管理層發言
Greetings, and welcome to the TeraWulf Second Quarter 2026 Earnings Conference Call. Please note that this call is being recorded. I will now turn the call over to John Larkin, Senior Vice President and Director of Investor Relations with TeraWulf. Please go ahead.
Good morning, and welcome to TeraWulf's Second Quarter 2026 Earnings Call. Joining me today are Chairman and CEO, Paul Prager; our CTO, Nazar Khan; and our 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 will turn the call over to our Chairman and CEO, Paul Prager.
Thanks, John, and good morning, everyone. The second quarter was defined by execution and expansion. At Lake Mariner, we converted additional contracted capacity into operating infrastructure and recurring lease revenue. At the same time, we expanded the platform through the acquisition of Muskie. Following quarter end, we executed a 401-megawatt lease with Anthropic at the Justified Data campus and entered into an agreement to monetize our interest in the Abernathy joint venture. Taken together, these developments demonstrate the model we've been building, which is to secure power-advantaged infrastructure, contract with high-quality customers, deliver capacity in phases and selectively recycle capital into the next generation of growth. Our number one priority remains execution. You see that most clearly at Lake Mariner. CB-3 was fully delivered and generating lease revenue in early July, bringing total revenue-generating critical IT capacity at the campus to 102 megawatts. That delivery also satisfied the applicable conditions for $600 million of Google's credit support for FluidStack's obligations to become effective. CB-3, therefore, represents more than just the construction milestone. It is another building delivered, another contracted revenue stream online and a significant portion of the credit support behind the project now effective. Following quarter end, we also amended certain FluidStack leases. Those amendments increased contracted capacity, added rent associated with tenant-requested scope changes and established updated delivery schedules on a data hall-by-data hall basis. Nazar will cover construction and commissioning in greater detail, but the key milestones are straightforward. At CB-4, we remain on track to begin energizing the first data hall in late September. At CB-5, we expect to begin energizing the first data hall in very early January. The first data hall at CB-4 is already in commissioning, and we continue to work closely with FluidStack, Google and the broader project team to align infrastructure readiness with hardware deployment. The important point here is that CB-3 is online and generating revenue. CB-4 is in commissioning and CB-5 is advancing against the updated customer-aligned schedule. That is execution. And it is increasingly visible in our financial profile with high-power compute leasing representing the majority of our revenue during the quarter. While Lake Mariner continues to deliver, the second major theme is expansion. Nowhere is this more evident than in Kentucky. Following quarter end, we executed a long-term lease with Anthropic for approximately 401 megawatts of critical IT capacity at our Justified Data campus in Hawesville. The agreement expands our relationship with Anthropic and represents approximately $19 billion of contracted revenue over the initial 20-year lease term. The economics are highly attractive for our shareholders and reflect the value of controlling large-scale power-secured infrastructure in a market where capacity is increasingly scarce. But the significance of the agreement goes well beyond its size. Lake Mariner demonstrates that we can take a legacy industrial power site and convert it into a large-scale operating high-power compute campus. Justified demonstrates that we can repeat the model in a new region. We secured the site, control the power infrastructure and converted that position into long-duration contracts with one of the leading companies in artificial intelligence. That is our model: control the infrastructure, contract capacity, finance it against long-duration revenue and deliver it in phases. We also expanded our Kentucky platform through the acquisition of the Muskie Data campus in Eastern Kentucky. Muskie is a gigawatt-scale development site and a prime example of the utility-partnership path to power that we discussed on our last earnings call. The campus is located within an established industrial park and is being developed in partnership with investment-grade Kentucky Power, an AEP company. Our electric service arrangements were entered into under a data center tariff approved by the Kentucky Public Service Commission, which provides for 1 gigawatt of electric service. Kentucky Power is expected to construct a new 345-kilovolt substation connected to AEP's existing 765 kV transmission network with initial electric service expected in the fourth quarter of 2028. This is not simply land with a queue position. It is a utility-supported development pathway with contracted electric service, defined infrastructure obligations and a state-approved framework for large-scale data center development. The market too often treats a queue position or inclusion in a batch study as equivalent to available power. It is not. The relevant and important questions are, when can the power be actually delivered, under what contract or commercial framework and with what degree of infrastructure certainty. Muskie provides considerably greater visibility for each of these points. Muskie also builds on the substantial momentum and relationships we've developed in Kentucky. The Commonwealth is increasingly attractive to prospective tenants because of its