Terawulf Inc

NASDAQ:WULF

Earnings Call Analysis

TeraWulf's 2024 Q2: Efficiency and Strategic Growth

TeraWulf reported a transformative Q2 2024 with notable achievements amid a challenging environment. They mined 699 Bitcoins, down 34% from Q1. Revenue fell 16% to $35.6 million, while gross profit dropped 23% to $21.7 million. Despite lower mining output, TeraWulf maintained industry-leading power costs. They advanced in AI and HPC, completing key projects like Building 4 at Lake Mariner, enhancing their operational efficiency. Strategic moves included amending Bitmain purchase agreements and eliminating debt. Looking ahead, their growth plans remain fully funded without needing additional equity financing.

Company Overview and Recent Transformations

TeraWulf, a leader in energy and digital infrastructure, is focusing on sustainable, low-cost, predominantly zero-carbon energy to power its operations. The company operates two main facilities: the Lake Mariner facility in upstate New York and the Nautilus Cryptomine powered by nuclear energy. In the second quarter of 2024, TeraWulf experienced significant operational changes and financial milestones despite challenges in the Bitcoin mining industry, particularly after the Bitcoin reward halving in April.

Quarterly Performance Metrics

In Q2 2024, TeraWulf mined a total of 699 Bitcoins, equivalent to about 7.7 per day, a decrease of 34% from the previous quarter's 1,051 Bitcoins. The company reported GAAP revenues of $35.6 million, a decline of 16% quarter-over-quarter. The average value per Bitcoin mined was $65,984, contributing $46.1 million to total revenues. The company's gross profit, exclusive of depreciation, totaled $21.7 million, down 23% from the previous quarter, reflecting a higher energy cost per Bitcoin mined of $22,954, compared to $15,501 in Q1.

Cost Management and Financial Guidance

TeraWulf has been successful in managing its power costs, achieving an average of $0.037 per kilowatt hour in Q2, down from $0.041 in Q1. For full-year 2024, the company expects average power costs of $0.035 per kilowatt hour. Total operating expenses are projected at $13.5 million. The company also aims for a marginal cost of production of about $40,000 per Bitcoin in the second half of 2024. Furthermore, TeraWulf has updated its SG&A expense guidance from $27.5 million to $30 million for the year due to hiring requirements for its HPC and AI initiatives.

Strategic Developments and Infrastructure Growth

The completion of Building 4 at Lake Mariner enhanced TeraWulf's mining capacity to over 10 exahash per second. The company is also advancing projects in high-performance computing (HPC) and artificial intelligence (AI), including the WULF Den project, which supports advanced GPU technology. An investment of approximately $8 million has been allocated for procuring NVIDIA GPUs, bolstered by financial arrangements with a major original equipment manufacturer (OEM) to minimize capital outlay.

Debt-Free Position and Future Opportunities

In a significant strategic move, TeraWulf has fully repaid its debt as of July 2024, enhancing flexibility for future growth initiatives without the burden of interest payments. The company's focus in the upcoming period will be on establishing high-quality, long-term customers within the HPC and AI sectors. This transition is viewed as a vital step to maximize the value of its assets and to ensure sustainable growth and profitability.

Focus on Customer Engagement and Market Strategy

TeraWulf is actively engaging potential customers for its HPC/AI capacity and exploring partnerships that align with its energy infrastructure capabilities. The company aims to leverage its substantial low-cost power access to meet the growing demand for data centers, particularly as industries increasingly rely on AI technologies.

Conclusion and Investor Outlook

TeraWulf is positioning itself at the forefront of the evolving digital infrastructure landscape, making strategic investments in sustainable energy and advanced computing technologies. As it continues to focus on operational efficiency and securing lucrative customer contracts, the company expresses a robust outlook for growth. The elimination of debt and a solid cash position enhance TeraWulf's competitive advantages, aiming for long-term shareholder value.

Earnings Call Transcript 2024-Q2

[Operator]

Greetings, and welcome to TeraWulf 2024 Second Quarter Earnings Call. [Operator Instructions] As a reminder, this conference is being recorded.

It is now my pleasure to introduce your host, Jason Assad, Director of Corporate Communications. Thank you. You may begin.

[Jason Assad, executive]

Thank you, operator. Good afternoon, and welcome to TeraWulf's second quarter earnings call. With me today are Chairman and Chief Executive Officer, Paul Prager; and our Chief Financial Officer, Patrick Fleury.

Before we get started, I'd like to remind everyone that our prepared remarks may contain forward-looking statements, which are subject to risks and uncertainties, and we may make additional forward-looking statements during the question-and-answer session. When used in this call, the words anticipate, could, enable, estimate, intend, expect, believe, potential, will, should, project, and similar expressions as they relate to TeraWulf are such forward-looking statements.

Investors are cautioned that results may differ materially from those anticipated by TeraWulf at this time. In addition, other risks are more fully described in our public filings with the U.S. Securities and Exchange Commission, which may be viewed at sec.gov and in the Investors section of our corporate website at terawulf.com.

Finally, please note that on today's call, we'll refer to certain non-GAAP financial measures. Please refer to our company's periodic reports on Form 10-K and 10-Q and on our website for full reconciliation of these non-GAAP performance measures to the most comparable GAAP financial measures. An updated version of our new investor deck may also be found at terawulf.com. We'll begin today's call with prepared remarks from Paul and Patrick, then we'll proceed to Q&A.

