Terawulf Inc. Q4 FY2025 Earnings Call

Greetings, and welcome to TeraWulf Fourth Quarter and Full Year 2025 Earnings Conference Call. As a reminder, this conference is being recorded. I would now like to turn the conference over to your host, Mr. John Larkin, Senior Vice President, Director of Investor Relations. Thank you. You may begin.

Thank you, operator. Good afternoon, and welcome to TeraWulf's Fourth Quarter and Full Year 2025 Earnings Call. Joining me today are Chairman and CEO, Paul Prager; CTO, Nazar Khan; and CFO, Patrick Fleury. Before we begin, please note that our remarks today may include forward-looking statements. These statements are subject to risks and uncertainties, and actual results may differ materially. Words such as anticipate, expect, believe, intend, estimate, project, could, should, will and similar expressions are intended to identify forward-looking statements. For a discussion of these risks, please refer to our filings with the SEC available at sec.gov and in the Investor Relations section of our website. We will also reference certain non-GAAP financial measures. Reconciliations to the most comparable GAAP measures are available in our earnings release and filings. With that, I'll turn the call over to our Chairman and CEO, Paul Prager.

Thank you, John, and good afternoon, everyone. 2025 was a defining year for TeraWulf. We said we would transition this company into a scaled power-backed AI infrastructure platform, and we did. Our strategy is simple and disciplined: control energy-advantaged sites, engineer infrastructure around power, and contract long-term credit-backed AI capacity. Everything we did in 2025 supports that strategy. First, we acquired 100% of Beowulf Electricity & Data, eliminating related party complexity and fully integrating power generation expertise into our platform. In today's market, power is the gating factor. If you don't control power, you don't control your destiny. We do. Second, we secured long-duration site control at Cayuga, up to 400 megawatts at a retired coal facility with real grid infrastructure already in place, brownfield, power-backed, scalable. That's our model. Third, we signed a 450-megawatt lease with Fluidstack supported by Google's credit. That was a platform-defining deal. It validated our model, our execution capability, and our ability to contract at scale. With Google's warrants, it will be our largest shareholder. That alignment speaks for itself. Fourth, we replicated the model in Texas through the Abernathy joint venture, proving portability across power markets. And fifth, we executed the WULF Compute and Flash Compute financings. These transactions demonstrated that contracted credit-enhanced AI infrastructure supports scalable and repeatable capital structure. Operationally, we delivered WULF Den and CB1, began recording HPC revenue and have now delivered CB2A for Core42. We are building, delivering, and contracting simultaneously. And since year-end, we added approximately 1.5 gigawatts of additional power back capacity in Kentucky and Maryland. Now let's talk about what actually differentiates us: power and execution. The AI build-out is not constrained by GPUs. It is constrained by power, interconnection, transmission, and increasingly new generation. That's why Morgantown matters. Morgantown is not just another data center site. It is a former coal generation facility in the Washington, D.C., Northern Virginia corridor, one of the most power-constrained data center markets in the world. Our Phase 1 vision includes approximately 500 megawatts of new dispatchable generation, 250 megawatts of battery storage, and 500 megawatts of data center load, followed by a similar Phase 2. Critically, the site is being engineered to operate as a net generator to the state. We are not just consuming capacity. We are adding it in constrained markets, and that is the only sustainable model at scale. This industry is moving towards integrated bring your own generation campuses. We are well ahead of that curve because we are fundamentally a power company that builds and operates digital infrastructure, not the other way around. We know how to permit generation. We know how to build generation. We know how to operate generation. We understand grid behavior, and we know how to integrate generation, storage, and compute in a way that works for customers and regulators. There are very few teams that can do that credibly and at speed. We are premier among them. Turning to Kentucky. Demand is extremely strong. We are engaged with every major hyperscaler and several large AI compute platforms. A data room is open; diligence is active. Conversations are robust and substantive. This week, we met directly with Governor Beshear and state leadership. The alignment at the state and local level is clear and constructive. Kentucky understands the economic and strategic import of power-backed AI infrastructure. This is 480 megawatts, a campus with immediate power availability, expansion potential, and strong state support. We are excited and highly confident in the long-term value of this asset. While we execute on the platform, we are consistently evaluating a constant pipeline of additional opportunities. We review hundreds of sites, and most do not meet our standards. We are disciplined around four key elements: power control and durability, scalability in 250 to 500-megawatt phases, signed credit-backed contracts; we do not speculate, and fourth, capital efficiency. We turn sites away all the time. But when we do find one that meets these criteria, we move decisively. Importantly, we already control the sites necessary to deliver our targeted 250 to 500 megawatts of contracted capacity annually through the end of the decade. The runway is in place. Growth from here is execution. Finally, we are building the team to match the ambition of the platform. We recently added a senior data center construction lead for Meta, and we have strengthened origination, project management, commissioning, and cybersecurity capabilities. This business is one in the trenches of engineering detail, construction discipline, and operational rigor. We are staffed accordingly. So in summary, the strategy is clear. The sites are secured. The capital is in place. Customer demand is strong. The team is built. Now it is about disciplined delivery, turning contracted megawatts into energized capacity, and durable recurring cash flow. With that, I'll turn it over to Nazar.