power infrastructure, business environment and the constructive engagement we have seen from state, utility and local stakeholders. Justified provides our near-term contracted delivery opportunity in Kentucky. Muskie provides the next gigawatt-scale platform in our pipeline, and we are actively advancing commercialization discussions for the site. Given its near-term power availability, we are increasingly optimistic about the potential to expand the Muskie campus to as much as 2 gigawatts and accelerate portions of the current development timeline. In addition to Muskie, our pipeline includes additional expansion opportunities at Lake Mariner and Lake Hawkeye in New York, Chesapeake in Maryland, and there are many other sites we are actively evaluating. Each is at a different stage, but collectively, they provide multiple paths to power, customer contracting and phased delivery rather than dependence on a single market for an interconnection process. As we expand the platform, we are also being disciplined about where we spend our time and capital. That's what drove our decision on Abernathy. Following quarter end, we entered into an agreement to sell our entire interest in the Abernathy joint venture for approximately $530 million. Abernathy is a great project. But at this point in TeraWulf's development, it is simply not the right project or the best project for us to continue to own. Our strategy is increasingly focused on large-scale opportunities where we control the site, control the power infrastructure, the development process and the customer relationship. All that drives long-term economics. The Abernathy transaction allows us to focus our management resources and capital on those opportunities. It also demonstrates our ability to create value through development and selectively recycle capital into larger scale projects that we directly control. The Anthropic lease demonstrates our ability to create long-duration contracted value. The Abernathy transaction demonstrated our ability to realize value and redeploy that capital into the next generation of growth. We've also recently cleared an important milestone at Chesapeake. On July 29, the Federal Energy Regulatory Commission authorized our proposed acquisition of the Morgantown site. That approval clears a significant regulatory condition towards closing. The site includes approximately 210 megawatts of existing grid-connected generation, substantial electrical infrastructure and meaningful long-term expansion potential in one of the most power-constrained regions in the country. Subject to the remaining closing conditions and required consents, Chesapeake offers the potential to develop an integrated generation, storage and data center campus capable of supporting up to 1 gigawatt of data center capacity while serving large-scale compute demand in the most competitive region and supporting regional grid reliability. We have developed our pipeline this way deliberately. Regional diversity gives us access to different power markets and utility partners, but it also provides greater operational and security resilience. We do not want the platform dependent on one grid, one regulatory regime or one source of generation. Our portfolio includes utility-supported grid-connected campuses, sites with existing generation infrastructure and locations capable of integrating generation and storage or supporting additional generation on the broader grid. We view behind-the-meter power primarily as a bridge to utility-supported grid-connected campuses. Over time, we believe the most reliable, resilient and economically sustainable power solutions will be those interconnected with a larger utility system. That concept is becoming increasingly important. The constraint on AI infrastructure is not demand. It is power, transmission, interconnection and the ability to bring new infrastructure online responsibly. We have consistently said that data center infrastructure should be an asset to the grid, not a burden to it. And that is how we view Governor Hochul's recent executive order in New York. We do not believe the executive order will disrupt our development timing at either Lake Mariner or Lake Hawkeye. The WULF Compute build-out at Lake Mariner is already permitted, and Lake Hawkeye is in the early stage of development and has not been impacted. Importantly, the governor has described the order as an effort to establish a framework, not a permanent ban on future data center development. We welcome such a framework. Responsible development at this scale necessarily requires careful consideration of environmental impact, grid reliability, cost allocation and community priorities. We do not view those considerations as a binary choice between development and responsibility. Instead, we work constructively with regulators, utilities and local communities to address legitimate concerns, operate responsibly and be an active and valuable long-term member and neighbor within the communities in which we operate. A clear framework should establish the rules of the road around reliability, cost allocation, resource use, generation, storage and community impact so that credible projects can move forward responsibly. We believe increasingly formal requirements for large load development are inevitable across all major power markets, not only in New York. The better approach is to prepare for reasonable standards now rather than assume projects can avoid them indefinitely. That means developing sites that can demonstrate credible and redundant power delivery, assume appropriate cost responsibility and where required, support additional nameplate generation capacity or otherwise help bring incremental generation and grid resources