It's my pleasure to now turn the call over to TeraWulf's CEO, Paul Prager. Paul?

[Paul Prager, executive]

Thank you, Jason, and good afternoon, everyone. We appreciate your attendance today as we review our second quarter 2024 financial results. To start, let me provide an overview of who we are and highlight some key aspects of our business.

For those new to TeraWulf, we are an energy and digital infrastructure company dedicated to utilizing predominantly zero carbon energy to power our operations. Our 2 premier data centers are the Lake Mariner facility in upstate New York, which utilizes over 91% zero-carbon energy sourced from the grid, and the Nautilus Cryptomine in Pennsylvania, a joint venture with Talen that is directly powered by nuclear energy. We believe our strategic focus on scalable, zero-carbon energy infrastructure gives us a unique and competitive edge.

In our last earnings call, we identified a significant shift in the industry, emphasizing that low-cost clean energy, operational efficiency, and profitability are becoming more critical than simply scaling operations. This insight will frame our discussion today as we review our second quarter results and strategic positioning.

In the second quarter of 2024, we achieved several milestones that significantly strengthened our position as one of the most efficient public Bitcoin miners and digital infrastructure owners.

First, we successfully completed the construction of Building 4 at Lake Mariner, pushing our total mining capacity to over 10 exahash per second. With a fleet efficiency of 23.7 joules per terahash and industry-leading power costs projected, at $0.035 per kilowatt hour for 2024. We have solidified our position as one of the sector's most efficient public miners.

Second, we strategically amended our Bitmain purchase agreements to specify the delivery of S21 Pro miners. We opted to acquire approximately 5,000 of the 30,000 miners available under the option purchase agreement, monetizing our option at an attractive rate of $16 per terahash.

Third, we made significant progress in our AI and high-performance computing initiatives through the WULF Den project. We committed to purchasing a 128-GPU cluster from NVIDIA, financed by an industry-leading OEM. This arrangement minimizes the equity required from the company preserving capital for future strategic investments.

And finally, we achieved a pivotal milestone by streamlining our capital structure, eliminating debt, and converting approximately $29 million out of $41 million lender warrants into common shares. This deliberate restructuring strengthens our balance sheet and positions us for sustained future growth.

With that in mind, let me touch on our operational expansion. At Lake Mariner, we currently deploy approximately 200 megawatts of operational infrastructure for Bitcoin mining, achieving industry-leading unit economics, as disclosed in our updated investor presentation found on our website.

We also have an additional 300 megawatts of near-term expansion capacity to meet the growing demand from HPC and AI data centers. Lake Mariner's location offers the ideal trifecta for scaling operations, low-cost power, ample land, and abundant water for cooling.

At the Nautilus Cryptomine, we currently have 50 megawatts of operational capacity for Bitcoin mining. And earlier this year, we announced plans to expand to 100 megawatts in 2025. Our partnership with Talen, underscored by the recent sale of the Cumulus Data portion of the campus to Amazon Web Services, highlights the strategic value of our direct connection to a baseload nuclear power station.

Amazon's acquisition not only validates the importance of leveraging sustainable energy, but also signifies the inevitable convergence of sustainable energy with the growing demand for high-performance computing.

From day 1, our model has been to strategically locate our operations, where we have access to scalable, low-cost, predominantly zero-carbon power. Both of our sites embody this approach.

Recently, market participants and analysts have emphasized the importance of bringing data center capacity to locations that have access to power. This has been our strategy for many years, bringing our operations to locations with access to predominantly zero-carbon generation resources.

In addition to our current operations, we are exploring a pipeline of opportunities outside of the company, including certain properties owned by Beowulf Energy, a private company wholly-owned by me. While our primary focus today is creating value with the organic scalability within TeraWulf, we are continuously assessing the contribution of additional assets. Any such contributions will be made in a fair and reasonable manner, consistent with our Board and committee charters.

Moving on to our financial performance, TeraWulf continues to leverage our low-cost infrastructure to drive profitability and create value for shareholders. Our commitment to shareholders has always been to deliver on our promises, and we've been successful in doing so quarter after quarter.

In July, we fully repaid our debt, well ahead of maturity, and are now completely debt-free. This significant milestone not only strengthens our balance sheet, but optimizes our strategic flexibility. This is particularly advantageous as we aim to deploy our Lake Mariner facility to support the surging demand for data centers driven by the proliferation of data-intensive applications, such as artificial intelligence, machine learning, and big data analytics.

In the second quarter, we achieved a GAAP gross profit margin of 61% and non-GAAP-adjusted EBITDA of $19.5 million, translating to an EBITDA per exahash of approximately $2.4 million. Our SG&A expenses totaled $11.9 million, and only $7.1 million, excluding stock-based compensation expense, materially lower than our peers.

Looking ahead, as Patrick will detail shortly, our growth plans for the remainder of 2024 are fully funded. This underscores our commitment to capital efficiency and ensures the continuation of our strategic initiatives without the need for additional equity financing in the near-term.

As regards dilution in the second quarter, we took advantage of the favorable stock price and market liquidity to safeguard our future valuation creation path and to facilitate our entry and growth in the HPC/AI sector in the most capital-efficient manner. These data points highlight TeraWulf's competitive advantages.