Thank you, Paul. Let me start with delivery and risk compression. WULF Den and CB1 were delivered in the third quarter and generated revenue throughout the fourth quarter. CB2A is operational, and CB2B is expected to be fully online in March. By the end of the first quarter, all Core42 capacity will be energized and revenue producing. Following contract execution, Core42 requested incremental fit-out enhancements. We adjusted sequencing accordingly. The monthly recurring charge was revised; no penalties were triggered, and revenue commencement remains aligned with the customer's deployment schedule. Turning to the Fluidstack buildings, CB3 is expected to deliver in mid-May. After signing, the tenant finalized certain design optimizations, which we incorporated without changing building footprint or lease economics. The associated revenue timing impact has been incorporated into the financial model. Importantly, CB4 and CB5 were designed collaboratively with the tenant from inception. These buildings reflect a fully standardized, repeatable design and represent the majority of contracted WULF compute capacity. Several structural improvements materially reduce execution risk. First, electrical redundancy has been optimized and standardized. Second, trade stacking and sequencing have been refined to minimize rework. Third, long lead equipment was procured after final design alignment. And fourth, mechanical and electrical systems now follow a repeatable installation model. Execution risk declines as design standardizes, and CB4 and CV5 reflect a mature optimized build. Both buildings remain on schedule with targeted lease commencement dates of the third quarter and fourth quarter of 2026, respectively. Through design optimization, we increased critical IT capacity from 162 megawatts to 168 megawatts per building without impacting the base construction budget. That incremental 12 megawatts across the campus is expected to generate approximately $200 million of additional lease revenue over the initial term. Finally, Abernathy, our joint venture, remains aligned with its fourth quarter 2026 lease commencement and continues progressing under fixed EPC structure, which further limits construction cost variability. Execution requires managing scope, timing, and cost in real-time. We have incorporated adjustments transparently and remain focused on disciplined delivery. Large-scale AI infrastructure requires active management of scope, schedule, and cost. We have incorporated refinements transparently, preserved economics, increased capacity, and maintained budget integrity. Execution remains disciplined and on track. With that, I'll turn it over to Patrick.

