管理層發言
Good afternoon. My name is Jeannie, and I will be your conference operator today. I would like to welcome everyone to CleanSpark's Fiscal First Quarter 2026 Financial Results Call. Thank you. Harry, you may begin your conference.
Thanks, Jeannie, and thank you for joining us today to review the first quarter 2026 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-Q 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 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 turn it over to Matt.
Good afternoon, and thank you all for joining us. This quarter represents a meaningful step forward in CleanSpark's evolution into a digital infrastructure and data center development company. One that builds on the strengths of our mining operations while expanding the set of opportunities our assets can support. We continue to operate a large-scale fundamentally sound Bitcoin mining business that generates durable cash flows and balance sheet strength. What is different today is what those cash flows now enable. CleanSpark is no longer a single track business. We are building an infrastructure platform with multiple independently valuable earning streams, all anchored by scarce utility-grade power. Bitcoin mining funds the platform. AI monetizes it and digital asset management optimizes it across all cycles. To frame how we think about AI development, we see three phases. First, securing scarce power and land; second, tenant-driven technical and commercial alignment; and third, structured long-term monetization.
We are now firmly in the second phase across multiple assets. As a result, when we look forward, we increasingly see a company defined not just by hashrate, but by the quality, scale, and flexibility of its infrastructure and by its ability to allocate capital into the highest return opportunities available at any point in the cycle. As we evaluate the opportunities for expansion into AI, we are seeing improving economics per megawatt, driven by scale, power quality, and contracting structures even as capital intensity increases. Despite this evolution, Bitcoin mining remains foundational to our business. We are fully operational, passing every day and generating strong cash flows from a scaled mining footprint of more than 50 exahash per second. During the quarter, despite challenging Bitcoin price action and rising network difficulty, we generated more than $180 million in revenue at a gross margin exceeding 47%.
Those cash flows allow us to fund growth deliberately. They give us the flexibility to hold assets in a fully monetized state while we complete diligence and commercial alignment rather than being pressured into speculative development. We've built this strategy to perform across a range of market conditions, including lower Bitcoin prices, slower AI deployment, or tighter capital markets without forcing reactive decisions. In November 2025, we completed a $1.15 billion convertible offering as part of our strategic evolution. Part of the use of proceeds was used to repurchase $460 million worth of shares, bringing total share repurchases to over $600 million since December 2024. This resulted in approximately 20% of our shares outstanding being repurchased because we believe dilution is not a strategy; discipline is. Turning to our power and land strategy. Historically, we built CleanSpark by acquiring and optimizing a large number of sub-100 megawatt sites.
Those assets continue to perform well and have appreciated meaningfully as energized land has become increasingly scarce and valuable. As we evaluated the AI market, we recognized an opportunity to capitalize on the demand for larger sites. Until recently, Sandersville with approximately 250 megawatts of already live power was our only large-scale asset capable of supporting hyperscale workloads. That has changed. In October 2025, we acquired 271 acres in Austin County, Texas, along with 285 megawatts of contracted power fully approved by ERCOT with certainty on energization and the potential gas capacity for significant behind-the-meter optionality. In January, we followed with a second development initiative in Brazoria County, Texas, supported by a transmission facilities extension agreement enabling an initial 300-megawatt demand load expandable to 600 megawatts. Together, these assets establish a Houston area infrastructure hub with almost 900 megawatts of aggregate potential utility capacity, assembled intentionally to support multiphase AI campus deployments.
As we look ahead, we expect to move from portfolio formation into commercialization milestones. Those milestones will take different forms, site-specific announcements, development partnerships, and structured long-term offtake agreements, but they all reflect the same underlying reality. Our assets are being pulled into the AI market, not pushed. We believe that over time, as those options convert into contracted visible cash flows, the market will increasingly recognize the embedded option value in our power and land portfolio. At Sandersville, we further strengthened our position with the acquisition of a 122-acre parcel in direct proximity to our substation and power infrastructure. These additions were made in close consultation with a select group of potential counterparties. Importantly, these discussions are no longer theoretical. We are operating from tenant-driven specifications, not internal assumptions.
We are now past initial screening and into advanced diligence across multiple sites, including power studies, cooling validation, and commercial structuring. The decisions we are making today around substation design, cooling architecture, and campus layout are not reversible, and they reflect confidence in where demand is heading. What excites us about AI monetization is not just scale, but the duration, predictability, and capital alignment of those cash flows relative to traditional compute. Throughout this process, we are expanding responsibly. That means being infrastructure-first aligned with customer requirements and disciplined in capital deployment. In this market, moving too fast is often riskier than moving deliberately, and we are intentionally optimizing for durability rather than velocity. As we plan this evolution, we have established an optimized operating model that allows us to continue running our mining infrastructure right up until load transition.