Low-cost, zero-carbon power, access to some of the lowest-cost, predominantly zero-carbon energy in the industry, profitability, higher EBITDA per exahash than any other player in the space, requiring less capital for growth, efficiency, a lean team, and efficient management drive maximum profitability for shareholders, and scalable infrastructure for AI/HPC.

Wulf has access to over 300 megawatts of infrastructure to scale into high-value and high-performance compute data centers. The true value in TeraWulf lies not only in the Bitcoin we mine profitably, but in the quality of our asset base and our ability to generate industry-leading profits while expanding into HPC/AI.

Even in times of market volatility, as we all have experienced in the last week, our focus on maintaining robust infrastructure and profitable operations ensures sustainable growth and long-term returns.

With that segue, I'd like to now address our strategic advancements in AI and HPC. WULF Compute is in the final stages of our 2-megawatt proof-of-concept project, the WULF Den, which is on track for completion in early September. This state-of-the-art project is designed to support the latest generation GPUs, featuring advanced liquid cooling systems and high rack density.

In addition, we are advancing the development of CB-1, a co-location project set to deliver 20 megawatts gross and 60 megawatts of critical computing power. This building will be tailored to support the next-generation of GPUs and is projected to be operational by the end of the year. We are actively engaging with several co-location customers regarding the 300 megawatt expansion capacity at Lake Mariner.

Our team is thoroughly evaluating each opportunity to maximize the potential, of what we believe is one of the premier data center locations in the country. Our primary focus here is on securing the most suitable customers to achieve the highest possible valuation multiple for TeraWulf common shareholders, among whom management is one of the largest groups.

As we look ahead, TeraWulf is at a pivotal moment, as strategic positioning and substantial infrastructure scale offer significant opportunities for value creations. With our large-scale facilities, we are well equipped to meet the growing demands of the data center market, positioning us to capitalize on emerging opportunities and drive substantial growth. We are excited about leveraging our infrastructure and expertise to enhance value for our stakeholders in the coming quarters.

Now, I'll turn it over to Patrick to discuss our financial performance.

[Patrick Fleury, executive]

Thank you, Paul. As Paul stated, the second quarter and beginning of the third quarter was a transformative time for Wulf, with strong financial results even in a challenging fundamental business environment following the Bitcoin reward halving in April. In the second quarter of 2024, we self-mined 539 Bitcoin at Lake Mariner, and our net share of Bitcoin mined at Nautilus was 160, for a total of 699 Bitcoins or about 7.7 Bitcoin per day, a 34% decrease over the 1,051 Bitcoin mined in 1Q '24.

The hosting agreement at Lake Mariner terminated in February 2024, and as a result, we received zero and an additional 6 Bitcoin in 2Q '24 and 1Q '24 respectively from associated profit sharing.

Our GAAP revenues were down 16% quarter-over-quarter at $35.6 million in 2Q '24 from $42.4 million in 1Q '24. Our value per Bitcoin self-mined in this quarter, a non-GAAP metric that includes Bitcoin mined at Nautilus, averaged $65,984 per Bitcoin for a total of $46.1 million, as detailed and defined in our monthly operating reports, press releases, and MD&A section of our 10-Q.

As a reminder, there is a key difference between our GAAP financials and the monthly operating reports and 2024 guidance. Due to our 25% ownership in the Nautilus JV, the revenue, cost of revenue, operating expenses, depreciation, and amortization at Nautilus are not consolidated into our GAAP financial statements.

Instead, the financial impact of the Nautilus JV is reflected in the equity in net income or loss of investee net of tax line item on the GAAP income statement. Our GAAP cost of revenue, exclusive of depreciation, for 2Q '24 was $13.9 million, a 3% decrease over $14.4 million in 1Q '24.

The quarter-over-quarter decrease was due to a 14% decrease in the average cost of power at Lake Mariner from $0.049 per kilowatt in 1Q '24 to $0.042 per kilowatt in 2Q24, an expected demand response proceeds of $1.9 million in 2Q24 versus $1.3 million in 1Q '24.

Our gross profit, exclusive of depreciation, decreased by 23% quarter-over-quarter from $28 million in 1Q '24 to $21.7 million in 2Q '24. Our power cost, or cost of energy per Bitcoin mined, a non-GAAP metric that includes Bitcoin mined at Nautilus, was $22,954 in 2Q '24 compared to $15,501 in 1Q '24.

As a reminder, in our GAAP financials, unlike our monthly operating reports, the company records proceeds received and to be received for demand response programs as a reduction in cost of revenue. These expected proceeds total $1.9 million in 2Q '24 and $1.3 million in 1Q '24.

As disclosed in our 2024 guidance, we expect to achieve an average power cost, including demand response revenues and the impact of Nautilus' $0.02 power of $0.035 per kilowatt hour in 2024.

For 2Q24, we achieved an average power cost of $0.037 per kilowatt hour compared to $0.041 in 1Q '24. This is consistent with historical power price variability in upstate New York where the Lake Mariner facility is located.

Operating expenses were stable quarter-over-quarter at $1.7 million in 2Q '24 and 1Q '24. As disclosed in our 2024 guidance, we expect $13.5 million of operating expenses in 2024, which includes operating expenses at Nautilus. Of the $13.5 million total anticipated for 2024, approximately 50% is expected to be incurred at Lake Mariner and 50% at Nautilus.

