in the county. And so we secured an additional plot of hundreds of acres of land that's immediately adjacent. So what you would see would be construction that is parallel with Bitcoin mining continuing and when we're ready to energize, we literally flip the switch, de-energize the Bitcoin mining and migrate that out and go to compute. Now specific types of compute. Jeff has built a model. Obviously, you have the Giga campus, which is large-scale training, those are generally close to 1 gigawatt and above. Then you have the mega campus, which is that kind of sweet spot 200 to 800 megawatts. And that's generally perceived to be kind of a combo site where it's inference and training or primarily inference depending on the offtake client. The last mile or the low latency, real -- critical sites like what you've seen in and around Metro Atlanta would be kind of the exception to that rule. And those would obviously be low latency inference-type operations. So this is going to be -- I think Sandersville going to be an interesting case study because, quite frankly, the demand that we're seeing is for multiple 190 to 200-megawatt critical IT loads and the off-takers are asking for 2026 delivery. So there's some real challenges in getting that tipped up in time. But as we saw, even with companies like Meta, for example, they're putting tents up and using behind-the-meter gas at $0.12 a kilowatt hour because the demand is such that there is no sensitivity to those utility rates. So we really feel like we're uniquely positioned in this Reggie. And you -- when you first launched coverage on CleanSpark, we talked about the fact that when Clean Spark entered the space, there were a handful of household names that were the standard from Bitcoin mining. We mined our first Bitcoin in December of 2020. And as of today, we have more hash rate in the United States of America than anybody else or of time is second to none. And I think you can see, you can count on seeing that same type of operational excellence and efficiency rolled into our next strategy. And Jeff is just the perfect guy to lead those initiatives.

And if I can get 1 more in. I'm not trying to nail you down to a time line, but I kind of read between the line but I think about cipher and they purchased a property in Texas a year ago and kind of just now announced a deal. And obviously, it takes a while to spot these types of things out. But I'm curious like how you're thinking about it, if it took you a year to sign a deal, would that be your satisfactory or kind of disappointing based on which you're seeing from a demand perspective now like you just should we think of something much sooner than that? Like any color you can help on how you're thinking about that internal personally, that would be great.

Yes. So that's a phenomenal question. And I can tell you that you called me, I was at my kid's basketball game, you called me when Corsico, we've announced their deal. And we talked about what the demand portfolio or the demand profile looked like back then. And at that point, it was we're going to convert these megawatts and we're going to identify a customer. I think there's been a complete paradigm shift in the space. And now you have customers knocking at the door because they have loads that need to be served very rapidly. So what I can tell you with -- I would say, a strong amount of certainty is you'll see a lease executed much quicker than what you've seen in the space. and the flexibility that's now come, I mean, we look at some of our peers that have extended their energization schedules because they're falling behind on construction, et cetera. The hyperscalers and the end users that we're having conversations with. We've made it abundantly clear. We're constrained like anyone else for the MEP side, but we have a distinct advantage. So I think what's likely to happen, and Reggie, quite frankly, there are 2 different off-takers that want to sign lease agreements by year-end. Is that going to happen? It's hard to say. But the demand is there. It's real. And as I mentioned in earlier comments, we were working on the script in the slide deck that we showed today, and I left here at 8:00 last night and the Global Director of site selection for a hyperscaler was calling me wanting to confirm that they were still in the running. So I don't think there's any question that you're going to see a lease much quicker than a year.

Your next question comes from Brett Knoblauch with Cantor Fitzgerald.

Maybe, Matt, just on the land and power side of the equation. Could you comment on what you had to pay for the new site in Texas that you guys just announced -- and I think you guys one out there that are looking to go out and find additional land that is energized soon. I think all of your peers need to hyperscalers probably looking to do it themselves. Like I guess how hard is it, how expensive is it? And how much is there out there that you think you can go out and buy that is kind of turnkey ready similar to the site that you guys announced it?

So yes, I could say what we paid for it, but I'm not going to. And I'll tell you why. There are some very fertile hunting grounds in the ERCOT region, and we don't want to price ourselves out of the market. What I can tell you is that the acquisition cost was a combination of equity and cash. We obviously filed the proper -- the appropriate filings for the share issuance. But the purchase price of that land and power came in line with what you're seeing in the market towards the low end of that range. Our advantage, I think, in securing land and power and speed to market is really because we've continued to state that Bitcoin mining is going to be a part of what we do going forward. And 18 months ago, when we were invited to sit down at Maralago with Donald Trump during his candidacy, he brought in Senator Hagerty from Tennessee. And because the question that came was, can Bitcoin miners actually plow the road, so to speak, can Bitcoin miners go in and monetize megawatts for a utility that needs to generate revenues now or for an energy developer that needs to monetize their power while they're waiting for an intercept agreement. And so Mr. Trump asked Bill Hagerty, can Bitcoin miners do that and what he said is, unequivocally, they not only can they but they do in TVA and CleanSpark is one of them. And so when we talk about Cheyenne, we're driver 9 iron across the street there from Warren Air Force Base, and there's another $1 trillion company, trillion market cap company that's in our in neighborhood. They were bidding for those megawatts. And we won -- we didn't win because the utility thought that our balance sheet was prettier or we were a better credit risk. We won because we said if you sell us those megawatts, we'll start buying them in 6 months, not 1.5 years or 2 years. So I think long answer your question. I think being a Bitcoin Miner with diverse mining portfolio and the flexibility of the modular deployments that we've done, some of the sites that you've actually seen. It gives us an advantage to jump in, monetize those megawatts on a small portion of the campus while we're tilting up the powered shell in the background. So I think our speed to market is complemented not only by the modular approach with the Submer partnership to a signing to go in and buy the power today.