online. Many of the principles outlined by the governor are entirely consistent with how TeraWulf already approaches development. We understand power markets, generation, transmission and believe clear standards will ultimately favor experienced, well-capitalized developers with credible infrastructure solutions and the demonstrated ability to execute. Increasingly, data center demand is also serving as a catalyst for long overdue investment in the nation's transmission and generation infrastructure. That is where TeraWulf is positioned. So when you step back, the progress is clear. We've delivered CB-3 and brought another contracted revenue stream online. CB-4 and CB-5 are advancing against updated customer-aligned schedules. We've expanded our Kentucky platform through the Anthropic lease at Justified and the acquisition of Muskie. We've agreed to monetize Abernathy so we can focus our capital and management attention on larger scale opportunities that we directly control. And we've cleared an important regulatory milestone at Chesapeake. Together, these developments reinforce both the consistency and the depth of our strategy, build a regionally diversified pipeline with credible paths to power, contract with high-quality customers, deliver capacity in phases and recycle capital where doing so creates long-term value for our shareholders. Based on the strength of the platform and the customer engagement we continue to see, we reaffirm our target of contracting an incremental 250 to 500 megawatts of critical IT capacity annually. We will pursue that growth with discipline and remain focused on power control, customer quality, execution certainty and shareholder returns. We have the sites, the capital, the people and the customer demand. Now it is about delivering. With that, I'll turn it over to Nazar to discuss construction, commissioning and the updated Lake Mariner delivery schedule.
Thank you, Paul. I'll focus my remarks on construction, commissioning and delivery at Lake Mariner. As Paul noted, CB-3 is fully online and generating lease revenue. With that building complete, our focus is now on executing against the updated delivery schedules for CB-4 and CB-5, which were developed in close coordination with our tenant. From an execution standpoint, the two most significant variables have been electrical labor availability and ongoing design optimization for our customer. Electrical labor remains highly constrained across the data center industry. As the design and electrical scopes became more fully defined, we added a second electrical contractor and scaled the workforce to support approximately 1,000 electricians at peak. That additional capacity is important to maintaining the targeted delivery schedule. At the same time, working closely with FluidStack, we continue to optimize the electrical, cooling and operational requirements while design, procurement and construction were already underway. Although this added complexity to execution, it enabled our customers to establish a deployment standard tailored to the latest generation hardware. With those resources and updated requirements now incorporated into the plan, we have greater visibility into the remaining work and remain confident in the revised delivery schedules. Let me now turn to commissioning and explain where CB-4 stands today. Commissioning generally progresses through six levels from Level 0 through Level 5. Level 0 covers design and planning, while Level 1 involves factory acceptance testing of major equipment before it is shipped to the site. Our current focus is on Levels 2 through 4. Level 2, or installation verification, confirms the major electrical, mechanical and cooling equipment has been properly installed, connected and configured. This is also when the tenant begins installing and tuning its cooling distribution equipment. Level 3, for start-up and prefunctional testing, is when individual systems are powered and tested under operating conditions. The customer also begins bringing server racks into the data hall and integrating them with the building's power, cooling and control systems. In practical terms, Level 3 marks the transition from construction into live systems commissioning. Level 4 functional performance testing is the contractual delivery milestone. Once testing is complete, the data hall is turned over to operations and begins generating revenue. The first data hall at CB-4 is currently in Level 2 commissioning. We expect to begin Level 3 in mid-August with the customer server rack scheduled to arrive shortly thereafter. We remain on track to reach Level 4 and begin generating lease revenue from the first data hall in late September. CB-5 is also progressing against the revised schedule with the first data hall expected to begin energizing in very early January. Across the project, we continue to have a highly constructive working relationship with FluidStack, Google and the broader project team. The revised schedules were developed collaboratively to align infrastructure readiness with customer hardware deployment. That coordination is critical on a project of this scale where construction, commissioning, equipment delivery and systems integration must all progress together. In summary, CB-3 is operating and generating revenue. CB-4 has been commissioning and remains on track for initial delivery in late September. CB-5 is advancing towards initial energization in very early January. We remain confident in the revised delivery schedules and focused on converting the remaining contracted capacity at Lake Mariner into operating infrastructure and recurring revenue. With that, I'll turn it over to Patrick to review the financial results for the second quarter.