When that transition occurs, we expect to redeploy miners elsewhere in our portfolio where they can continue to operate profitably. Earlier, I said that Bitcoin mining will always be core to our business. And that's because it continues to provide us with a strategic advantage and power acquisition. That advantage is now translating directly into differentiated positioning in AI infrastructure. We have seen this movie before. The discipline that allowed us to scale mining profitably across multiple cycles is the same discipline we are applying here, only now with larger contracts, stronger counterparties, and materially longer duration cash flows. Before turning to digital asset management, I want to briefly comment on the AI lease market. We believe there are meaningful second mover advantages in AI infrastructure, similar to what we experienced in Bitcoin mining. Lease economics have continued to improve across multiple dimensions.
Rates have risen, risk-sharing terms have become more balanced, and credit markets supporting these projects remain deep and constructive. When negotiating large-scale contracts, we are balancing lease rates, delay provisions, capital structures, and counterparty quality to optimize the holistic return profile. Our goal is not to win a single deal but to build durable scalable relationships that monetize our growing portfolio over time. I also want to briefly touch on digital asset management. DAM is not a trading function. It is a capital allocation and liquidity management capability with defined mandates and risk limits. During the quarter, DAM generated over $13 million in premiums and cash. That represents about 24% of normalized adjusted EBITDA and improving capital efficiency across our business. These results are process-driven and fully integrated into our broader financial framework.
As we look forward, we see multiple paths to value creation unfolding in parallel, continued strength in our operations, increasing visibility into AI monetization, and disciplined balance sheet management that preserves strategic flexibility. With that, I'll turn the call over to Gary.
Thank you, Matt. Let's dive right into the numbers for our fiscal first quarter 2026. For the quarter, our revenue grew year-over-year by approximately $19 million, an increase of almost 12%. Our Bitcoin production was relatively flat, where we saw revenues of almost $100,000 per Bitcoin in the quarter compared to $84,000 in the same quarter last year. Our gross margins declined slightly from approximately 57% a year ago to 47% this quarter. This decline was mainly driven by the year-over-year increase in network difficulty. Power prices also increased marginally to $0.056 per kilowatt hour, up from $0.049 a year ago. However, this reflects our decision to continue hashing during higher cost, higher revenue periods instead of curtailing based solely on an arbitrary power price threshold. This quarter, we recognized a net loss of approximately $379 million compared to net income of approximately $247 million a year ago.
This change was driven primarily by mark-to-market adjustments to Bitcoin's fair value at the end of each respective period. Our adjusted EBITDA was negative $295 million compared to positive $322 million a year ago, also driven primarily by mark-to-market adjustments. Turning our attention to the performance of the first quarter versus the immediately preceding fourth quarter, revenues declined approximately $43 million or 19% to $181 million. This drop was primarily due to a combination of two external headwinds: rising network difficulty and softer Bitcoin prices. Because of these pressures, we experienced some of the lowest cash prices in history during the quarter, underscoring the importance of having a fleet with high uptime and efficiency. Quarter-over-quarter, our cost per kilowatt hour decreased marginally from $0.059 in Q4 to $0.056 in Q1, partially offsetting our 19% revenue decline.
As a result, our gross margins remained healthy at 47%. With respect to our overhead expenses, it is important to note that the prior quarter includes approximately $25 million of expense related to separation from our prior CEO. As mentioned on last quarter's call, we do expect that our professional fees, payroll, and G&A line items will increase as we execute on our AI strategy. Additionally, I want to underscore 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 adjusted EBITDA was negative $295 million for this quarter compared to positive $182 million for the fourth quarter. It is important again to note that the difference relates to noncash mark-to-market adjustments, for which the current quarter includes approximately $350 million of these charges. On a normalized basis, taking these mark-to-market adjustments into account, our normalized EBITDA would be $55 million or approximately 30% normalized margin for this quarter.
