Greetings, and welcome to the TeraWulf Investor Update Call. At this time, all participants are in listen-only mode. The fireside chat session will follow introductory remarks, and as a reminder, this call is being recorded. It is now my pleasure to introduce your host, John Larkin, TeraWulf's Director of Investor Relations. Thank you, you may begin.
Thank you, and welcome to the TeraWulf Investor Update Call. 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 Q&A session. When used on this call, the words anticipate, could, enable, estimate, intend, expect, potential, believe, 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 investor section of our corporate website at terawulf.com. Finally, please note that today's call will refer to certain non-GAAP financial measures.
Please refer to our company's periodic reports on Form 10-K and Form 10-Q and our website for a full reconciliation of these non-GAAP performance measures to the most comparable GAAP financial measures. We'll begin today's call with introductory remarks from Paul and Nazar, and then we'll proceed to the Q&A portion of the call. It's now my pleasure to turn the call over to TeraWulf CEO, Paul Prager. Paul.
Thanks, John. Good morning, everyone, and thank you for joining us today. At TeraWulf, our mission has always been clear: maximize the value of our scalable and sustainable digital infrastructure. We began by delivering low-cost, predominantly zero-carbon Bitcoin mining, a strategy that has served us exceptionally well. Today, we are proud to announce the next step in our evolution: our strategic expansion into high-performance compute to meet surging market demand. Over the past nine months, we've been disciplined and deliberate. We've engaged with nearly 25 counterparties, conducted exhaustive due diligence, and prioritized certainty and execution over speed. For us, this is not about signaling potential. It is about delivering results underpinned by finalized and tangible agreements. Today marks a milestone for TeraWulf: our first HPC hosting agreement with Core42.
The Core42 lease covers 72.5 megawatts of gross capacity, the equivalent of 60 megawatts of critical IT load, with an option to expand by another 135 megawatts of gross capacity or 108 megawatts of critical IT load. The data center lease is executed through La Lupa, a newly formed and wholly owned TeraWulf entity created to support this strategic expansion into high-performance compute. As a quick aside, the name La Lupa carries special significance. Inspired by the mythological she-wolf who nurtured Romulus and Remus, La Lupa represents strength, resilience, and the foundation for greatness. Just as she enabled the creation of Rome, this data center lease with Core42 marks the cornerstone of a significant extension of our digital infrastructure at TeraWulf, one powered by AI-driven compute infrastructure. Why Core42 and TeraWulf? Well, the collaboration between Core42 and TeraWulf represents a powerful alignment of strengths, expertise, and vision.
Core42, a wholly owned subsidiary of G42, is a leader in the global digital transformation and compute markets, actively working to build a worldwide footprint of AI compute capacity. The data center lease with Core42 is backed by a parent guarantee from G42, further strengthening our financing strategy. To ensure we have this right, we partnered early with top-tier advisors: Milbank as legal counsel and J.P. Morgan and Morgan Stanley as financial advisors. At TeraWulf, we sought a long-term tenant that we can grow with, one that shares our vision for innovation, sustainability, and scalability. Equally important, our alignment on operational priorities and growth potential makes this strategic relationship a natural fit. The opportunity for expansion is a key driver behind this agreement.
Beginning with a 72.5 megawatt deployment, which represents 60 megawatts of critical IT load, the data center lease allows for significant scaling potential of up to 135 megawatts of additional gross capacity or 108 megawatts of critical IT load. After an exhaustive review of sites across the United States, Core42 selected TeraWulf for two critical reasons: our future-ready, purpose-built infrastructure and our proven ability to deliver and operate large-scale high-performance projects. To that end, over the past year, our team has collaborated extensively with NVIDIA, Dell, and Hewlett Packard Enterprise to design future-proof data centers capable of supporting next-generation GPU clusters.
This announcement highlights the strength of that work and our commitment to providing state-of-the-art digital infrastructure that evolves with our tenants' needs. With that, I'll turn it over to Nazar Khan, my partner and TeraWulf's Chief Technology Officer, to walk you through the details of the data center lease. Naz?