Thank you, Nazar. The second half of 2025 was transformational for the company. We secured over $12.8 billion of HPC lease agreements, executed $6.5 billion of debt and equity-linked financing, and materially strengthened our balance sheet liquidity while carefully managing and minimizing dilution. Let's begin at a high level before diving into the financial details. We are a business in transition and executing on rapid growth. The 2025 results still reflect a meaningful contribution from Bitcoin mining and its inherent volatility, including commodity pricing and complex network difficulty dynamics. Over time, that volatility will decline as long-term credit-enhanced HPC revenues become the dominant driver of results. Importantly, while mining introduces revenue volatility, its flexible load profile has been strategically valuable at Lake Mariner, supporting demand response participation and power cost management. Mining is not a long-term growth focus, but it has enabled the transition. The 2025 results of operations reflect that transition in motion, and the balance sheet reflects that we have the capital structure to execute. In the fourth quarter of 2025, revenue was $35.8 million, down from $50.6 million in 3Q '25, primarily driven by lower Bitcoin production. Importantly, HPC lease revenue increased to $9.7 million in Q4, up 35% from $7.2 million in Q3. For the full year, revenue increased 20% to $168.5 million from $140.1 million in 2024, with digital asset revenue of $151.6 million and HPC lease revenue of $16.9 million. We commenced HPC leasing in July 2025 and have energized 18 megawatts of critical IT capacity as of year-end. As additional buildings come online, the revenue mix will continue shifting towards stable contracted HPC revenue. The cost of revenue, exclusive of depreciation, increased 10% from $17.1 million in Q3 to $18.9 million in Q4. Demand response proceeds recorded as a reduction in the cost of revenue decreased due to technical difficulties in Q3. For the full year, demand response proceeds also increased year-over-year from $8.6 million in 2024 to $17.7 million in 2025. Operating expenses increased as we scaled the platform to support HPC deployment. Quarter-over-quarter operating expenses rose to $8.8 million from $4.5 million. Full year operating expenses increased to $19.7 million in 2025 from $7.6 million in 2024, reflecting staffing and operational readiness. For context, TeraWulf finished 2024 with under 100 full-time employees and will exit 2026 with close to 300 full-time employees. Let me address the HPC leasing segment profitability as presented in Note 19 of our 10-K. The as-reported annual segment profit margin is approximately 42% versus our long-term guidance of approximately 85%. That difference is driven by three factors: first, $1.2 million of tenant fit-out revenue and associated costs during 2025. PFO carries a modest margin as provided for under the HPC leasing. Second, $4.1 million of development and pre-revenue operating costs; and third, partial period revenue contribution as buildings ramp. Adjusting for those factors yields approximately 77% segment profit margin in 2025, which is consistent with ramp expectations and converging towards our 85% steady-state margin guidance as utilization stabilizes. SG&A expense also increased as we scaled the platform to support HPC deployment. Quarter-over-quarter, SG&A expense rose to $66.6 million from $16.7 million. Full year SG&A expense for 2025 totaled $147.8 million from $70.6 million in 2024. After adjusting for stock-based compensation, SG&A increased from $39.7 million in 2024 to $94.5 million in 2025. This increase is primarily attributable to an incremental $47.5 million of new hires, strategic growth performance, and milestone-based employee compensation in 2025, reflecting the notable scale of execution achieved during calendar 2025. Adjusting for this item results in total SG&A of approximately $47 million in 2025, in line with our prior guidance of $50 million to $55 million. Depreciation increased to $88.6 million in 2025 from $59.8 million in 2024, reflecting infrastructure placed into service and accelerated depreciation of $19.6 million associated with certain mining assets transitioning to HPC use. Interest expense in Q4 was $62.4 million compared to $9.8 million in Q3, and we recognized interest income of $31.5 million in Q4 compared to $4.1 million in Q3. Annual interest expense for 2025 and 2024 was $80.2 million and $19.8 million, and we recognized interest income of $39 million and $3.9 million, respectively. The increases in net interest expense were expected following our capital raises at TeraWulf and WULF Compute in the second half of 2025. Actual cash interest paid during Q4 and the calendar year 2025 was $6.9 million and $13.9 million, respectively. The change in fair value of warrant and derivative liabilities in 2025 was a loss of $429.8 million, primarily related to the Google warrant. This is a non-cash loss and therefore does not affect our liquidity. Equity and net loss of investee net of taxes for 2025 was $4.1 million, which represents TeraWulf's 50.1% share of the net loss of the Abernathy Joint Venture, which was formed in October 2025 and has not yet commenced operations. Our GAAP net loss in 2025 was $661.4 million compared to a net loss of $72.4 million in 2024, with the increase primarily driven by non-cash fair value adjustments related to the Google warrant and non-cash depreciation. Our non-GAAP adjusted EBITDA in 2025 was negative $23.1 million, down from positive $60.4 million in 2024. As a reminder, these results are inclusive of significant increases in SG&A and operating expenses over the past 12 months as we invest heavily in our HPC business. Turning to the balance sheet and liquidity. As of December 31, 2025, cash and restricted cash totaled $3.7 billion. Total assets amounted to $6.6 billion, with total liabilities of $6.4 billion. Regarding liquidity, as detailed in our fiscal year-end 2025 investor presentation on the slide titled Capital Structure, as of January 31, 2026, the HoldCo parent entity had approximately $500 million of available cash or approximately $300 million pro forma for the Kentucky acquisition announced on February 2. Regarding project-level capital positions and construction progress, both WULF Compute and Abernathy are fully funded through substantial completion with long-term fixed-rate financing, eliminating construction funding uncertainty and reducing reliance on near-term capital markets access. Importantly, we do not anticipate the need for additional equity to fund our currently contracted development. As of January 31, 2026, WULF Compute had approximately $3 billion of gross cash or $2.6 billion net of debt service reserve and interest-earning construction accounts, with $850 million of CapEx spend complete and $2.38 billion remaining. That leaves approximately $200 million of cash cushion, which is incremental to the substantial contingency embedded in the financing structure. As Nazar noted, schedule adjustments resulted in approximately $16 million less projected revenue in years 2025 through 2026. However, design optimization has increased capacity from 162 to 168 critical megawatts across CB4 and CB5, generating approximately $200 million of incremental revenue over the initial lease term. The net effect improves projected cash flows and reduces expected debt and maturity by approximately $45 million versus prior projections. Finally, with regard to the Abernathy JV, as of January 31, 2026, the JV had approximately $1.5 billion of gross cash or $1.2 billion net of debt service reserve, interest during construction, letter of credit, and HoldCo LockBox accounts, with $268 million of CapEx spend complete and $1.1 billion remaining. That leaves approximately $70 million of cash cushion at Flash Compute with a further $100 million liquidity reserve at the parent JV, supported by a $1.35 billion lump sum EPC contract. With respect to Kentucky, we have proposals in hand for secured loan facilities to fund pre-lease development to preserve HoldCo parent liquidity. Demand for near-term power remains strong, and we are targeting 480 megawatts online in the second half of 2027. We do not build on speculation. In summary, 2025 reflects a business transitioning from volatile Bitcoin mining revenue to stable contracted HPC revenue. Mining has strategically supported that transition. Contracted HPC revenue is ramping, liquidity and contingency are strong.