Thank you, Nazar. As Paul outlined, the quarter demonstrated both sides of our capital model, creating long-duration contracted value and selectively recycling capital into larger scale opportunities that we directly control. Following quarter end, we entered into two transactions of significant financial importance. First, we executed a direct 20-year lease with Anthropic at the Justified Data campus, representing approximately $19 billion of contracted revenue over the initial term. Second, we entered into an agreement to sell our 50.1% interest in the Abernathy joint venture for approximately $530 million, representing a 20% internal rate of return on our initial investment. I'll focus my remarks today on our second quarter results, the financial impact of the FluidStack lease amendments, our updated Lake Mariner capital outlook and the liquidity available to fund our growth. Revenue for the second quarter was $44.8 million compared with $34.0 million in the first quarter, primarily reflecting additional HPC capacity coming online. HPC lease revenue increased 52% quarter-over-quarter to $31.9 million from $21.0 million and represented approximately 71% of total revenue. As Nazar described, the first data hall at CB-3 achieved ready-for-service status in late June and the second data hall followed in mid-July. As a result, the second quarter included only a partial revenue contribution from CB-3, while we entered the third quarter with 102 critical megawatts operating and generating lease revenue at Lake Mariner. Importantly, completion of CB-3 also satisfied the applicable conditions for $600 million of Google's credit support for FluidStack's lease obligations to become effective. This represents an important credit milestone and further strengthens the contracted revenue profile of the Lake Mariner build-out. The FluidStack lease amendments executed in early July also have several important financial implications. The revised commencement dates were mutually agreed with our tenant, and we remain on schedule with the revised timeline. TeraWulf will contribute approximately $150 million to address tenant fit-out costs incurred through June 30, 2026. In return, WULF Compute expects to receive more than $300 million of incremental lease revenue over the initial 10-year lease term. Together with the previously announced increase in contracted capacity from 162 to 168 critical megawatts at each of CB-4 and CB-5, the amendments are expected to generate more than $500 million of incremental lease revenue for WULF Compute over the initial lease terms. Turning to operating results. Cost of revenue, exclusive of depreciation, increased to $12.4 million from $2.4 million in the first quarter. The increase primarily reflects lower demand response proceeds, which are recorded as a reduction in cost of revenue and declined to $2.8 million in the second quarter from $14.1 million in the first quarter. Operating expenses increased to $23.4 million from $11.2 million as we continued scaling the platform ahead of additional HPC capacity entering service. The increase primarily included $5.8 million of additional site-level expenses, including security, labor and pre-revenue operating costs, $3.3 million of site preparation and demolition costs associated with future development and $2.2 million of minor equipment repair costs related to assets expected to be returned to service or sold. Reported HPC leasing segment profit margin was approximately 28% during the quarter compared with our long-term target of approximately 85%. The reported margin includes approximately $2.8 million of tenant fit-out revenue and associated costs, $6.8 million of pre-revenue operating costs at WULF Compute and $6.0 million of development costs across our portfolio of uncontracted development sites. Adjusting for these three items, HPC leasing segment profit margin was approximately 80%. We expect margins to progress toward our long-term target as additional contracted capacity enters service, pre-revenue operating costs decline and the contribution from mature HPC lease revenue increases. SG&A expense decreased slightly to $126.9 million from $127.8 million in the first quarter. Excluding stock-based compensation and charitable contributions, adjusted SG&A was $28.6 million compared with $26.3 million in the prior quarter. We continue to expect full year adjusted SG&A to remain within our previously disclosed guidance range of $75 million to $100 million. Depreciation decreased to $21.2 million from $28.5 million in the first quarter. The first quarter included $11.9 million of accelerated depreciation associated with Bitcoin mining assets whose useful lives were shortened as portions of the Lake Mariner campus transitioned primarily to HPC use compared with $2.6 million in Q2. Interest expense was $56.4 million compared with $67.1 million in the first quarter, and we recognized interest income of $28.9 million in Q2 compared to $29.4 million in Q1. Cash interest paid was $125.7 million compared with $5.3 million in the first quarter, reflecting the first semiannual interest payment on the WULF Compute senior secured notes in April. We recorded a $755.7 million noncash loss from the change in fair value of the Google warrants compared with a $216.3 million noncash loss in the first quarter. The change was primarily driven by the increase in TeraWulf stock price and had no impact on our liquidity. Equity in the net loss of the Abernathy joint venture was $11.1 million, generally consistent with the $11.5 million in