This represents cash generated from our operations. Bitcoin value as of our September 30 balance sheet date was approximately $1.5 billion, and as of December 31, it was $1.15 billion, which the difference is attributed to the noncash mark-to-market adjustment of $350 million I mentioned earlier. Turning our attention to the balance sheet, you'll see our cash balance increased over $400 million compared to Q4. This is due to the $1.15 billion 0% convertible transaction we closed in November. As you know, we used a portion of the proceeds to pay off the outstanding balances on our Bitcoin-backed lines of credit and also repurchased $463 million of stock. This left approximately $420 million of net cash proceeds, the majority of which we will still have on our balance sheet. In addition to our cash balance, we had approximately $1.15 billion of Bitcoin value as of the end of Q1. Our total debt is approximately $1.8 billion, which on a net debt basis is approximately a 1.1 debt to liquidity ratio.
Most importantly, the converts do not come due until 2030 and 2032, and numerous options remain available to us for capital. It is also important to note that our outstanding share count has decreased almost 20% in the last 15 months, as we have not issued a single share of equity on the ATM or other offerings—dilution is not a strategy; discipline is. Turning our attention to our balance of over 13,000 Bitcoin. I want to point out that we are one of the first, if not the only company, which has scaled operations that is also using Bitcoin as a productive capital asset. On the last call, we discussed in detail our DAM strategy in its first full quarter. You may have also heard us previously talk about our 'crawl-walk-run' approach, which I'm happy to say we're now fully in the 'walk' phase. We're at full utilization of the portion of our Bitcoin balance we expect to use for yield generation, which is 40% or approximately 5,200 Bitcoin.
Our DAM strategy generated $13 million in cash returns on the Bitcoin model during the quarter when Bitcoin price was down mark-to-market. I want to highlight several key members who speak to our core DAM strategies. We overlay a covered call derivative program on our monthly production and sales of Bitcoin, which resulted in an uptick of $7,700 or 8% per Bitcoin over the average sales price of approximately $97,200. Overall, the $13 million in total premiums also represents an annualized return of 4.2% on our average total balance, which surpasses our target of 4%. We accomplished this all within six months of our first trade. Importantly, this is all achieved by monetizing elevated volatility, especially in October, while keeping the average delta below 20. I also want to point out that we have added an additional tool to our treasury management to belt. The Basis Trade is a market-neutral strategy that captures the difference between the forward price of Bitcoin and the spot price.
Importantly, this strategy takes no price risk and generates returns from the same types of market structure dynamics that we noted in our thesis in the first place. This basis trade allowed us to put our cash balances to work and exceed the risk-free rate by almost 200 basis points, as we saw an annualized yield of over 5.5% on the cash allocated to the basis trade. While these opportunities are cyclical, we will continue to be opportunistic based on market dynamics, filling out the flywheel we initially envisioned when we launched our DAM team. On a final note, I'd like to take some time discussing our capital strategy going forward, especially in light of our expansion into AI data centers. From a capital perspective, I'm confident the capacity and appetite for financing an AI data center with a grade A tenant is strong. We saw a high-yield deal from our friends at Cipher, which priced at an attractive 6% and 8%, indicative of the quality of recent leases being signed and the capital available in this market.
The recent $2 billion bond had approximately $13 billion in demand, an oversubscription of 6x. While we have not committed to one specific means of financing our AI data center builds, we are focused on building a capital stack that minimizes dilution. This continues with the sale of monthly Bitcoin production to cover our operating expenses. Between our current cash balance and capacity on the Bitcoin-backed lines of credit, we have over $800 million of liquidity available without selling any of our Bitcoin hurdle. This liquidity provides us optionality, and we will continue to use the lines of credit opportunistically in the marketplace for accretive purposes. Matt spoke about our current efforts and where we are going, and we are excited to share on future calls the relationships and ecosystem we are building—one that is a more fulsome approach than exists in the market. While we are early in the innings of our AI data center journey, the market is moving quickly, and CleanSpark is responding decisively. Our conversations with grade A credit quality tenants are ongoing, and it is not a matter of if, but when. With that, I will hand it back to Harry to lead us into Q&A.
Thanks, Gary. We will now open the floor to questions from the analyst community. Operator, please provide instructions and manage the queue for the Q&A session. Your first question comes from Mike Grondahl with Northland Securities.
分析師問答
I was wondering if you could talk a little bit about the demand environment you're seeing for HPC? And maybe how that's changed in the last 90 or 100 days? And kind of what attributes are you looking for most in a lease partner?