We are ready to take questions.

First, I just wanted to start with Kentucky. That site sounds like the reception has been strong. Could you give us a few more details on the site and what an ideal customer or lease would look like there?

Mike, this is Paul Prager. It's a fantastic site. This was the site of a former smelter. It's at a transmission super highway. Multiple utilities can service the property. What was most compelling about it was its central location and the immediate availability of power in scale. Demand for the site is extremely strong. Our data room is open. Every major hyperscaler and several large AI compute platforms are doing diligence. The conversations to date have been substantive, with some written term sheets already coming over the desk. Earlier this week, I was down in Kentucky. I met with Governor Beshear and state leadership. The alignment at the state and local level is clear, very constructive, very supportive of business. The importance of this project for the local community, particularly the public schools, which Kentuckians are very proud of, is massive. We held a community information session two nights ago. We had a job fair yesterday. We filled the auditorium and then some. We're extremely excited and confident in the long-term value of the asset. As Patrick has said from day one, we're all focused on the best financial credits to be our long-term customers. I think we've operated that way in the past, and that's the message going forward. I think you'll see a world-class credit as our next customer for what we're hoping to be a 10- to 15-year deal. I think we'll see that deal happen pretty soon. I don't want to give you a drop-dead date, but we're in very active and substantive discussions.

Great. And then secondly, the Maryland site seems like a lot of potential up to 1-plus gigawatts, but a little bit more complex. And kind of playing into what you guys have talked about, bring your own power, how does that play into some of TeraWulf's strengths?