Your next question comes from Jim Missouri with Chardan Capital.

Is Sandersville the only existing mining site you've identified for critical IT applications or the first one and there's going to be others?

Yes. Jim, thanks for the question. The answer to your question is B, it's the first one. the inbound inquiries we've had for the 100 megawatts surrounding the Atlanta Airport are second in urgency only to Sandersville. And Sandersville is because it's 250 megawatts energized operating today. the demand for College Park and Norcross is because it's low latency in the most dense compute environment in North America outside of Northern Virginia. So there's a tremendous amount of demand there as well as some of our sites in Tennessee. So it's really just a sorting process. And as we mentioned in our prepared remarks, we've done a portfolio analysis to kind of determine we don't want to move Bitcoin mining infrastructure into a facility that's going to be rapidly pivoted to AI HPC deployment. So I think the answer is Sanders Village is the low-hanging fruit that everybody wants. The Metro Atlanta stuff is second, and then we've got a whole bunch of third place sites.

And the way you described it, it sounds like that flipping of the switch from mining to AI can take place before the Sealy facility is energized. Am I understanding that correctly?

Yes. So think about it this way, Jim. We're -- our facility in Sandersville is purpose-built Bitcoin mining. We have a couple of hundred acres adjacent. We're going to build on that land, while we're still mining. Now the speed to market, really the summer is a big differentiator. As I mentioned before, there are hyperscalers that are popping up tents because they need access so quickly. So I think it's a relative question. The Sealy part, the Sealy project is very appealing because we've done all the analysis, all the engineering is done. We've gone to the levels of completing the survey and finding where there are easements for the gas lines on site, et cetera. So we can configure the footprint based on the needs of an end-use customer. So we spent a great deal of time in NVIDIA with some of their teams, and they have a giga site. They have all the reference architecture for a GB300 deployment for a gigawatt of power and everything is detailed down to the inch of fiber runs. So that type of build is obviously much more detailed and going to take a longer period of time than tilting up a powered shell and slotting in a modular solution like you'd see from a sub or like you'd see from a company like Integra out of Houston. There are a number of these companies that provide that full MEP turnkey solution. So I think what is likely to happen is we'll probably execute on both simultaneously. The delay on Sealy, and it's not really a delay. It's just the inertization schedule on ERCOT is fixed. The cool thing about that is the large load studies are done. There's no it's -- just when. And the first 207 megawatts energized is the first half of '27. Their commitment is April, but they have flexibility for the first half. So I think the conversations we've had with offtakers for Sandersville, they're looking to get something in the book fast with Sealy, it's also high demand and with the understanding that by Q2 '27, it's energized and you can build in the meantime.

Understood. And just 1 more, if I might. You talked about increased expenses and given that as well as the recent prices of Bitcoin, will you need to sell the entirety of your Bitcoin production in order to cover your expenses?

Not at all. We're generating 600, 700 bitcoin a month, and we're doing sort of 54%, 55% gross margin. So the expenses we talked about and as we're building, I think it really depends on the lease we put together and the ability to leverage that lease for financing. But what we're seeing is that the -- depending on the credit quality of the end-use tenant, the LTVs are anywhere from 15% to 30%, having just put up the inverse of that, I'm sorry, 70% to 85%. Having just put up that $1.15 billion 0% bond, we're sitting on a pretty healthy stock of cash. And then as Gary mentioned in his prepared comments, we have $400 million in low single-digit interest unused capacity on Bitcoin back to credit facilities. So I think you'll see us take more of a hybrid approach rather than sell the stack. And then the last thing, with regard to the compressed margins in the environment, having the highest uptime and the most efficient fleet in the nation means that as energy prices press up margin compression happens to everybody. We just happen to make more money out of the same megawatts because of fleet efficiency and uptime. So I tell the story, it's like when my grandfather told me when 2 guys are camping in a tent and a bare walks into the camp site and the 1 guy puts on his shoes and the guy says, why are you putting on your shoes, you can't outrun a bar and he said, I don't have to run the bear. I just have to outrun you. And that's really what we see as our Bitcoin mining advantage. They're still industrial scale miners operating fleets greater than 20 jewels per terahash with not significantly better power pricing than we do. So we have a ton of flexibility and I really like the position that we're in to continue using Bitcoin Mining.