the first quarter. GAAP net loss attributable to TeraWulf was $939.9 million compared with $427.6 million in the first quarter. The increase was primarily driven by the noncash fair value adjustment associated with the Google warrants. Non-GAAP adjusted EBITDA was negative $18.3 million compared with negative $4.1 million in the first quarter, reflecting continued pre-revenue operating and development costs incurred ahead of additional contracted HPC capacity entering service. Now turning to the balance sheet. Cash and restricted cash totaled approximately $3.0 billion as of June 30. At the parent level, we held approximately $1.2 billion of unrestricted cash at quarter end. Including the initial $250 million payment received under the Abernathy transaction in July, unrestricted parent liquidity increased to approximately $1.45 billion. We expect to receive an additional $150 million on or before December 31, 2026, and approximately $130 million on or before April 30, 2027, subject to the terms of the transaction. At WULF Compute, we had approximately $1.9 billion of gross cash at quarter end or approximately $1.5 billion after accounting for debt service reserves and interest-earning construction accounts. Approximately $2.3 billion of project capital expenditures have been completed with approximately $1.7 billion remaining. Approximately two-thirds of the remaining expenditures are committed, providing increased visibility into the remaining capital requirements. As a reminder, our original cost guidance for the WULF Compute financing and deployment was $8 million to $10 million per critical IT megawatt. We currently estimate total project costs at approximately $9.1 million per megawatt within that original guidance range and modestly above the approximately $8.6 million per megawatt financed in October 2025. The execution factors Nazar discussed, including electrical labor constraints and evolving customer equipment and operating requirements, are reflected in this updated estimate. Following the FluidStack lease amendments and TeraWulf's additional capital contributions, we expect the project's pro forma capitalization to be approximately 32% equity and 68% debt compared with approximately 26% equity and 74% debt at the time of financing. At the Justified Data campus, we had contributed approximately $353 million of equity as of June 30, including the $200 million site acquisition cost. These investments have funded the acquisition and early development work supporting the Anthropic lease and the planned project-level financing. Based on our current plans and assumptions, our existing liquidity and expected Abernathy proceeds provide the capacity to fund our remaining Lake Mariner commitments, planned equity investment at Justified, interim letter of credit requirements at Muskie, the proposed Chesapeake acquisition and other new sites we are actively pursuing without accessing the equity capital markets. Importantly, this plan also preserves a substantial liquidity reserve during the construction and delivery of these large, complex projects. We remain focused on matching capital deployment with contracted customer demand, maintaining financial flexibility and selectively recycling capital when doing so improves control, scale and long-term shareholder returns. Over time, we also continue to target investment-grade credit profiles at the TeraWulf parent and each of our finance subsidiaries. In summary, the second quarter reflects a financial profile increasingly driven by long-term contracted HPC revenue. We entered the third quarter with 102 critical megawatts operating at Lake Mariner, greater visibility into our remaining WULF Compute capital requirements and substantial liquidity to complete our contracted developments and fund the next phase of growth. With that, operator, we are ready to take questions.
分析師問答
Our first question comes from Nick Giles of B. Riley Securities.
It's good to see the service agreements with Kentucky Power. I was hoping you could speak to potential partnerships with utilities. What kind of framework should we have in mind? Should we expect that utilities are willing to share the economics of the project or kind of stop what we're seeing here with Kentucky Power?
Nick, it's Nazar here. So with the utilities, I think we're seeing a couple of different flavors come about. With the integrated utilities, Kentucky Power being an example and AEP being an example, I think they are independently solving for contracting for the load. So they've signed up a gigawatt of capacity with us under a letter of agreement and transmission arrangement, and then they go find the generation to support that. And so whether that comes from their existing resources, they contract for it, they recently acquired a power plant or they build more power plants, they're independently solving for that. And so what they're looking for us to do is to commit to both whatever the transmission build-out is for that project. And so that will be project dependent depending upon the location of that and where that sits within their grid and then a backstop for the energy. And so part of the credit that we posted to AEP was for the energy for the initial 500-megawatt allocation as well. And so again, the projects really depend upon the location, which will drive the cost of the transmission required to support it. And then it will be a market price on the cost of energy. When you put those two things together, that's where you see the credit that we posted to Kentucky Power and AEP.