Thank you, Paul. Let me outline the key components of the data center lease. Under the terms of the deal, TeraWulf will deliver 72.5 MW of gross capacity equivalent to 60 MW of critical IT capacity of state-of-the-art data center infrastructure at our flagship Lake Mariner facility in New York. The rollout will occur over the first three quarters of 2025. Under this data center lease, TeraWulf will provide low-cost, predominantly zero-carbon power and operate and maintain purpose-built infrastructure. Core42 will supply GPUs and associated equipment, including servers, networking, and compute equipment. This agreement marks a major milestone in G42's U.S. expansion. Our role is to maintain a secure facility, operate the data center infrastructure, and pass through all power costs directly to the tenant. I'll now walk through the economics of the deal. From an economic perspective, the fundamentals of the data center lease are very compelling.
Key highlights include a 10-year initial term with two 5-year extensions, 60 MW of critical IT load, which is equivalent to 72.5 megawatts gross, assuming an approximate PUE of 1.25, revenues of $125 per kilowatt month, which equates to $1.5 million per megawatt per year with a 3% annual escalator, and finally, a 12-month revenue prepayment credited back through a 50% reduction until fully repaid. Additionally, Core42 has the option to expand their footprint by up to 135 megawatts gross or 108 megawatts of critical IT load at market rates. This option also includes a 12-month revenue prepayment. When you look at the return profile, this data center lease becomes even more compelling. With $1.5 million of revenue per megawatt annually on approximately $6 million per megawatt build cost at a 70% EBITDA margin, we estimate a 17%-18 unlevered yield with an extremely well-capitalized, high-quality tenant.
These metrics underscore the tremendous value unlocked from our infrastructure and highlight our disciplined capital deployment strategy. With that, I'll turn it back over to John.
Thank you, Nazar. So in addition to Paul and Nazar, we're also joined by our CFO, Patrick Fleury, and our Chief Strategy Officer, Kerri Langlais. So I have a number of questions that have come into the team over the past month, specifically on the HPC AI compute business from the investment community. Also, as we referenced in our press release, we're receiving questions at info@terawulf.com. So I'll be mixing these questions in today. Paul, I'm going to start with you. Can you tell us more about Core42 and G42 and their focus on AI?
Sure. G42 is a global leader in AI and digital infrastructure, and Core42, their wholly owned subsidiary, is a key enabler of AI compute infrastructure. G42 is committed to cutting-edge infrastructure globally with a focus on sustainability and scalability. Both G42 and Core42 view this lease as a key milestone in their mission to empower nations and enterprises with AI solutions.
Patrick, could you discuss the economics of the first 72.5 megawatts from the TeraWulf perspective?
Yeah, of course. Nazar just touched on it, but I'll revisit, but based on our current budgets, the numbers are broadly consistent with what we've been discussing since putting our slide 16 in our investor deck in May of 2024, and it's a slide we've repeated each quarter, and the focus is on that middle column, co-location, and so just to repeat, it's $5 million of CapEx per gross megawatt, excuse me, and then adjusting for a 1.2 PUE drives CapEx of $6 million per megawatt of critical IT load. $1.5 million of revenue per megawatt at a 70% EBITDA margin is $1.05 million per megawatt of EBITDA, so as Nazar mentioned, we think about everything in yield.
When you think of those figures, it's $1.05 million of EBITDA per megawatt divided by $6 million of CapEx per megawatt of critical IT load, and that's effectively a 17% to 18 unlevered return. Just as a reminder, the 12-month revenue prepayment is recorded as deferred revenue, and that'll be credited back to Core42 over approximately two years.
Nazar, we've gotten this question many times over the past month. Were there any specific sticking points in the negotiations?
Thanks, John. There were no specific sticking points. The discussions were more focused on coordination of various design elements, operational parameters, and delivery timelines. The lease document is a fairly complex agreement. Each line in the document represents economic value, both for our shareholders as well as the Core42 stakeholders as well. And so it takes time to work through this agreement. And let's not forget, this is a 10-20-year agreement. So every line in that agreement was diligently reviewed and analyzed by both sides.
Okay, Kerri, we have an email question about what happened to the other parties given how many conversations TeraWulf had over the past nine months. Where do those broader conversations stand given that one customer spoke for all the initial capacity?