SG&A expenses decreased quarter-over-quarter from $14.9 million in 1Q '24 to $11.9 million in 2Q '24. The decrease is almost entirely due to $6.9 million of stock-based compensation expense incurred in 1Q '24 versus $4.8 million in 2Q '24.

Adjusting for stock-based comp, SG&A decreased 11% quarter-over-quarter from $8 million in 1Q '24 to $7.1 million in 2Q '24. As I've indicated previously, this decline is expected as SG&A spend is more heavily weighted to the first quarter of the year versus following quarters.

With our entry into HPC and AI and need for more staff, we are updating our 2024 SG&A guidance from $27.5 million to $30 million of SG&A in 2024, as indicated on Page 14 of our August investor presentation available on our website.

Depreciation decreased quarter-over-quarter from $15.1 million in 1Q '24 to $14.1 million in 2Q '24, which was the result of a quarter-over-quarter decrease of $2.5 million in accelerated depreciation related to certain minors, of which the company shortened their estimated useful lives based on expected replacement by April 30, 2024, partially offset by increased assets placed into service and depreciated as a result of our continued infrastructure buildup.

Gain or loss on fair value of digital currency is a new income statement line item for us in 2024, given our early adoption of the new FASB accounting rule, which marks the company's Bitcoin holding to the fair market value as of the filing date with changes in fair value recorded in net income.

In 2Q '24, we incurred a loss of $0.7 million compared with a gain of $1.3 million in 1Q '24. It's critically important to note the 6-month 2024 net gain of $0.6 million is substantially all realized, meaning it's realized cash in our bank account, not theoretical mark-to-market gains on paper, which many of our peers at HODL have reported.

GAAP interest expense in 2Q '24 and 1Q '24 was $5.3 million and $11 million respectively, which includes cash interest expense and amortization of debt issuance costs and debt discount related to the term loans advancement. However, cash interest paid during 2Q '24 was $2.5 million, down over 33% from $3.7 million in 1Q '24. This decrease is the result of repayment of $30.2 million of debt in 2Q '24 and $33.4 million in 1Q '24.

As detailed in our press release on July 9, and debt footnotes of our 10-Q, we repaid the remaining $75.8 million of debt with approximately $19.8 million of excess cash flow sweeps as defined in our credit agreement and $56 million of equity proceeds from our ATM facility.

In connection with the voluntary prepayment of $56 million of debt in July, a third quarter 2024 event, the company incurred prepayment fees of $0.9 million, rolled off unamortized discount of $3.3 million associated with the principal repaid, and recorded a loss on extinguishment of debt of $4.2 million.

In 2Q '24, we reported $0.8 million in equity and net income of investees, net of tax, as compared to $5.3 million in 1Q '24. These amounts represent TeraWulf's proportional share of net income of the Nautilus joint venture.

Let's take a moment to review in detail how Nautilus impacts our GAAP financial statements and non-GAAP adjusted EBITDA. Please refer to the hypothetical quarter on the Nautilus illustrative P&L and financial statement impact Page 10 of our August investor deck to follow along.

We run approximately 1.85 exahash of total capacity at Nautilus, and pay $0.02 for 50 megawatts of power, assuming 98% uptime, a 610 network hash rate, and $60,000 Bitcoin price. This results in quarterly revenue of $7.4 million, cost of revenue, excluding depreciation, of $2.2 million, and therefore gross profit of $5.2 million.

Subtracting quarterly and other operating expenses of $1.7 million, which as discussed previously is approximately $6.75 million annually, results in operating profit of $3.5 million. This would effectively be the net distribution of Bitcoin to Wulf from Nautilus and the figure we add back in the non-GAAP adjusted EBITDA calculation.

Including quarterly depreciation of $5.6 million results in equity and net loss of investee, net of tax of negative $2.1 million, which is what would appear on our GAAP income statement and be stripped out of our non-GAAP adjusted EBITDA calculation.

Our GAAP net loss for the second quarter was $11.2 million, compared to a net loss of $9.9 million in 1Q '24. Our non-GAAP adjusted EBITDA for 2Q '24 was $19.5 million compared to $32 million in 1Q '24.

Now, turning our attention to the balance sheet, as of June 30, we held $104 million in cash with total assets amounting to $480 million and total liabilities of $93 million. Pro forma for our debt repayment on July 9, our adjusted cash and Bitcoin balance was $28.4 million.

Our net working capital as of June 30 was a positive $18.5 million. As disclosed on Page 14 of our August investor deck, we achieved a marginal cost of production, including every single cost in the company of $41,587 per Bitcoin in 2Q24, and expect to achieve approximately $40,000 per Bitcoin in second half '24.

Regarding our capital position and growth plans for the remainder of 2024, we are fully funded. On Page 12 of the August investor deck, you'll find a capital sources and uses bridge for 2Q '24, as well as our expectations for second half '24 on Page 13. We are now debt free and in pole position to maximize the value of our assets as we diversify into HPC and AI.