Mike, thanks for the question, and thank you for the recent initiation. We're glad to see Northland covering us. I can tell you that six months ago, when I reassumed the role of CEO, we entered a market where there was a lot of enthusiasm around signing a deal. And what we're now seeing is some of the punitive components of the early leases, such as losing a significant amount of revenue for a day late on an RFS date, and differing terms that are backstopped only at the site level rather than at the top-level. This has given us an opportunity to really sit back and evaluate what's out there. And I can tell you that Gary, Harry, and I, along with some of our team attended the Pacific Telecom Conference in Hawaii. The feedback we received by presenting an end-to-end solution was overwhelmingly positive. We've been very pragmatic about the assets that we've accumulated—the location, the distance away from fiber networks, and the access to behind-the-meter generation.
As a result, we've now been entertaining multiple trillion balance sheet companies that are interested in long-term leases on some of these assets. So we're seeing the demand continuing to escalate. And I might add, we saw Amazon earlier today talk about their commitment to invest $200 billion in AI infrastructure in 2026, exceeding the $140 billion estimated by the Street. So looking at the demand behind that, we feel very solid about it. And if the inbound inquiries and conversations we're having with hyperscalers are any indication, the fear of a bubble is highly overstated.
Got it. And then maybe just as a follow-up, your three sites—Sealy, Sandersville, and Brazil—would you say it's equal demand for all three? Or is there one that sticks out amongst those? How would you handicap that?
I think probably the highest demand right now is Sandersville. Quite frankly, because it's 250 megawatts, we already built a substation; it's already energized. The Sealy site energization is Q1 '27 for the first 207 megawatts, so we're seeing strong demand there. And obviously, the next site has also been very appealing. But I would say that the data center environment in Georgia and the energized site are very compelling to the offtake clients.
Your next question comes from the line of Brian Dobson with Clear Street.
So just as a quick follow-up. You mentioned there have been some really positive CapEx comments from companies like Amazon. To me, that signals rising demand for AI data centers. Would you say that that's indicative of demand, call it, across the sector from various hyperscalers that you're speaking with? Or are people getting more cautious at all?
Yes. I would say it's an emphatic yes. Just as a quick aside, Jeff Thomas, who leads our AI venture, has been in the office with us this entire week, and more often than not, he's excusing himself to go into his office and close the door to field an inbound inquiry. So I would say demand is escalating rapidly.
That's certainly good news. And I know you guys mentioned that you're looking for a mix of quality and scalability among clients, given construction commitments that you've already made, how confident are you that you'll be able to, call it, sign a contract in the relatively near future?
We're very confident, Brian. I'll be honest with you. The delay in—I wouldn't even call it a delay; I mean when we did this six months ago and then we had our earnings call seven or eight weeks ago, we said that we would expect to sign a quality lease in less than a year. And I would say that, that's highly accelerated. But the discipline that we're taking, you look at some of the leases that other Bitcoin miners have put up and they're very highly redacted in the public filings. And that's a result of the punitive nature of some of the delay provisions. So as we contemplate this, we're actually working on a basis of design with the offtake customer. We're designing it in advance and then assuring that we can meet the delivery timelines to remove that potential overhang of failure to deliver risk. So being disciplined about this—and building specific to the basis of design for the offtaker, including the implementation of the approved reference architecture from the chip manufacturers—will allow us to have that certainty to secure the supply chain before we enter into these commitments to ensure that we don't have that fail to deliver.
Excellent. Excellent. And then just one final one on Bitcoin mining, if I may. Given your efficiency, you're better positioned than those heading into the next halving, I guess, has your thought process changed at all as far as operating Bitcoin lines in tandem with your expansion into HPC?
That's a great question, Brian. And what we found is that as new energy sources are energized, some of these communities, especially the smaller communities, are incentivized to monetize those metal lots very rapidly. The challenge is to build a data center for a hyperscaler with the approved basis of design and incorporating that reference architecture is a 12-month best case, 18- to 24-month kind of average case delivery timeline. We can use the infrastructure that we have for Bitcoin mining, like we did in Cheyenne, Wyoming, where we secured a 100-megawatt lease over a hyperscaler. We did that simply because we committed to start paying power bills inside of six months, not inside of one and a half years, and that makes a difference to these communities. So we'll continue to use Bitcoin mining as that tool. You heard us talk about on the call, something that we haven't published yet because it wasn't material; and that is we have a 122-acre parcel adjacent to Sandersville.
What does that mean? That means I can operate 11 exahash to a profitable Bitcoin mining up until the day we cut the power over to support the data center for our end-use clients. We also have on the map of our projects, something that we haven't talked about is a 15-megawatt site in South Dakota. The utility there had introduced a blockchain-specific tariff that, with an interruptible load, gives us the lowest cost per kilowatt hour of almost any site in our portfolio. So that flexibility allows us to migrate that mining to a profitable location once we've spun up a data center behind us. So we see it kind of as a loss leader, but it makes money.