Thanks, John. So as Paul mentioned in his introductory remarks, we started this process nine months ago with more than 2,000 parties, and over the course of our diligence, had sort of whittled that down to less than a handful of people at the very end. So we did have other counterparties who were very close to the finish line here on this initial 72.5 MW. And while we opted for one customer to take that full amount of 72.5 MW, we do remain in dialogue with the other parties. These discussions are ongoing, and we're actively exploring opportunities for them to become tenants at Lake Mariner or potentially other sites that we are currently evaluating.
Nazar, we have a few questions focused on the option for additional megawatts. So maybe if I can just dial out and ask higher level, could you just talk to the specifics on the option for additional power?
Sure. As noted earlier, the agreement reflects an option for 135 megawatts of gross incremental capacity, 108 MW of net critical IT load. This option reflects a balance between Core42's potential near-term scaling needs and capacity that we can deliver over the next 12-15 months. Given the supply-demand balance in the market, most compute customers want to option as much capacity as possible. But our focus has really been on delivering tangible capacity, and so we really focused on where we could get to over the next 12-15 months as well as where Core42 thought their business would lead them. In terms of pricing, as Patrick mentioned, we think about these agreements in terms of yield. We're spending CapEx to put a building up, and we're earning a fixed revenue over a long period of time from a high-quality tenant on the back end.
So our initial deal, as well as all future capacity, will be targeting that similar 19 unlevered yield.
Paul, a timing and capital allocation question, if I could. Can you share your thoughts on the deployment timing on the remaining megawatts that are allocated to the AI compute business? And then maybe at a higher level, can you share your thoughts with everyone on capital allocation between Bitcoin mining and AI compute?
Sure. Sure. I think of the deployment of the 500 MW that we've currently circled for AI compute, and that is not all of our development pipeline. But I think that takes three to three and a half years. So I think it's safe to assume 100 to 150 MW a year. Everybody is in a race to power right now. And the question really is, is it a 100-meter race, a 400-meter race, 1,600-meter race? We've been in infrastructure for 30 years. Infrastructure is not a 100-meter race, though what seems to be a driver for the customer today is immediately available power. Regarding the megawatts that we currently use for Bitcoin mining or the 250 MW that we've told the world that we have allocated to it, we expect at some point there's going to be a natural fork in the road for us.
We think it's probably going to be the next halving, where given the required substantial CapEx upgrades, we would, if the market is then as it is today, we would likely shift those megawatts over to high-performance compute. But in the meantime, the 250 MW will generate a ton of free cash flow for us. And as you know, John, we've just recently upgraded our machines, so we're state-of-the-art, and I think we're at 18 joules per terahash in terms of efficiency. In the meantime, I think it's very helpful to have the flexible load of Bitcoin mining alongside the less flexible load of AI compute as we ramp this business, and I know that has given some comfort to the grid operator.
Thank you. So we fielded many questions on the two big variables in the AI compute model, specifically PUE and gross margin. So I'm going to split those in two. Patrick, maybe you can go first. Can you talk to TeraWulf's PUE guidance of 1.25 and then specifically why this first slug of capacity is coming in very slightly better than that?
Yeah, sure. So as mentioned, this is coming in at 1.2, and the reason for that is we're through the design stage. We've got a very good sense at this point on how the design will use power and how much power will actually make it to the GPU. So I think as folks know, this is one of very few operational data centers, call it in the first half of next year, with this type of power density, liquid cooling, and the designs and layouts can be different by each client. So it's really difficult to predict the PUE with any precision. But look, I think there'll be a lot of discovery over the first 12 months by us and our customers. And I think, remember, relative to other locations, Lake Mariner is really blessed with a very favorable ambient temperature environment.
Pool on the shores of Lake Ontario, that really helps. We've had no 100-degree days ever recorded at the site in over 40 years, and then just for folks that are looking further out, just note that our PUE guidance on the option and future capacity remains at 1.25 because we don't yet know the specific design specs for that additional capacity.
And then, Kerri, could you talk to how investors should think about gross margin guidance of 70%, given that investors are very aware that another signed deal has been guided to 80%? And so if you could break down the major drivers, the cost structure to help investors think through the gross margin structure of our AI compute business.