Sure. Just to be clear, Chesapeake Data is about the power and load differentiation. It's a gig site with certainly a gig data center load capability, 500 megawatts of battery storage capability. It's a former coal generation campus. It's in the Northern Virginia corridor, which means prices are exceedingly competitive. It's designed to be a net contributor to the Maryland grid. It's very well supported by Governor Moore and MDE. Again, both Maryland and Kentucky, by the way, have very sophisticated brownfield programs, which make it really easy for a company like us who has been around the block and owned and operated coal-fired power plants shutting down to do everything the right way. These two states have very progressive programs on how we do that at a commercial feasibility level. Listen, the market is moving to bring your own generation. We said that a year ago. In December, Alphabet bought Intersect for almost $5 billion. In January, Microsoft raised dedicated generation as part of their five-point infrastructure strategy. President Trump in January floated the notion that PJM emergency auctions needed to incentivize new generation. New generation projects would go to the top of the queue for interconnect. And then just the other night, in the State of the Union, the President stated data centers need to build and fund their own generation. So that's where the world is going. We have a real and growing power shortfall. Morgan Stanley says potentially 47 gigawatts from 2025 to '28. The hyperscalers are openly stating that power is the binding constraint. Look at anything recent public commentary by Colette Kress, NVIDIA's CFO; Sundar, Alphabet's CEO; certainly Jens Huang, NVIDIA's CEO. I mean, it's all about we need power. So delivering generation alongside that load solves the problem. We could bring incremental megawatts to the grid or the system, dispatchable generation using CCGTs, not just bridging power. And we have a history of partnering with grid operators to solve reliability and adequacy challenges. What's our core competency? We've been talking about this. This is the only team out there with over 25 years of developing, building, renovating, and rescuing generation. We have deep expertise in siting, interconnection, and generation development. And that's why we could go and take on a project like Morgantown and have the support that we can from both the local and state communities as we pursue this. We're really excited about Morgantown. It's a big job, but it fits into our schedule. And again, we've told the world we're looking to deliver 250 to 500 incremental megawatts of data center load every year, and this fits right in.

Do you think the pricing and terms and conditions in Kentucky can be materially better than what you've done in the past? And the construction schedule seems pretty ambitious. Do you have the labor and all the equipment you think you'll need, and I guess, the permits to get it done?

In terms of the labor, yes, Kentucky is a fantastic site. You have not only the construction expertise and the trades expertise, but you have people that really want to work. We brought on for that job Fluor, which I consider to be a world-class contracting party. We've dealt with them in the power space for a long time. They have relationships at every level in our C-suite. We are already ahead of the curve in procurement. We have a proven design that our hyperscaler customers really like. And we think because it's immediately available power, we know that because we've got very robust conversations going on right now with folks that want to come in and be our customers at competitive pricing. Nazar?

Just to add to what Paul said, as Paul said, we brought in Fluor on the EPC side. So we're already hitting the ground running with the overall development of the project. The timelines that you see for the second half of '27 are reflective of ongoing and meaningful discussions already with the EPC. So we feel pretty good about the overall timeline. The other big component is gathering the labor that's going to support the project as well. There have been efforts already underway for both the mechanical, electrical, and civil scopes to get the requisite labor to support the project as well. So that date is informed by quite a bit of discussion that we've had with Fluor. In terms of the overall economics, we've said this on prior calls; we really think about these projects on an unlevered yield basis. And so if you look at where we've been historically, we think we'll be in that ZIP code or maybe better over time. But again, as we think about the overall economics of the project, we're always focused on what's the unlevered yield that we're developing at and pushing to kind of maintain and continue to increase that as well.

Congratulations on the progress here. I wanted to ask on a couple of questions on the power side. First, I noticed that the PUE across all these sites is right in line with the initial deals at 1.25. And you mentioned in the slides that that's best-in-class. My understanding was there were certain aspects of Lake Mariner and Cayuga that drove that 1.25, but maybe I'm mistaken. It seems like it's standard for TeraWulf to operate at that incredible efficiency. Can you just give us a little color there on why you're able to sort of run circles around your competitors when it comes to a PUE standpoint?