Your next question comes from the line of Mike Colonnese with H.C. Wainwright.
Matt, first one maybe for you. I appreciate your comments on the HPC business with start being advanced discussions or diligence stages rather potential tenants here. And that you're currently looking on a basis of design. Curious about what milestones should we be on the lookout for next and some of the expected timelines you see as we come across the next couple of quarters here?
So I think the process when you're dealing with a hyperscaler is we could rush in and sign a lease to get a headline. And then we're facing potential losses for a failure to deliver. So as I mentioned in my prior comments, we're working towards that basis of design. And one of the things I think that is a key differentiator that's maybe gone a little bit under the radar. And that is we put out a press release announcing an MOU with Subaru. Mike, you've been around our company long enough that we don't ever make a material disclosure unless we've got a firm contract. And we felt that that was important as we head into some of these discussions because Subaru has been very successful in building a modular MEP. So mechanical, electrical and plumbing, all the fiber runs—all according to the reference architecture required by the chipset manufacturers. So our solution will be to build the gray space, to build a tilt-up shell, and then slot in the reference architecture.
That also gives us flexibility. So if you have a hyperscaler that wants to modify from one particular type of chip to another, we have that modular approach. It also shortens the timeline because we've all heard the horror stories about some of our peers that have a couple thousand tradesmen all working at the same sites in West Texas, and they're struggling to provide housing and food and bathroom facilities. We look at this differently. We build instead of a one-off data center that's stick-built. We build the shell according to the specifics required by the end customer and then we build the MEP portion in a factory. So it's consistent and duplicatable and scalable, which is differentiated from anything else in the space. So it's important to us to establish all of those build parameters ahead of time so when we put pen to paper, there's absolute certainty that we can deliver the product as expected on time.
Helpful color. Matt, I appreciate that. And Gary, maybe one for you. Does this recent downturn in Bitcoin prices change your huddle approach at all? I know from covering the name for a while here, you guys have historically had a very dynamic huddle approach, one that tended to adjust based on prevailing market conditions. So curious how you guys are thinking about the huddles back here?
Thanks, Mike. Since you've been around a while, you know we've built this business on optionality. So that option is still on the table if we wanted to dip into the huddle and part with some of those Bitcoin, I'll tell you that's not something we're planning on doing even at these levels. We think the strategy is still intact, and part of the hedge for us is really selling nearly 100% of our monthly operating production. So as of right now, there's really been no change in that strategy. I'd also conversely say that we're not expecting to hold 100% of the operating Bitcoin production either because that would mean that we'd run through our cash a whole lot quicker. As we had mentioned when we were raising the convert funds, we expect to use the majority of those funds to expand in AI data centers because we think that's the future of the company.
Mike, maybe that's just a little—another layer to the Bitcoin mining side. At our last disclosure, we were at 16.07 joules per terahash, and Taylor and his team are actively deploying the 13.5 joules per terahash machines in immersion-cooled containers in five different locations. So we expect our fleet efficiency to continue to improve. In the last cycles, I mean we've been through this a few times, we see the kind of washout of the sorting process, and the less efficient fleets tend to unplug. So we also believe that there is a very strong opportunity for us to organically grow a share of the network hashrate just by default as other less efficient miners unplug.
Your next question comes from the line of Greg Lewis with BTIG.
Just thinking about the move forward in the HPC, I know Jeff joined the team a few months ago now. Gary, you alluded to potentially higher SG&A over time as we kind of build out the team and get ready to pivot into this new business. Like how should we think about costs and processes? And where are we in terms of—we've seen other companies go out and build teams. What we're realizing is we have a lot of capable, talented people already inside the company. How should we think about growth at the employee level here?
Greg, thanks for the question. It's a great question. We get it quite often from investors. I'll tell you, it's hard to give guidance on that because while we have a plan to bring on a certain number of FTEs, the timing of when those hit is really what's going to drive what the numbers are going to be for the fiscal year. Additionally, we have optionality to where we can rent services. So if we need services from someone, we could bring in outside consultants or contractors to help fill that void while we're waiting to bring on full-time talent. That could be different from what it would be to bring on a burden employee. So we're not prepared to give out numbers right now. I don't think it's anything that's material that's putting us at risk or anything. I think we've been pretty measured about bringing on people right around the time we will need them. So I think you'll see that slowly uptick throughout the remainder of the year.