Sure, John, and I think this largely dovetails with the question that was just fielded by Patrick in terms of our guidance on PUE. I mean, the reality is there are no legacy data center designs that are applicable here, right? We're building the data centers of tomorrow, and there are very few, if any, facilities operating at this power density with liquid cooling, so given how advanced these data centers are and the lack of sort of a track record for operating them, we are guiding on the very conservative end of gross margin, and it's driven by really two key buckets of costs. There's the people costs, obviously, that includes site labor, maintenance, the repair, security, both in terms of cyber and physical security, and technical support.
And then there's sort of the bucket for other costs, and that's things like insurance, audit and compliance, lease expense, and fiber or water costs. And so when we think about all of that together and relative to our guidance of 70%, I mean, we're very confident in our ability to execute there. And I think over time, we expect that we will likely be ahead of that. But I think at this point in time and where we are in the development and construction and rollout of our facilities, that's where we're comfortable guiding our investors.
Paul, I know you fielded this question many times, but I think it's an important one. How sustainable do you think the $1.5 million per megawatt revenue number is going forward?
I think it's entirely sustainable. I remember a dinner, I think at the Cantor Conference, where everyone in the space was around the table, and they asked if the CoreWeave deal was a unicorn. And I think I was in the minority, and I said, "No, that deal is doable again, if not at better terms, and we're here now." I think the market's gotten tighter over the last few months. Specifically, we continue to hear from potential customers who need near-term access to power. And we, as you know, have the access to power with infrastructure, land, and water. Our sites are in a zone with an oversupply of power. It's also 91% green power generated in that zone. And our power costs are typically below market averages of about $0.045.
So given our power cost advantage, our cost-to-build advantages given our existing infrastructure, I'm certain that this is sustainable given the demand outlook that we currently see. I mean, at the end of the day, revenue per megawatt is not the primary focus, as it entirely depends on build costs. As Nazar has said, Patrick has said, Kerri just said, we're focused on yield. We will continue to focus on achieving high-teens, unlevered yield on our data center developments.
That's a good segue. Nazar, we have a question coming in regarding build costs. Could you just talk about how confident do you feel about the $6-8 million per megawatt estimated build costs going forward?
Sure. There's the current cost, and we've guided to we're delivering capacity in the first and second quarter of next year. And so the budgets for CB1 and CB2 are fairly well-defined. And so we are pretty confident in that range for the existing capacity. As we look out, there are certain long lead-time items that are a significant component of the overall construction cost. And so, as Paul mentioned, the markets are getting tighter. And so we're seeing some pressure on those things. So depending upon when we would go back to market to lock in pricing for those things, we may see some increases. But generally, we're pretty confident that being able to deliver on that $6 to 8 million per megawatt for buildings beyond CB2 is possible, again, if we're kind of back in the market here within the next three, four months or so.
It's important to remember that part of this is also kind of we have a fairly dynamic market with respect to the GPUs and equipment that we'll be hosting. We're going from Hopper to Blackwell to Rubin on the NVIDIA roadmap. The designs that we have are flexible enough to kind of roll with those upcoming changes. NVIDIA is kind of out there saying that every six to eight months, they're going to come out with a new design. Again, as everyone's mentioned, our focus is really delivering a high-teens, unlevered yield on cost. Again, where we sit today and the work that we've done on CB1 and CB2 gives us comfort that we'll be able to maintain that general CapEx target.
Kerri, could you take this one? Investor asking if we could discuss TeraWulf's development pipeline and how actively the team explores new opportunities?
Sure. We are constantly evaluating opportunities for additional sites. Our management team has decades of experience in developing, financing, buying, operating, selling power projects. So we're capable of evaluating a lot of different opportunities and sussing out pretty quickly whether they're appropriate for TeraWulf. So many of these sites come with complicating factors. It's actually sort of our bailiwick in terms of being able to work through potential site issues. Maybe it's a site that was a former power plant that's been retired, or maybe it was a former paper mill, things of that nature that take time to evaluate and sort of work through. But that's where our skill set really sets us apart. So that expertise allows us to identify and pursue projects that align with our strategic goals and ensure that we maximize both value and efficiency and additional scaling of our platform.