Sure, Chris, it's Nazar here. There are a couple of factors in that. As you noted, one is just the geographic location. Being more in the northern half of the country versus the southern half provides us benefits from an ambient conditions perspective with respect to the design and our ability to meet that peak PUE. You'll see the Abernathy site is at 1.4 PUE, which is reflective of its geographic location. In addition to that, we have invested heavily in our cooling solutions as well; thus, we give ourselves extra room in the design that we have on cooling to maintain that lower PUE. So in general, as we think about where we are for generally kind of in the northern half of the country, where we have these off-seasons during the winter and spring where the summer isn't sustained heat, we think we can maintain that 1.25 PUE.

Okay. Was there also a redundancy aspect to it? Are you still not using big diesel generators at these sites?

That's correct. And again, that gets back to these brownfield sites. Typically, you're seeing us play at brownfield industrial sites. The way to think about it is that when the smelter was there, there was significant investment to design and build a smelter at that site. The last thing that Century wanted was a single point of failure on power coming into the site, no different than a data center. So there are five different lines coming into that site, which provides a considerable amount of redundancy. When we look at sites like that, we see that there are multiple independent pathways for the delivery of power, which obviates the need for on-site backup generation. Morgantown, similar. It used to be a former coal-fired power plant; that 1.5 gigawatt coal-fired power plant did not want a single point of failure in power delivery. We're utilizing that same system now to bring power back in. The big benefit with these industrial former brownfield sites is that they usually have that built-in redundancy that data centers want as well.

I just wanted to thank you for that question regarding PUE because we put a page in the deck called critical IT capacity, the metric that matters. What we're going to be trying to do is really ensure that the street and our retail shareholders understand gross megawatts measure scale, but it's critical IT megawatts that measure monetized execution. We think of TeraWulf as an execution story. We're really going to be reporting more in the context of critical IT as opposed to just gross megawatt capacity. We think it's far more indicative of our performance in the marketplace.

Great. One more quick question is the 0.5 gig of battery power sort of caught my eye. Can you just give us a little color on why and what that means?

As we're adding these large loads, if you see in the statements we've made, we want the site to overall be a net supplier of energy back to the grid and to the state. So having that battery storage allows the load at the site to operate in a way that really mitigates peak demand. It acts as a critical peak demand shaver, which we believe makes the loads that are there assets back to the grid rather than burdens. So we think the right composition is about 0.5 megawatts of storage per megawatt of load. That's why you see in each of the phases at Morgantown, for every megawatt of load, there's about 0.5 megawatt of battery storage.

Guys, I want to congratulate you for the progress across all fronts. Maybe just following up on that last one, you mentioned the battery storage, and there just appears to be some different moving parts at Chesapeake. Can you just give us a sense of how CapEx could differ from Mariner and what you're doing to really de-risk that development?

When we think about the composition of Morgantown, there are a few different legs here versus what we're doing at Lake Mariner or other sites. On the data center side, we're in that $8 million to $10 million per megawatt range for CapEx, which is consistent with what we've done at Lake Mariner and Abernathy. As we look to contract Kentucky, we're in that same range. In Morgantown, we're adding two incremental legs: one is power generation and the second is battery storage. On the power generation side, we'll be somewhere in the $2 million to $3 million per megawatt range for the fully delivered power plant. Part of that cost will be time, part will be the type of turbines we are deploying, and we're working on several different alternatives for that site. But that will be around $2 million to $3 million per megawatt for that capacity. There is about another million or so on the critical IT load. So when we think about what we're offering back to our customer, it's the data center that they're already paying for in other deals that we have, that $8 million to $10 million per megawatt range. In addition to that, now it's the components on the generation side, meaning $2 million to $3 million per megawatt on dispatchable gas-fired generation, then another $1 million or so on battery storage. So all in, we think it will be around $13 million to $14 million per megawatt on a fully loaded basis. Included in that, we will be seeking to charge that full cost back in the capacity payment back to the underlying customer. They'll effectively now be long the value of the generation as well as long the capacity in the data center.

Nazar, I really appreciate all that detail. I'm sure others will have follow-ups. I just wanted to squeeze one in on the regulatory side. I think you still need approvals from FERC at Morgantown. Can you just walk us through what the timing looks like there and how we should think about what you ultimately need to get across the finish line from a third-party perspective?