Okay. Great. And then I think it's been understood that we were going to acquire more land at Sandersville for at least a few months. How does now owning that additional land at Sandersville, does that go at all in changing the kind of conversation that it seems like we're focusing—part of this call is you focusing on potential terms of some of these HPC contracts? Does not owning some of the land potentially leasing it kind of a nonstarter?
Greg, it's Harry. I think you're exactly right. So we view the closing of the land expansion at Sandersville as a very orderly process in progressing the AI data center project there. It allows us to move into a very specific basis of design alignment exercise, which is underway. And it also brings a level of specificity to the compute and power ramp for the data center deployment as well because there's complexity to these projects that extend beyond standing up the data center for our state. There's a lengthy commissioning process that the tenants typically take on in the context of the overall project life cycle. Being able to map out those timelines and those work streams in detail is critical as we move through the full commercial scope of the discussions; they are, in many cases, governed by some of those technical pieces in the ramp process.
Your next question comes from the line of Stephen Glagola with KBW.
Thanks for the question. Can you maybe provide some insight into how ERCOT's proposed large load bag study process may impact the energization timeline for the Sealy side as well as the approval and development schedule associated with your Brazoria County, Texas project?
Stephen, yes, Harry, again. Happy to do that. So I think the first piece of it is that the study process that ERCOT is proposing to roll out has not gone final yet. They're still in a comment period where they're taking member requests for how they want to influence that process and how it's going to be brought to market. So we're waiting to see kind of the final form of that. But given the early news there, we've had a lot of detailed discussions with a number of counterparties that we're working with there. That includes the substation developer, the utility, some of the political folks, and obviously some of the teams that are caught as well. I think the assets that we have in the state are in a very favorable position relative to this new piece of the process for a handful of reasons. The first is that the large load studies that have been done at Sealy are complete, and at the second location it’s in a deeply progressed state.
We've received the notice to proceed language at both of them. So that's kind of point one and two. The next is that the interconnect and the FDA pieces are executed. The third at Brazoria is that the CAIC has been funded, and at the Sealy location, the substation is already under construction. These are significantly progressed projects. And what we've seen is that the view of the batching and the study rollout is largely being informed by project maturity as well as location. And what we've gotten feedback from the utility at both of these locations is that the points we've selected are locations in the overall ERCOT transmission system that are going to be the least impacted by this type of reevaluation process. So we feel very, very positively about where these two assets sit within the system and how they're going to be treated. But until ERCOT comes out with final language, we can't have 100% visibility into that yet.
Your next question comes from the line of John Todaro with Needham & Company.
As you think about longer-term pipeline, adding more power, are there other power markets that are now starting to look more attractive relative to Texas, and where could we see that? And then I have a follow-up on the HPC tenant side.
Thanks, John. I think that we have always had a strong heritage of diverse portfolio construction. We see it in the way that we enter into power agreements today. We've got a significant footprint deployed and operating in Georgia. We've got significant presence in Tennessee. Wyoming and Mississippi as well are the smallest two, but they're by no means small. So I think that what we're going to be able to accomplish is a continued expansion in those markets because of the relationships and community quality that we've engaged in to date. Additionally, I think that the other side of the question you’re asking is do large-scale data centers skew toward in front of the meter power or behind the meter power. And we're asking these questions internally along both vectors. So we think that there's a huge amount of opportunity inside Texas and outside on the in front of the meter profile. We have a team that has become exceedingly expert in sourcing, negotiating, and closing on that power.
We are also strongly evaluating the capacity for behind-the-meter power as well in places where we're either able to get a smaller in front of the meter load or there's a particularly rich commodity environment by which we could power behind-the-meter generation and then the associated data center. So we're people-first business fundamentally, and our power and land teams are prepared to expand the portfolio very broadly in a diversified way, but also add that potential for behind-the-meter to the repertoire as well.
Great. And then just on the HPC tenant discussion. I guess just trying to gauge how far advanced we are in the positioning. Are we down to one potential tenant that seems much further along? Or is there kind of three that are in final competition stages? Just a little bit more color there?
Ask me a question that I can answer. What I would say, John, to be honest with you, is there are multiple potential offtake tenants for Sandersville to begin with. I would say there is a specific front runner by an order of magnitude to the extent that our team is collaborating with their team on the site placement. You may have seen a slide in our deck that had a mockup of the layout of the data center. So we've advanced it significantly with a particular offtake, but by no means is it committed elsewhere. I mean there, the competition for megawatts and land right now is stronger than it was when we announced this strategy to expand into AI. So we're not closing any doors, but I would say there’s a clear frontrunner there.