Patrick, we have a few questions regarding the TeraWulf deal versus another announced deal, I'll say. So I'll paraphrase a bit. As Paul mentioned, many people thought the CoreSite, CoreWeave deal from an economic perspective was a one-off. Your thoughts on how the deals compare?
Yeah, look, I'm going to focus on our deal, which I think, as folks hopefully can see, we got great terms on our deal. It's very consistent with what we've been disclosing to the market since May. But it's also a very different structure, and we've been steadfast about that since we started talking about this earlier this year. And that is that we funded the CapEx on our construction. But look, we think it's a fabulous deal for Wulf and Wulf shareholders. I think that last point, though, I just want to drill in a little bit more there. So we sold an asset, right, in our Nautilus facility, reinvested that at Lake Mariner, and then we also executed on a convertible financing in October. And those two things allowed us to fully equity fund Wulf then, CB1, and CB2.
Now, given we have this lease, we will look to project finance, in other words, take ourselves out of that equity in the future with debt. Then we will also look to project finance future projects so that we do not need to use that equity funding in the future. I think if folks remember, on page 13 of our last quarter investor deck, we showed if we accomplish that, we have over $380 million of unallocated cash available next year to fund returns to shareholders. I think that is a very big difference that, again, we have been steadfast on since we first started talking about entering this business. There are other aspects of the deal that I think are maybe viewed as minor but are really quite a big deal for us. For example, we have a 3% annual escalator.
When you compound that over 10-20 years, that's a big deal, and I think our counterparty quality is extremely high. We have a corporate guarantee from G42, which is backed by a sovereign government with a credit rating and CDS that trades virtually on top of U.S. government CDS. Moreover, G42 is a beneficiary of a very large investment from Microsoft alongside some of the most successful and largest fund managers in the world, so we feel very strong about the credit quality here, about Core42, G42, and the sovereign that supports them, and just to be clear, John, I know I just thought we got a couple of comments here. Sorry, but to be even more clear, we do not need any additional financing for this first 72.5 megawatts. Again, that was equity funded.
Now that we have a lease, our intent is to actually debt finance ourselves out of that equity funding, just to be really clear. We think, look, it's early days, but we think that cost in the beginning, and it depends on, frankly, whether we go to the typical bank project finance market or Term Loan B financing markets. We expect that will be single-digit cost of capital and amortizing around within sort of the life of the loan. Remember, we do have two five-year extensions beyond the initial ten-year term. I would expect that the term of that debt is maybe 10 to 15 years.
Great. And then any thoughts on timing on the first financing kickoff?
Yeah, look, like we said, we've hired Morgan Stanley and J.P. Morgan, and we expect to begin working on that in Q1. But I think the execution of that will likely be mid-year.
Got it. Nazar, I think a very good question that we received this morning. Will the next lease be easier because TeraWulf has this one under the belt for Core42 and also if it happens to be a different tenant?
Yeah, the simple answer is yes. With Core42, we have an option embedded within the lease agreements, the 108 megawatts that we've talked about. A couple of questions have come in around timing of that. Again, that's geared towards the end of the year, first quarter of 2026, end of 2025, early 2026, in terms of the delivery of that. So yeah, with the existing tenant, there's an option there. It's fairly straightforward. And with others, it's also we've been now through a very detailed negotiation on every single parameter that comes up with respect to these agreements: service level credits, the operational parameters, what these customers care about in terms of humidity, temperature, efficiency. And so all of those things have been negotiated in a very detailed document.
Yes, on this next go-around, to the extent that there are other customers that we bring in at this site or others, we've got a pretty good framework to start working from.
Kerri, we haven't really talked much about the power source in New York Zone A. Do you think investors and/or customers appreciate the value of the green power source?