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

As you might imagine, most of my questions have been answered. But I was just wondering if you could maybe opine about there are others in the space that are trying to do the same thing that you're doing, take taking Bitcoin sites. And moving them over to HPC and AI and they've been telling us they're going to do this, and it's been a year has gone by and there's no like lease. Why would -- could you opine as to why it would be taking so long when there is so much demand?

I'll tell you from my perspective, and then I'd certainly invite either of my colleagues to chime in on it. What we learned when we spent some time with NVIDIA and AMD, there's -- there are very specific reference architecture that is required for specific clusters. And I think that the challenge that we're seeing, and I'm certainly not casting aspersions on any one strategy. But I think if there is so much demand for hyperscaler but they want a specific cluster, be that the new Google chips or AMD or NVIDIA. The site needs to be specifically designed for those clusters. And I think building a site and then suggesting that it's flexible for somebody else to reconfigure or modify it, it could be useful, I think, is a little bit of a challenge. So to have it purpose built to the specific architecture of the offtaker, I believe, is a real advantage. And having all these sites that are already energized that we're currently using those megawatts to mine Bitcoin, give us that flexibility. I'm not in a rush. I don't have to do anything, quite frankly, until we have a lease I don't even have to start construction because I want to make sure that it's built to suit for the offtake customer. So I guess my perspective, John, and again, I invite Harry or Gary to comment. I think that build and they will come mentality doesn't apply if it's not to the specific architecture requirements of the end user.

I'll just add one, Jon, which is just that the market today is different than the market a year ago. The demand profile has accelerated. The crunch for power is tighter than it was. And so we're seeing some of the hesitation that some of these offtakers might have had 12 months ago. The sense of urgency is just more significant, and that's going to represent a difference in execution time line today than they would have looked like back then.

Okay. And then I just have 1 question. On Sandersville, when you get the AI part build and you want a static the switch, what happens to the Bitcoin mining site? Would they have more power -- would you?

Yes. No, a couple of opportunities there. First and foremost, the ASICs would be migrated to a facility that needs them right now, and there's always plenty of demand for that. the facilities that we built. And Jon, I don't remember if you visited -- on our Analyst Day -- the facilities standards. Okay. So they're identical to what we built in Jackson, Tennessee as an example. So the cool thing is we build all the foundational stuff and then what goes vertical is basically bolt together. So we have the ability to repurpose those buildings based on any number of different factors. But no, it wouldn't be a write-off in mothball it. It would be repurposing those assets for deployment elsewhere.

But will you need more power?

Yes, for sure. That's why we would move it somewhere else like, for example, Wyoming or Tennessee or even other sites in Georgia, we would, for sure. Now Sandersville is a bit of an anomaly because there are some opportunities for power expansion there. And that's something that's still out for discussion, but the demand is significant.

And with your last question, it comes from Nick Giles from B. Riley Securities.

Great. You spoke to a 1 gigawatt pipeline. And I was hoping you could break that down a bit. I mean, how many of these opportunities would you describe this late stage? Or how soon could we see those drop down? And then which of your existing power markets do you see most of these opportunities?

Yes, Nick. Great question. And I want to give you a historical example to kind of illustrate why we don't always give direct pipeline granularity as our business. So if you look back to the prior quarter's call, we talked about pipeline and it was contemplated in that environment. What wasn't contemplated in either of those numbers was the 285 megawatts that we purchased in Texas. And that's because the relationships that we have across the utility and infrastructure space are diverse, and they're all very warm and deep. And so there are opportunities that come out of the woodwork along the way that leapfrog to the front of lists that we fold very set. And so it's part of our capital strategy to have dynamic flexibility and execution with speed. And it's also part of the pipeline and relationship management that we do work on across the infrastructure and utility partners that we have. And so that type of dynamic flexibility is why we've been successful acquiring power and land and developing it with the quality that we have. And so when we look at that multi-gigawatt pipeline, a lot of those regions where we see expansion opportunities are places where we have deep relationships, the Georgias, Tennessees, Wyomings and now Texas is of the world. But some of the utilities that serve those regions, I'll use TVA as an example because I'm a homer, Tennessee Valley Authority serves 7 different states. And so while it's the same utility partner, it bleeds outside the lines of the great State of Tennessee. And so those are the types of dynamics that we see replicated across those relationships and part of why we feel so good about the pipeline growth opportunities and why we capitalize the business to hunt power and land just like Gary said.

Understood. Well, I appreciate the update and have a good health.

Nick, happy Thanksgiving.

And with no further questions in queue, I'd like to turn the call back over to Harry for any closing remarks.

Everyone, thank you again for joining today's earnings call. We look forward to being 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, America's Bitcoin Miner.

This concludes today's conference call. You may now disconnect.