Let me take a moment to point out a few key items in the second half '24 capital budget on Page 13. We expect the following approximate capital expenditures in second half '24. Number one, $8 million on Wulf Compute's 2-megawatt HPC/AI Wulf debt, which includes the purchase of 128 NVIDIA H100 GPUs, consisting of direct liquid-to-chip cooling with backbone of a full cluster of which approximately 50% of the purchase price is being financed by a large OEM partner.

Two, $14 million on LMD Site Electrical to allow expansion to 500 megawatts. Three, $23 million on construction of Building 5, a 54-megawatt Bitcoin mining building expected to be substantially complete by year-end 2024.

And four, $30 million on construction of Wulf Compute's CB-1, a liquid-cooled, redundant, and high-power density 20-megawatt HPC/AI infrastructure expected to be substantially complete by year-end. The $30 million spend constitutes Wulf's equity contribution for the 20-megawatt building, and we expect to finance the remaining 70% in the project finance market.

As discussed on our 1Q '24 earnings call, updated on Page 15 of our August investor deck, and as a peer public co-location company has announced, we are targeting a customer contract with a 1-year revenue prepay, which would return Wulf's $30 million equity contribution.

Given the quality and quantity of customer interest in our assets, we are comfortable paying this forward, so we can deliver sizable HPC/AI capacity by year-end 2024 with additional rapid growth in 2025.

At Wulf, our financial objectives remain clear and simple, maximize profits, secure long-term, high-quality customers in HPC/AI to minimize Wulf's future equity needs, and return value to shareholders while providing investors access through transparency and accountability.

With that, I'll hand it over to the operator and look forward to answering your questions.

[Operator]

[Operator Instructions] The first question is from Mike Grondahl from Northland Securities.

[Mike Grondahl, analyst]

How would you characterize the demand environment for the 2 megawatts, the 20 megawatts, and then 50-plus going forward? Could you just give us a little bit of flavor of who you're talking to and what you're hearing?

[Patrick Fleury, executive]

Yes, Paul, I think you're lines muted. Do you want to take that or do you want me to?

[Paul Prager, executive]

Sure. Mike, we're very active. Listen, I can tell you that we're talking to a range of customers and a variety of customer class, if you will. We're talking to a couple of the hyperscalers. We're talking to end users. We're talking to enterprise customers of our OEM.

And I think that what's exciting to people is the immediacy that we're able to meet their goals. As you know, we've advertised that we'll have megawatts available by year-end. And we've given a very programmatic schedule for folks in '25.

So I think interest is at the beginning, at the tip of the iceberg, not even close to sort of the end of it. I think that the market's continuing to evolve. And a lot of folks are talking to the hyperscalers about what their needs are. And a lot of people don't want to be working with hyperscalers. So they want direct access. But we are engaged daily on this. And for us, I think it's very, very important, Mike, that we make the right decision and have a customer that's the right credit quality. We don't want to go out there with, if you will, somebody that isn't somebody that we could scale with over time or that we think won't be around in a year or 2.

So I can only tell you that, we're very excited about the prospects. We're very excited about the potential customers we're speaking to. But they're coming in all shapes and sizes. And again, our primary concern is credit quality and value ultimately to our shareholders in the deal that we strike. I think most of all, that should be indicated by the fact that we're funding this $30 million here. I think that should indicate to you real confidence on our part.

[Mike Grondahl, analyst]

No, that's great. And then just maybe as a follow-up to that, in the last 90 days, as you've taken part in a lot of these discussions, what have you maybe learned the most or what's been surprising or interesting to you, Paul?

[Paul Prager, executive]

Patrick, I'm going to let you start and I'll give my sort of closing remark on that.

[Patrick Fleury, executive]

Yes, look, Mike, as you know, I always tend to answer things too honestly. So I would just say, look, I'm the CFO, right? My job is to get the numbers right and get things financed and financially engineered. I think for me, we have a -- I won't use the word, but we have a kick-butt team. And I got to tell you, like, being on the phone with some of the largest and most accomplished data center folks in the world, and having our team impress those people, I mean, that to me has honestly surprised me the most. Because again, I mean, we're really good at what we do. But you know, we're getting into a bit of a new lane here. And we've been studying it extremely hard.

So I think, to add some fuel to Paul's fire on the customer side, I mean, the customer conversations are red hot, our phones are ringing off the hook. And I think our team is really doing an incredible job.

[Paul Prager, executive]

I guess for me, what I've learned most in terms of over the course of the last 90 days is how different the requirements are for each one of these customers. And yet the 1 uniform thing that's missing, as sophisticated as these customers are, is a lack of familiarity with power and energy infrastructure. And as we have these discussions, and they've advanced, it's been great to watch these potential customers appreciate how important solving the energy infrastructure problem is, and the value of partnership in getting that endeavor completed.

So I think we're spending a lot of time educating these folks about energy infrastructure, how we do it, how we suggest they do it. And they're teaching us a lot about their requirements. And so I think it's pretty fascinating. And for me, it's far more than, it's not fascinating because, oh, it's interesting. It's fascinating because I believe that it will jettison TeraWulf to be an important company, I think, as we continue to expand into HPC/AI, given the backbone of the energy infrastructure that we have.

[Operator]

The next question is from Joe Flynn from Compass Point Research & Trading.

[Joseph Flynn, analyst]

And to kind of piggyback off the earlier question, I was wondering if -- I mean, would you expect to have a 20-megawatt customer by year end? And with the forward guidance on, the building -- build of co-location Building 2 in future capacity, is this, kind of a, build it and they will come situation? Or would you expect to have customers secured beforehand?