Thanks, John. I think right now, candidly, I don't think we're being fully rewarded for the value of our access to clean power. As Paul mentioned, and it's kind of been reiterated throughout responses to a number of the questions, time to power is universally right now the biggest challenge for our customers and tenants, and so as they prioritize immediate access to infrastructure, I think sustainability may not be the number one deciding factor. However, as enterprises increasingly scrutinize the power that they're consuming and align that with their own environmental goals, I expect the value and importance of our access to what is predominantly zero-carbon power to become a much more significant determining factor for capacity going forward, so in short, I don't think we're getting full credit for it today, but I do think it's going to be an elemental piece of the puzzle going forward.
Thanks, Kerri. Paul, I'm going to direct this one to you. Why do you like the high-performance AI compute business so much?
Listen, we've been in energy infrastructure for almost 30 years, and this team's been together for the vast majority of that time, and so in the power business, what we used to do was we used to find an off-taker for electricity. We'd want them to be a good credit, and we'd find a fuel supplier. We'd capture a spread. We would build the plant. We'd operate the plant. We'd finance that in the most efficient manner, and then we had predictable revenue for 15-20 years. That's what this business is, and in this case, we've got a long-term off-taker who's a phenomenal credit. Core42 is backed by G42, both top-tier customers of NVIDIA, both partners with Dell, G42 backed by a sovereign wealth fund, a tremendous reputation.
So we've got a long-term off-taker of phenomenal credibility, approved, if you will, by our financial advisors, J.P. Morgan and Morgan Stanley. We've got the site. We've got long-term supply of electricity. And we are building and operating. So Patrick, we'll get to work on a project financing so we could wash, rinse, and repeat. I mean, this represents really just about 10% of our capability and capacity over the next three to five years. So we've got a lot of work to do, but the revenues are tremendous. The credit and the partnerships are tremendous. And it's just like what we did in the power space for two decades.
That's helpful. Thanks, Paul. Patrick, this has been a recurring question that we've gotten. Can you give us your thoughts on TeraWulf's ability to become a REIT down the road?
Yes, sure. So that is something that we will most certainly explore. We've already begun implementing from a structuring perspective. But just to note, it's really important that folks realize that the optimal time to convert to a REIT is when the business is moving from a non-taxpayer to a taxpayer, given that when you convert to a REIT, you have to purge the company of retained earnings to become a REIT. We have a pretty significant amount of NOLs, and we're also poised to generate a pretty significant amount of free cash flow over the next three to four years. So given our NOLs and then that earnings profile, I think just very roughly the timing could line up pretty well with the next halving. So yeah, we have explored that on a preliminary basis. Nothing precluding us from becoming a REIT.
I think we'll expect to explore that in more details once we reach operational scale.
Okay. To finish off, we have a few of what I call micro questions that have come in. So we'll have a bit of a speed round of sorts. Answers can be brief. So here goes. Nazar, when will GPUs and servers begin arriving at the facility?
As early as January. So we'll start to receive equipment in January. Again, we've been guiding that initial building will come online towards the end of the first quarter of the year. And so equipment will be arriving on site very shortly. And then over the course of the first two quarters of next year, there's going to be over $2 billion of hardware that's going to show up at our site that's going to populate the 60 MW of critical IT load.
Kerri, what are the benefits of being in one site?
Thanks, John. Listen, there are several operational benefits to being at one site. We are blessed to be in a single-state ISO. We can leverage our operational synergies across the site, which streamlines our processes and reduces complexity. I think importantly, and we've talked about this before, the people who operate the Lake Mariner data center facilities are many of the same individuals who worked at the site when it was a 700 megawatt coal plant. And so that knowledge, that inherent knowledge of our infrastructure, how our facilities integrate with the grid, how to manage significant loads is really invaluable as we scale that business. So that, I think, is something that can't be replicated. We know every risk at the site and how to address it. We've had many, many years of doing just that at this location.
So the other thing I think that kind of dovetails with this is, as we have been through these conversations with our tenants over the last several months, we're also seeing value in zonal deployment of HPC data centers. So nearby data centers can share computing resources. For instance, tasks that might require less immediate processing time can be routed to less congested partner facilities, if you will, in the same regional zone. And so that also plays into our thinking as we think about or evaluate potential additional sites that we might be able to bring into the TeraWulf platform to further expand and optimize our footprint in that region.