It's a pretty pro forma approval process that we've done countless times. It happens anytime you transfer an existing power facility to another party. They typically look for things like whether you're a monopoly in the area. We are not. We consider this to be a routine process. We would expect FERC approval within three to six months.

Could you update us on any remaining zoning requirements or state and local approvals needed to move forward with the Hawesville data center campus? That's one. And then, separately, while Cayuga has already received zoning approval, are there any additional permitting or regulatory steps still outstanding for that site?

We'll go backwards. I'll start with Cayuga. We had our first informal session with the planning board a couple of nights ago; it went very well. They were engaged and asked good questions. This process will take another month or so to finalize our application. They will then meet to review it. We will come to an agreement on what that permit should look like, solving for certain conditions like whether you are drawing water from the lake, what the noise levels will be during operation, and what we are doing in terms of beautification. These are all reasonable questions. It's what we do. It's what any company in redeveloping a former power plant or industrial complex does. The Cayuga process, as you know, first had to go through zoning board approval to ensure that it was consistent with the definition required to achieve a permit, and it won that day. With respect to Kentucky, Nazar?

On Kentucky, from a permit perspective, we have to obtain the requisite building permits. That being said, we had a town hall this week in Hawesville. We had a workforce session for potential employees. The entire judge executive, the economic development director in Hawesville, are fully on board and very eager to get us going. As Paul mentioned earlier, it's an extremely supportive environment. While we don't have the permits in hand, everyone is fully aware of what we're doing. We have briefed them on the scope, size, and scale of the buildings we're bringing. Importantly, as part of what we're doing at Hawesville, we're also decommissioning and taking down the old aluminum smelter. There's a big cleanup happening at the site as well, which the town is very eager to see happen. We have previewed several components of this project, and we're hopeful it will be a smooth process to get approvals.

Just two more things. One is, with respect to both of these projects, there's not just one permit you get. There's a principal notion of a permit for a facility. But throughout the process, you have a number of specific permits, like demolition permits or building permits for particular structures that you need to obtain at the local level. The second thing I wanted to mention about Cayuga is interesting; as the state has become a popular place for people, and they look forward to doing more data centers there, they are starting to ask, 'What about power?' The beautiful thing about the Cayuga site is that it was a former power site. It can be again. We have the ability, both in terms of the landlord's huge site, which is 400 acres, to bring in our own power generation if that was ever something important to state or local constituents. So Cayuga is a much better site than just any other one in terms of that part of the world because of its ability to house a power plant on it if it's needed.

First one, if I may. Could you characterize the competitive process at Hawesville, maybe comparing that to Lake Mariner? Are we talking about one entity taking the 384, or is it going to be multiple entities? And then I've got a follow-up.

I'll start by giving you context. In the environment that we were contracting for Lake Mariner, I think we were very confident we had a really special site. But I think the hyperscalers were on a growth curve in terms of education and figuring out design. When we started Lake Mariner, I think people were not as solid in their design concepts or terms of how they would fill out their customers' ledger. Now we are in a very different environment. We're in an environment where hyperscalers, for example, Meta, are very aggressive. Google's got a great program. Amazon's got a great program. Microsoft is back into the market. They're extremely competitive, together with several neos that are seeking larger and larger facilities. We find ourselves in a much more competitive environment now where customers have more definition of what they need. With that, the beauty of Kentucky is that its immediate availability has enabled a robust dialogue for us in filling out our customers' needs.

The only thing to add, Darren, is we're targeting one, maybe two customers for the entire site.

That's helpful. And as a follow-up, you've raised your guidance to the 250 to 500 range. What are the one or two deterministic factors that drive that range? Is there any reason we should think there may be upside to that range? I'm trying to get an understanding of what that context really means.

The 250 to 500 is a very calibrated range that we're giving. It factors in operational and deployment capabilities. This requires gathering hundreds of people across multiple trades. It's also about procurement process around equipment and ensuring that we're not the last buyer, but we have room between what we need and available market capacity. Lastly, financing is what the company can bear from an accretive perspective. Thus, the 250 to 500 megawatts per year represents $2.5 billion to $5 billion of incremental capital we are guiding the market to spend annually for the next few years. We remain very confident that each year we will continue to sign up another 250 to 500 megawatts of critical IT load with customers and deliver that 12 to 18 months hence.