[Patrick Fleury, executive]

Yes, look, Joe, it's Patrick. I'll answer that. I mean, look, I would just say, we put a slide in the deck that updates on timing. And, we have -- we're having constant discussions where folks are looking for hundreds and hundreds of megawatts of capacity. So, I would just say, as Paul echoed, we are really -- every day, everyone on the team focused on making sure we've got the right customer or group of customers to achieve the maximum valuation multiple. And so, if that takes a little bit longer than expected, then it does. But I think, there's a balance here of we know power is in short supply. We have it.

And then we also know folks are scrambling for it. So, I think -- I'm not going to answer directly your timing question. But, there's kind of a sweet spot, as we all know, over the coming months. And that's what we're trying to align with.

[Joseph Flynn, analyst]

That's helpful. And then, I guess, as we look, like, into the future in 2026 and 2027, you talked about, like, the hundreds of megawatts type deals. Like, what kind of margin profile should we expect as, like the time moves out?

[Patrick Fleury, executive]

Yes, look, I think, you can see on Page 15, we did update that page around the edges, if you kind of compare it to our page from the last deck. I think, we've kind of stuck with 65%, 75% margins on this co-location business. I know some of our public -- or one of our public peers has put out a higher number. We're going to stick with that sort of 70% for now. And once we have some operations under our belt, we'll update that. But that's kind of -- that's what we're sticking with for today.

[Operator]

The next question is from Darren Aftahi from ROTH Capital Partners.

[Darren Aftahi, analyst]

Can you talk about how the Williston -- sort of the role it plays relative to the 20-megawatt facility you're building, I guess, asked another way, is this something that you would showcase for clients that are potentially interested in the bigger capacity?

[Patrick Fleury, executive]

Yes, that's exactly right, Darren. I think it's -- and again, on the -- as we talked about buying these GPUs, and you can see, as we've kind of laid out on Page 15, that is not our core business. But it is a show me story. And, some of the folks that we're in negotiations with want to come on site, want to see that we can run the greatest and latest technology. And so, that's our commitment there.

But again, these are single-digit millions of dollars, right, that we're spending on that business, on the sort of GPU as a service business, if you will. And our major and main focus remains co-location.

[Darren Aftahi, analyst]

Great. And then just 1 more. I noticed in the release, you guys talked about Building 5 as having optionality for either BTC or HPC. I guess, like, when it's complete, is it -- what's the calculus from your perspective that determines which way you go there?

[Patrick Fleury, executive]

Yes, look, I think, as we've said before, there's other -- right, there's -- it's not just, a capital allocation decision for us. So, we have contractors and subs that have been on this site for 3 years, right, and they've built Building 1, Building 2, Building 3, Building 4, now, they're moving to Building 5. So, part of that is we're competing with Buffalo Bill Stadium, for example. So, we need to keep those contractors and subs on-site working. So, we're moving ahead with Building 5. Yes, Building 5 does have some optionality, should we land a customer very, very soon where we could pivot that building. But just the order of magnitude, right, we're talking about, call it around $25 million to build that whole building. And, roughly 50% of that can be repositioned as, high-power compute AI infrastructure.

So, Darren, it's small dollars to kind of keep our people at the site, keep them working so that we've got runway for the next, 2, 3 years to build out hundreds of megawatts. So, that kind of comes into the calculus as well. And again, to kind of pivot, you're talking about sort of 50% of that CapEx that's reusable. Does that make sense?

[Operator]

The next question is from Lucas Pipes from B. Riley Securities.

[Nick Giles, analyst]

This is Nick Giles on for Lucas. Guys, I appreciate all the color so far. Your CapEx on the CB-1, I believe it implies a CapEx per megawatt of $5 million. Should we think about this as a good go-forward target? And, what levels of redundancy could influence this figure? I mean, Paul, you said it earlier that, each customer has different needs. So, just trying to think about how that figure could potentially flex?

[Patrick Fleury, executive]

Yes, Nick. So, it's a good question. So, yes, it's roughly $5 million. As you can see on Page 15, we did update our build cost per megawatt to $6 million to $8 million. Generally speaking, like, that's specific for our site. And what I mean by that is, if you were to add, for example, backup diesel reciprocating generators, which we're not planning to do, that would cost you at least another $2 million per megawatt. And so, again, our site is very unique in that it's a former pull plant, right, retired and reclamated.

So, we have grid infrastructure in place that you typically would not build for a data center. For example, we have 2 345 kV lines that come into our substation and multiple transformers, right? So, what that means is, unlike a sort of typical data center that has one line coming in, if we lose one of those lines because the transformer goes down, we just move to the other line. And so, when you think about that redundancy, I mean, really, you're talking about you're only going down in a blackout, which happened once in kind of the last 60 years, right? And so, that, again, makes our site very unique and our build cost unique to this specific location.

[Nick Giles, analyst]

Patrick, that's super helpful. My next question, obviously, the team has really deep experience in power and infrastructure, TeraWulf and even Beowulf included in that. Good to hear that potential customers are recognizing this. As it relates to data center design, are you using any third parties today, or is there any appetite to bring in any partners on either the data center design or operations side?