Nazar, I have a couple of questions there, maybe two closely related, basically about the integrators. And so how closely did the on-site TeraWulf team work with the OEMs, the integrators, and potential customers during the design stage? And how many site visits were there?
We worked extremely closely with those folks. Members of the NVIDIA team have been on our site multiple times. In the press release, you'll see there's a quote from the Dell team. Dell is the system integrator for part of the deployment that Core42 will be doing here, and so Dell's been getting multiple visits to the site. We've got a very large working group between Core42, the Dell team, and our team in rolling out the capacity here over the next couple of quarters, so again, this has been, I'd say, a very coordinated and iterative process with both the OEM as well as the integrators to make sure the whole chain works, and so I think, as Kerri mentioned earlier, these are data centers of the future. The rack densities that we're building to are different than where the market has been going back three, four, five years.
And so each component, each phase has been heavily reviewed, discussed, and kind of we're working together on that. And that being said, as we go along, that intimate discussion is going to need to continue, kind of bringing up the capacity that we're talking about. And as mentioned earlier, over $2 billion of hardware will be at our site. It's going to require very close coordination with all those parties to ensure that Core42 is getting the most out of the equipment they have, and they're also getting the most out of the design that we have for the infrastructure that we've built.
Gotcha, and the next part was about how long will they be on site when the data centers come online. It sounds like quite a bit.
Yeah, yeah. It's going to probably be over a couple of months. And so just, again, kind of tuning once all of the GPUs are installed. Again, CB1, we've outlined a 16-megawatt critical IT load facility. It'll probably take eight-to-nine weeks to fully tune all of those GPUs to ensure that Core42 is getting the most out of them.
Paul, you've talked about feedback received from potential customers after visiting our site. Can you elaborate on what was the feedback?
Sure. I mean, it all comes down to our team led by Naz and Sean Farrell, working with the customers alongside the OEMs and the integrators to get to a custom solution. We had a couple of dozen site visits, and there was not one party that visited the site that then did not approach us to enter into an arrangement with us. We entered into the deal with Core42 because, again, we said all along there were three things we were looking for: first-class quality credit, and again, Core42 backed by G42, supported by a sovereign wealth fund. That's all you could ever ask for. Two, we wanted people we could scale with. Core42 did a tremendous amount of diligence relative to some of the hyperscalers are out and about, and they're really trying to option space.
Core42 really dug into design, brought Dell and NVIDIA into the picture early on to work on diligencing the site, what we were capable of there. I mean, they were really all over it. And so culturally, we really liked them a lot because of their intense focus on our site and what was capable there. So we're highly confident we're going to be able to work with them. And of course, the third thing that we focused on was yield. We liked the Core42 sense of commerciality because they recognized the availability of immediate power. And so we were able to strike a commercial arrangement with them where it made sense for both parties, and we thought we were going to get a premium return. And so that's how we ended up here.
Thanks, Paul. Patrick, a question on exactly when does TeraWulf start earning revenues? Is it when the servers are installed, or is there another specific date or milestone?
Yeah. Hey, John. So yeah, when the building is ready, TeraWulf delivers a commissioning report to our tenant ahead of certain agreed-upon dates, and that's when revenues begin. And I just want to go back to, I think Paul just mentioned something I just want to put an exclamation point on. The yield that we're talking about is unlevered yield. So then when you add in typical project financing debt at reasonable rates and at reasonable term, you're talking about a yield, levered yield that's in the mid to high 20s. And that's where I think this gets really exciting for us and TeraWulf shareholders.
Thank you. Hopefully, this has been helpful to investors. And with that, I'm going to turn it back over to Paul for closing comments.
Thanks, John. Today's announcement further reinforces our core expertise: the ability to deliver high-quality digital infrastructure. This data center deal positions Wulf at the center of two of the most impactful uses of sustainable digital infrastructure: Bitcoin mining and AI-driven compute workloads. As significant equity holders, we approach this process with patience and precision to deliver maximum value. The outcome speaks for itself, and I am particularly proud to share this milestone as we close out the year. And yes, finally, it's beginning to look a lot like Christmas. Thank you for your continued trust and support. We look forward to an exciting and prosperous 2020.