That's super helpful. I appreciate it. Maybe switching gears. WULF has not done a deal direct with one of the high investment-grade hyperscalers. And so I was curious whether this is by design or what some of the key differences you see between the deals that you've done to date and one that would be with a high investment-grade hyperscaler?
Yes. Nick, it's Patrick Fleury. Thanks for the question. So as I think you've heard from us many times, we are highly focused on credit quality of the tenant; that's paramount to us. And so we have terrific tenants, both at Lake Mariner and at Hawesville. I think you saw us buy Hawesville in February and lease it less than six months later. That process was very robust and Anthropic was the winner of that. It's very strong economics to us, but there were a bunch of other competitors for that capacity in that process that we have moved over to Eastern Kentucky at Muskie. So I think I would just say to you, we are extremely happy with our existing partners. They are very strong credits. As you heard from us, we have the first slug of Google's backstop at Lake Mariner now effective. So I think it's a combination of credit quality of the counterparty and that's not just today, Nick, that's us thinking about the business model today, the competitiveness of that business model. And then the credit, not necessarily immediately today, but also three, four, five years down the road. These are 20-year leases, right? So you have to be front-footed and forward thinking. So as an example, as you've seen all the hyperscalers take on more debt, their balance sheets today are not going to be their balance sheets in five or six years. So I think there's an element of us solving for all of those things as we're looking at tenants. But I think stay tuned. And I think as we grow our site portfolio, naturally, the tenants will diversify.
Hi, it's Paul here. It's the same from an operation and execution perspective. We're just moving upstream to be closer to the ultimate customer. That enables precision, and that enables us to be a better service provider to our customers. So we get to deal direct now with our customer in Anthropic, and it just makes things that much more efficient in how we build out our facilities for them and how we operate the facilities for them.
The next question comes from Darren Aftahi of Lucid Capital Markets.
Just two, if I may. So can you maybe balance just the appetite for executing on more sites versus kind of digesting what you have? I guess, said another way, how kind of full do you feel you are? Are there limitations in kind of expanding the portfolio? And then maybe secondly, you talked a little bit about this, but you guys are operating in a few different geographies. Can you kind of talk about the calculus of what you look at from a sort of community and governmental risk perspective when you're thinking about entering into new or existing markets? And maybe if you had to characterize it, are there any markets where you feel more partial to versus others?
It's Nazar here. I'll take your questions in reverse order. First, with respect to sites, locations and geographies. We've expanded into Kentucky. We received clearance from FERC in Maryland. And so we're traditionally looking for jurisdictions and sites where there's been some activity previously. And so we've had a number of brownfield sites, which we think are well situated for further development. There's likely some significant electrical activity that occurred at that site, which bodes well for the future of the site with respect to repositioning that. And so we're looking for areas where some activity has occurred. The local communities understand what that means and can see the vision of what that transformation or repositioning of that asset could look like. And so fortunately for us in Kentucky, whether it's in Hawesville, west of Louisville, or in Muskie on the eastern side of the state, we have very strong local support in both of those communities with respect to bringing those projects forward. The other piece is that the team spends a tremendous amount of time engaging with the various local stakeholders and informing them of what we're up to. A lot of times, we find that it's a lack of information that is a challenge more than what the information is. And so the team spends quite a bit of time in each of those local jurisdictions, ensuring that we're properly conveying what we're seeking to do with respect to the projects. With respect to the overall portfolio, the guidance we've been giving is 250 to 500 megawatts per year of critical IT. We provide that guidance for a couple of different reasons. One is there's just an operational capacity. Each of these projects at the upper end of the range is still nearly $5 billion of total capital. We have to fund that with both equity and debt. We've got to hire a number of contractors, electricians at our site in Lake Mariner. We peaked at over 1,000 electricians at the site. And so as you scale up, the ability to procure the requisite labor gets more challenging. So that's one component. The second is I think we've seen in the market, especially here over the past couple of weeks and months, that these large-scale infrastructure projects may not work on a strictly linear scale. So a lot of times, people look and say, okay, what's happening in this quarter or that quarter and try to roll it out three or four years. Given the size of these projects, we think there will be more fits and starts. So trying to make sure that we pick our spots properly and have a good understanding is important. As Paul alluded to in his remarks, a position in the queue does not represent access to power. We really try to spend time ensuring that the projects we bring forward have the ability to deliver power and that we can capture it. Therefore, we can go to our underlying customers and tenants and give them very clear visibility with respect to their ability to start using that. So I think there are a number of things that we're looking at that could increase that. But for now, we remain focused on that 250 to 500 megawatts per year. Paul had some other thoughts as well to add.