Your next question comes from the line of Brent Noble with Cantor Fitzgerald.
This is Gareth on for Brett. I was just hoping you could go into detail on the two new sites in Texas. When are you guys expecting to have power available on those sites? And what do you think the timelines are kind of going forward there?
Yes, absolutely. So let's tackle the Sealy project first. The land secured is 271 acres. The gross power is 285 megawatts. The first 207 to 209 megawatts is coming first half of ‘27. Then it’s about 40 in ‘28 and 40 in ‘29. And that's driven by the transmission agreement by which we secure the power. The second project is in Brazoria County. Larger footprint with up to 477 acres at that location. The way that it's structured is that we've signed for that agreement, but we're not closed yet, and there are some closing conditions associated with it, that we expect to wrap up here. The timeline for the energization is a function of some of those closing conditions. We don't have the type of line in the sand clarity that we have at Sealy. But I think that Q4 ‘27, Q1 ‘28 is a reasonable range for that, and everybody internally is working to bring that energization date as close to the inside as possible.
Your next question comes from the line of Paul Golding with Macquarie.
Congrats on all the progress. Gary, you referenced a peer's recent capital raise. I just wanted to ask, as we think about the liquidity you have, but also you have substantial capacity going forward that you'll hopefully be growing into with leases. Should we take that to be—that comment to be indicative at all of how you hope to face a counterparty in terms of counterparty type face-to-face with hyperscaler able to do high-yield raises where the counterparty credit quality could yield that? I guess, how should we think about how you're selecting your counterparties given the context given around capital raising and cost of capital?
Thanks, Paul. I'll tell you that it's very important to us to have grade A credit quality tenants because we think that's the most financeable and the best cost of capital. So that's what we're focused on. In terms of the vehicle, I quote the recent Cipher deal because the high yield seems to be a playbook that a number of our peers have started to go down that path. We're open to that. We're happy to see that the terms are getting better, both with that and the contracts and leases backed by that bond. There are some other options as well, obviously, a little higher cost of capital. But at the end of the day, what's great about this is, and you've heard this word from us for quarters or years now, is optionality, right? We have a lot of options on the table. But I think it's safe to say we're going to follow a playbook right now that's probably proven in the capital markets, and it all starts with a grade A tenant.
And then maybe a follow-up. I believe, Matt, you mentioned when speaking about the Sandersville work and the 122 acres with tenant-driven specs in mind. How should we think about what that means for terms? You also noted that terms in discussion were seemingly more positive across the conversations you're having. Is there any kind of prepayment or deposit discussion involved in the conversations you're having, given that you are proactively using tenant-driven specs to set up the sites for HPC?
It's—thanks for the question, Paul. It's a bit—it's a little early in the discussions to comment on that directly. But I can tell you that we're—obviously, there are a number of different leases that have been put up. You've seen modified gross; you've seen triple net; you've seen posted agreements. You've seen miners that have committed to buy the chips themselves. So I can tell you that our focus initially—and I don't mean to downplay the quality of any other transaction—but as we look at the financing options, I think it's important to us to have a significantly better deal than I think the market would have otherwise expected. I think we're interested in putting a deal together directly with the hyperscaler, not necessarily with the Neo cloud backed by a hyperscaler. And we believe that we'll set the standard for the quality of the agreement, and the, I guess, win-win is a term we use in the company a lot. There will certainly be expectations our feet will be held to the fire, so to speak, to deliver, but it's not at the risk of an existential threat on a fail to deliver. So we're negotiating all those terms. And I think what you'll see when we announce the first lease is a basis or a model for what you can expect going forward.
Your next question comes from the line of Jim Milre with Chardan.
At the current Bitcoin prices, let's call it $63,000 or so. How much of your minor fleet is economic to operate? Or another way to ask it is how much of the minor fleet meets the hurdle rate in order to operate?
Great question, Jim. Thank you for that. So our fleet efficiency improved and then it got a bit worse. And it got worse by design because as mining economics improved, we actually started to scale up some less efficient equipment in our fleet. What I can tell you is that Taylor and his team are constantly running real-time analysis based on utility prices, network difficulty, and the price of Bitcoin. And they brought me in to Gary and myself this morning as we were working on this presentation today. They brought us in a list. I would say less than 10% of our fleet at the current half price is not profitable. So the vast majority of it is, and the small portion that is at or below the breakeven threshold are machines that we brought on to take advantage of $125,000 Bitcoin 1.5 quarters ago. So it's not punitive to us to unplug those. Having said that, as we unplug those less efficient machines or scale them down or underclock them, it increases the overall efficiency of our fleet.