[Patrick Fleury, executive]

Yes. So, Nick, we, and I'm the wrong guy to answer that question, and I know my team is spread out all over the place, so I'm not exactly sure who's online, nor am I sure what we're willing to disclose, but the short answer is yes, yes, and yes. We are partnered with a variety of different partners on all aspects of the HPC/AI, or what I would call, like, the data center of the future build. And literally, that's everyone from NVIDIA, Supermicro, Dell, HP, like, all the way down the line.

So, yes, that is a collective effort led by Sean Farrell, SVP of Operations on our team, and Nazar Khan, COO and Co-Founder, and their whole team, and augmented by multiple outside parties. So, yes, that is a big team effort, and like I said, been just really, really impressed with the team that Nazar and Paul have assembled there.

[Nick Giles, analyst]

Got it. Got it. Well, appreciate all the detail, and Patrick, to you and the team, continue best of luck.

[Operator]

The next question is from Brett Knoblauch from Cantor Fitzgerald.

[Brett Knoblauch, analyst]

And thanks for all the color you guys have shared. Maybe to start, for Lake Mariner, I know you guys have talked about that facility being able to go to 770 if you do an additional interconnect. Would that occur before you guys consider to drop down additional assets into TeraWulf from Beowulf? I guess, can you just help us understand, like, the timing and what would happen between the 2?

[Patrick Fleury, executive]

Yes, that's a good question. I don't know, if Paul wants to address it, but my answer is not necessarily, Brett. Some of the customers we're talking to, have demand. I mean, if you think about what we have, we have close to a gigawatt of low-cost, predominantly zero-carbon power. I mean, that is extremely rare, right? And that's, by the way, what's sitting, effectively in the public company and what we have access to also in the private company. But I don't know, Paul, if you want to add any color to that.

[Paul Prager, executive]

Yes, I think we've discussed this before, Brett. The priority is absolutely TeraWulf, and they have a call on -- the opportunity to expand. And obviously, processes are involved, right? You've got to figure out what's fair. You've got to go to your audit committee. You've got to get a fairness opinion. You've got to do all this stuff. But at the end of the day, we're going to be most responsive to customer demand and not assume, we don't want to assume a whole lot of risk here until we've decided the customer path. But we're not going to run out of the opportunity to access power anytime in the very near future or further out.

I mean, TeraWulf has what it needs. And obviously, to take that site and to continue on at it makes sense because of scale, familiarity with the site, and familiarity with the stakeholders there, which includes, some state agencies. So, we're most excited about Lake Mariner and our build costs there and our continuing ability to grow the usefulness of that site.

[Brett Knoblauch, analyst]

Perfect. Appreciate it. And then maybe just one follow-up on, I guess, just how we should be thinking about the AI/HPC capacity coming online. I think I'm reading in the fine print on one of the slides that you expect the 20-megawatt building to be substantially complete by end of the year. I guess, first, should we expect by the end of this year GPUs to be plugged in there?

And then secondly, I'm also seeing that you guys are expecting to have 150 megawatts of gross power capacity for AI/HPC in 2025, which is, I guess, a bit more than I was expecting or thinking about before this call. So, does that kind of assume that you'll have a couple additional maybe 50-megawatt buildings that you would look to construct in the back half of next year?

[Patrick Fleury, executive]

Yes, Brett. So, I hate to do this, but I was persuaded by both, our team and investors and analysts. But if you look at our deck on Page 16, we did put together an illustrative development timeline to address exactly that question. And so, I just want to point out, though, that that timeline can be fed up if we need to based on customer demand, or it can be slowed down. So, it's not absolute, but I think generally speaking, we feel pretty comfortable with what we put on Page 16 from a timing perspective. But again, like I said, it's very dependent on customer demand.

[Operator]

[Operator Instructions] The next question is from Bill Papanastasiou from Stifel.

[Bill Papanastasiou, analyst]

Congrats on another quarter of solid unit economics. For my first question, just wanted to touch on that 150 megawatts of growth in 2025, part of your WULF Compute business. Can you give a breakdown in terms of in terms of the makeup of that 150 megawatts? How much of that could potentially be coming from WULF Den operations versus your CB-1?

[Patrick Fleury, executive]

Yes, Bill. So, good question. Again, I'll point you to Page 16, where we kind of laid it out in WULF Den co-location Building 1, which is what we referred to as CB-1, CB-2, and then sort of additional co-location capacity. So, it's all kind of laid out there in black and white. Take a look at it. I think if you have questions, again, holler at us. But it's meant to at least provide a guideline of what we think is very, very achievable. And then, like I said, that could kind of be sped up or slowed down depending upon demand.

[Bill Papanastasiou, analyst]

Okay. Awesome. I see that now. And secondly, now that the term loans have been extinguished, how should we think about dilution going forward to fund Bitcoin mining and their aggressive HPC/AI growth strategy?

[Patrick Fleury, executive]

Yes. Look, Paul, do you want me to answer that and then you can chime in after?

[Paul Prager, executive]

Sure.

[Patrick Fleury, executive]

I mean, look, Bill, I think, there's a lot of different places to take that question. But as you know, I think we have among the largest, insider board, management team ownerships of the stock. So, we've always been very conscious about dilution. We've also been willing to use it when it makes sense. And so, I think paying down the debt, and then positioning us to really expand into high-power computing AI, our team's view was that is a very productive and accretive use of dilution.