This is Paul. I would only add that regional diversity is a foundational principle in the development of our portfolio. I don't want to be reliant on one grid. I don't want to be reliant on one regulatory perspective or horizon. I don't want to be reliant on one political mindset. I think our customers will become increasingly sensitive to the notion of security. I'm a Navy guy. I remember when we looked at submarine design versus what others were doing. Some designs put all major machinery in one part of the vessel; if you hit that part, the vessel was inoperable. American submarine design dispersed major machinery throughout the vessel to sustain operations even if one area was hit. I think security is critical. Our customers don't want to be vulnerable to something that happens in one grid, one region, or one fuel source. Labor sourcing, as Nazar mentioned, is important. Lastly, as you think about the move towards inference, folks want to move to the market as opposed to being in one place. Inference is more of a local consideration. That's why we're building the portfolio as we have, and we'll continue to focus on regional diversity as fundamental to what we do.
The next question comes from Michael Rollins of Citi.
So just building off of the last few questions. I'm curious, if you take all of that together, are you able to size what the incremental pipeline opportunity could be, maybe kind of thinking about it over the next decade. Currently, you've got 2.9 gigawatts of, call it, line of sight of lease capacity in pipeline. Just curious what the TAM is or how big that can get to? And then second, with certain regions seeing some new headwinds on timing, for example, recent developments in Texas or you mentioned earlier the implications of the New York Governor's actions. How is that affecting the conversation, demand, the interest to pre-lease with your portfolio?
This is Patrick Fleury. We are a long-time power team, as you know. I think I'll kick it off and then Paul can add. We always try to frame for investors the power demand and power challenges that the market is dealing with today. If you step back and think about different power regions, as an example, the state of California is on a stand-alone basis the fifth largest economy in the world. It has an 85-gigawatt installed grid. The Texas market has a roughly 95-gigawatt installed grid and average demand of around 65 to 67 gigawatts because it's a peak-year market and the system has to be designed for the highest draw day. The backlog in Texas is now over 400 gigawatts and batch 0 is roughly 60 gigawatts. So to frame that, you basically have to build another California in the Texas grid to meet all of the batch 0 projects. Combined cycles are roughly 500 megawatts each and take about three years on average to build. So you have to build hundreds of combined cycles. That frames how we look at and approach each market or utility partnership. It's power first: is the transmission grid set up to accommodate our load? Is the local generating system set up to accommodate our load? Can we get front-of-meter generation expansion at our sites as opposed to behind-the-meter, and work with local utility partners to increase our capacity over time, not by limiting our customers to behind-the-meter single or double points of failure, but connected to the grid that has thousands of points of failure while also not being a parasite on the grid and providing additional generation and even backup power during times of grid duress. That's generally how we approach each of these sites and markets from a power-first position.