Got it. That's helpful. And can you discuss CapEx plans for this year and next, both in from a dollar basis as well as an allocation between Bitcoin and HPC?
Jim, it's Gary. I'll tell you that our focus is going to be on deploying capital towards AI. That's in the range of $9 million to $11 million a megawatt, which, as you probably know, has been reported by most of our peers in the space right now, and that's the range that we're seeing. So what is deployed is really going to depend on—that amount is going to depend on the design to build and the customers, and when we sign those respective leases. But I'll tell you that the overwhelming percentage—the majority percentage of that is related to HPC. With respect to Bitcoin mining, I think that the investment, particularly at these levels, doesn't make a whole lot of sense from the sticker prices that we're seeing from the major manufacturers. If you look at our balance sheet as of 12/31, we had about $130 million of prepaid deposits on Bitcoin mining equipment and miners. And about $112 million of that was through September.
So we're still deploying some infrastructure, mainly emerging cooling and miners that will help drive down that efficiency. But we don't plan on spending a significant portion of our cash on mining unless economics change. We need to keep in mind that we are about two years off to the next halving. And in any cycle, as you get closer to having that ROI window closes rapidly. And right now, it doesn't make sense. So we want to be redirecting every dollar possible towards CapEx.
Your next question comes from the line of Matthew Capitalo with Maxim Group.
I'm filling in for Mac right now. I was just wondering if you guys have any insight or I guess, predictions on how we should be thinking about network difficulty in response to the current Bitcoin prices.
Yes, absolutely. So I think that what we've seen over the last weeks and the difficulty adjustment that's coming on Saturday is important to know is that it's the largest difficulty adjustment to the downside since the China mining ban in 2020. And that's a combination of two factors. The first is that there have been significant weather events across the entire country during that period of time where you're seeing either the demand response programs get engaged or you're seeing power prices move past the breakeven point of economics. And so that's certainly a contributor to this difficulty adjustment. The second piece is clearly Bitcoin prices are off considerably. And so that next coming difficulty adjustment is going to be a significant one. And then I think given the price action that we've seen in the latter part of the difficulty adjustment period that we're in the middle of right now, we could see additional downward pressure on difficulty in addition to that.
So I think that ultimately, this is the self-healing nature of the proof-of-work and Bitcoin mining system. And as you see these types of market forces, the network adjusts to be able to create that security model that's supposed to keep producing blocks and processing transactions. I just want to add one thing because if you look at this historically, when Bitcoin runs and hash price gets better, the global hash rate laps right? You'll have a period of time, usually weeks, maybe a month or whatever for miners to find a way to get plugged in because miners just don't sit around waiting for hash price to sit typically; they're just not on racks just waiting for Bitcoin price at a certain price. It's the opposite on the way down because most miners, not all, but most miners know what their breakeven is because that's a real punitive cash penalty since they have to pay their power deals. And so as Bitcoin mining economics go down, what we've seen at least historically is that cash rate comes off pretty quickly, pretty close to that.
And that's because miners say, well, hey, why am I going to take a lot of X amount of dollars when I could just go buy Bitcoin and I have more Bitcoin than if I actually mined it? So I think that given that routed in the fact that we have a decreasing fleet with decreasing joules per terahash, meaning our efficiency is going up—and we're more efficient and we're producing more Bitcoin for every watt that we're putting through these machines—we will be one of the last ones theoretically to turn off, and we'll mine more Bitcoin in terms of quantity as that global hash decreases.
And I guess as a follow-up on the Texas opportunities, do you guys have a timeline on executing behind-the-meter opportunity into your portfolio?
Yes, I appreciate the question. I think it's too early to have a view on the exact timing for that type of opportunity. What I can say is that we're evaluating a number of different behind-the-meter deployment types. Some of those energization schedules are longer. Some of them are much faster to market. Ultimately, those types of decisions will be made in concert with the tenant community, and we're really looking to be able to meet their needs and satisfy impute demand, whether that's through one form or behind-the-meter generation or another.
There are no further questions currently. Harry, I turn the call back over to you.
Thank you, and thank you, everyone, 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 from us at CleanSpark.
This concludes today's conference. You may now disconnect.