I would add a few points. First, we selected Fluor to be our contractor in Kentucky because we believe they're the most skilled contractors in the country. They're particularly good on front-end projects, ensuring that execution goes flawlessly. Choosing Fluor was about scalable growth so we could work with them on additional projects nationally. Secondly, we are all about execution. We could talk about our pipeline as much as you want, but if we don't deliver for our customers, we are out of business. Therefore, we need to be conservative and focused. We're still a nascent business. We have seen improvements in facility design just between CB2A, CB2B, CB3, and now in CB4 and CB5. As we learn and grow alongside our customers and our leading contractor in Kentucky, we will enhance execution capabilities. If there's an opportunity to grow beyond what we've said, we will, but we must stay focused.

Paul, you have multiple attractive sites that you could theoretically lease out now. Is there a reason why customers would want the Kentucky site over Lake Cayuga or the mega campus in Maryland? Has the order of preference for sites changed?

The demand is so significant that any party would likely want to take everything off the table at any one time. However, it's about time to power. That’s why Kentucky has become so crucial to the customers we are discussing with. The ability to achieve that scale in a short period differentiates it from others. One thing we've pushed for is regional diversity, and hyperscaler customers are starting to recognize that as well. They've experienced what happens when too many are in one place, and they're concerned about security and grid capacity. Therefore, having regional diversity is compelling for customers along with the immediacy of power availability. I think we have opportunities to expand our footprint both on the generation side and customer side in Kentucky. However, I have to remain focused on executing our projects and delivering facilities to customers. We're already issuing limited notices to proceed just because our customers want to get onto the property. Our focus is strongly on execution.

For Maryland and Kentucky phase build-outs, what is the target critical IT megawatt per building or hall you're envisioning today? Should we think about repeatable modules similar to Fluidstack? What drives that sizing decision?

In the Fluidstack context, the base design we worked on together results in about 168 megawatts of critical IT load being the benchmark. Typically, that gives you over 200 megawatts of gross capacity. As we have discussions with various customers, we usually look at that range of 200 megawatts gross, with around 160 megawatts net critical IT megawatts as a building block for deployment. For Kentucky, that would provide a capacity around 380 megawatts of net.

How are you viewing some of the pushback at the state level to data center builds? Are the media articles overblown or is there some risk there?

Thoughtful and careful engagement is key. How you bring load on is critical. From the beginning, we've indicated that there’s beneficial Bitcoin mining and a detrimental form. We've established a strong relationship with the energy supplier in Kentucky, developing an understanding of loading challenges. We've aligned our interests with them to provide visibility on where we stand. This allows us to ensure our loads are seen as assets, not burdens to the grid. We remain committed to thoughtful engagement around the challenges presented in media. What we focus on is being constructive and seeing our loads as assets. We hope Kentucky will serve as a model for others, showing how data centers can integrate back to energy grids. We are in the early stages for both Kentucky and Maryland projects. We currently have a team of around half a dozen people in Kentucky. Each of our sites aims for 100 to 120 people once fully operational. Hence by this time next year, we anticipate staffing 100 people in Kentucky, with a few in Maryland at the moment.

We also have added corporate-level staff to manage the larger portfolio and ensure effective management from procurement, legal, accounting, IT, and everything essential at the corporate level.

When do you think Phase 1 of Morgantown might be able to be turned on?

We're currently working through that. Morgantown involves several components: load, data center, gas generation, and battery storage, along with remediation. We've been scoping that with MDE and expect to refine the timing over the next couple of quarters.

Regarding the State of Maryland, they've experienced challenges by shutting down many of their great generation sources; however, with policy changes from the governor, we have received commitments for expedited permitting. This suggests to me that it won't be business as usual. Their reception has been very positive, and we're being encouraged to generate jobs at both the county and state level.

We have reached the end of the question-and-answer session, and this concludes today's conference. You may disconnect your lines at this time, and we thank you for your participation.