But I think now, right, if we can get the financial engineering correct with the right customer, then we have a free cash flow positive Bitcoin mining business, which by the way, if you look on Page 11, is among the most profitable of any of our peers. And, put a nice quote in there, it's not the size of the dog in the fight, it's the size of the fight in the dog. I mean, we're, small compared to these other folks that are, trying to build Exahash to the moon, and we're generating more profit.

And that's important, because if we can get the financial engineering correct, like one of our peers has done in the high-power computing AI business, then we have a business that's generating positive free cash flow, while we build out a massive high-power computing AI business. So, that is the goal. And if you kind of read between the lines, that would mean very little incremental dilution. So, that's what we're trying to achieve. Whether we get there or not, and how quickly, I can't answer that today, but that is the goal. Paul, do you want to add anything?

[Paul Prager, executive]

Yes, just 2 things. Listen, we made the decision to issue equity to pay down the debt, on the back of 2 things. One, we were uncomfortable how having the debt would affect our flexibility, if we wanted to make a deal in the HPC/AI space, and we didn't want the tail to be wagging the dog. I mean, the value proposition from HPC/AI is, it's just a whole different magnitude. So, we felt it was critical to pay off the debt.

In retrospect, we're delighted that we paid off the debt, given the volatility that the mining space has gone through. But more importantly, given the opportunity set that we've become aware of in the HPC/AI world, and to have the time to thoroughly review each one of the HPC/AI customer candidates, to review their credit, to be thoughtful about how long will they be in the business, and what commitments we should expect from them and they could honor.

I think that's a luxury that we would not have had, and we still had the lenders sitting out there. So, I get it, we had to dilute a little bit, but it was accretive, it was conscious, it was intentional. I think, I'm so proud of the team for being committed to do that, because I think having, having a company that's debt-free and the opportunity to have free cash flow at the levels that we do, and having the time to be able to come up with the right HPC/AI customer conclusion, as opposed to sort of just race and, marry the first girl who asks us out on a date, we ended up in a whole different place. So, that was our thinking, and I believe, ultimately, it was in the absolute best interest of our shareholders.

[Bill Papanastasiou, analyst]

Appreciate the color. Congrats again on the quarter.

[Operator]

The next question is from Kevin Cassidy from Rosenblatt Securities.

[Kevin Cassidy, analyst]

Just a question around, as you're building out the HPC/AI business, with, lead times on these GPUs are stretched out quite a bit. Do you have secure deliveries, or are some of your co-location, your customers, going to help you bring in the equipment?

[Patrick Fleury, executive]

Yes, good question, Kevin. It's Patrick, and thanks for joining the call. So, again, that's part of the customer, as Paul likes to say, this sort of get-to-know-you process, right? I mean, a lot of customers, as you know, are very well funded, but if they're not in the right queue with NVIDIA, then it doesn't matter. So, I think as part of that diligence process, we are, for example, like, we are in touch with NVIDIA regularly, because you think about, again, what's in the public company, what's in the private company, some other stuff we look at, like, I mean, we have close to 1,000 gigawatts of power. So, excuse me, 1,000 megawatts of power. So, from a GPU perspective, I mean, that moves the needle for NVIDIA. I mean, that's $30 billion of GPUs. So, yes, I think that is certainly one very important aspect of the diligence process that, we are in touch with when we're talking to customers.

[Paul Prager, executive]

Yes, I would like to add just a little bit to that. I can't speak for NVIDIA, but it would appear to us, talking to them and to their customers, it appears to us that they are keen to have allocations to customers other than, the hyperscalers, and as such, and therefore, diversify their customer base, right? So, it makes sense. So, I think that it won't be an issue to have a customer, if it's not a hyperscaler that if you will, has an allocation.

I think the second thing is, and this is great for the team here, is that I think NVIDIA is basically telling people, hey, if you want your allocation, show me where you're going to put them. Show me that you have the power, you know? And so, that's creating some real energy on the part of potential customers to strike a deal, because they move up, if you will, in their supply queue on their side of the equation when they have access to megawatts. So, I think that is a good, healthy pressure that we can be the beneficiaries of, as long as we think they're the right partner, have the right culture, and could ultimately scale with us.

[Kevin Cassidy, analyst]

Great. Yes, I understand that. That's good for the diversification. I'm sure NVIDIA doesn't want to have all of their GPUs going to companies that are just competing with each other. So, great.

[Operator]

There are no further questions at this time. I would like to turn the floor back over to Paul Prager, CEO, for closing comments. Paul, you can go ahead.

[Paul Prager, executive]

Sorry, figuring out my mute button. In closing, I want to thank you all for joining us today. TeraWulf continues to lead the industry with best-in-class assets, lowest-unit economics, unmatched scalability of our owned infrastructure. We remain committed to consistent growth and profitability that aligns with investor interests. I am personally, and as a company, we are entirely optimistic about the future and confident in our ability to deliver sustainable growth and value.

I want to thank you all again for your time, and we look forward to continuing to respond to you as you think about TeraWulf as an investment and any questions you may have. Thank you.

[Operator]

This concludes today's teleconference. You may disconnect your lines at this time. Thank you for your participation.