It's Paul. That's why I said in my remarks it's nice to know you're somewhere in the queue, but we're all about available contracted power and that's where we focus. You asked about how big this could go. I don't have a specific number. Demand is extremely strong. At our Muskie site, we have the most active data room we've ever had and with world-class credits as potential customers. But I focus on execution. The simple response is that as long as we do what we have contracted right, the customers will continue to come, and that is why Anthropic is a multiple repeat customer. We have guided the markets to 250 to 500 megawatts because that's what we're built for at this time, and we have reaffirmed that. Can we do more? Maybe with the right EPC contractor, the right region where labor is available, and with the right mature regulatory framework and policy, absolutely. That's why Lake Mariner and Lake Hawkeye are worth more today as a result of some of the things the New York governor is discussing. But we're focused on execution. And as long as we can handle what we have on our plate, then we can expand in the next round. Regarding dialogue and what's changed, for us not a whole lot. We think about dialogue in two primary ways: with the community in the region and with customers. Because we come out of power, we've always developed power plants with sensitivity to the region, the local community and to the regulatory framework. Our customers are very sophisticated; they know the difference between a queue position and a contract with an investment-grade credit to supply gigawatts of power. That is where they are focused, and that's how they're prioritizing their time. I think that's one reason Muskie is seeing significant activity in the data room today. People recognize when that project could come online and that it's a very credible project. We're excited about where the market is today and look forward to continuing to participate in its growth.
The next question comes from Tim Horan of Oppenheimer.
A question on execution. Where are you finding more electricians and HVAC trades, etc.? Do you have costs locked down for them, especially on the Anthropic contract? Do you have the construction contracts locked down at this point? One reason I ask is that there is talk in the industry of very large builds — how does the industry find people to do this, and how do you compete against others looking for the same workers?
If it's okay, I'll start. One of the things we're very excited about is our relationship with Fluor on the Kentucky project. They're a top-quality, world-class EPC that we have a history with on the power side. They are good at bringing in top-quality talent and securing them in contracts for EPC duties on site. That's one driver. Second, experience matters. We grew up in the trades building power plants, so we have experience working with the right subcontractors and contractors. We like to lock down pricing so we can be comfortable. Early on, that was tougher because there wasn't a reference design. Both the market and contractors needed to figure out manhours and scope. We have that reference design now, which helps our subcontractors understand the scope and take more risk with us to provide fixed-price contracts. Nazar, did you want to add something?
This is Nazar. To add to what Paul said, this ties back to our guidance around 250 to 500 megawatts. If demand is there and we wanted to do significantly more, finding the workforce is a challenge. Working with Fluor and other partners, leveraging relationships in our jurisdictions and properly sizing opportunities are critical to execution and feed into discussions with tenants. For customers with similar designs or hardware, our ability to map the required labor hours tightens over time. Procurement of equipment and procurement of labor required to support these projects is critical. There is some bound we have to work within, which is part of the reason for our guidance range.
The next question comes from Mike Grondahl of Northland.
This is Logan on for Mike. First, can you provide additional insight into the demand and discussions for the now potential 2-gigawatt Muskie site relative to Hawesville? Second for Patrick, can you touch on how we should think about financing that development?
Logan, it's Nazar. The discussions are robust. We are engaged with all of the usual suspects regarding that capacity. They are interested not just in the initial tranche but the total site over time as well. We're working through iterations on hardware, design and deployment since the full site could be a multiyear process. Demand remains very strong. The '28 power is becoming more prominent; much of '27 capacity is either sold or close to being sold. There is heavy emphasis on '28 and strong demand. All the usual suspects are working with us on that front.
Regarding financing, you'll see us follow the model we opened up last October, which looks like project financing — a bond that's amortizing or a term loan. Our approach is intentionally different than much of the market. Our projects are not levered to extreme levels; they are conservatively levered with a healthy equity layer. As I mentioned earlier, the WULF Compute project moved from roughly 75% debt and 25% equity to a pro forma roughly 68% debt and 32% equity. For the debt markets to continue to take paper, you have to distinguish yourself. Our approach is conservative leverage and significant deleveraging and amortization during the period. You'll see us do that at Hawesville and Muskie as well. Importantly, as I mentioned, we have a flush balance sheet with sufficient liquidity to do almost everything in our near-term order book without returning to the equity capital markets. We'll be back in the debt capital markets to finance Hawesville, and I expect likely sometime in the first half of next year for Muskie.
The next question comes from Stephen Glagola of KBW.
Paul, can you provide more color on the three to five sites currently in active pursuit and final due diligence, particularly around power capacity under evaluation, geography and energization timing? And Patrick, any updates on securing credit support on the Anthropic lease at Hawesville and who that counterparty is?
This is Patrick. Regarding the Anthropic lease, I would reiterate that the Hawesville lease will be supported by an investment-grade credit. We went to the market and completed a large equity financing in April, and with the Abernathy JV sale, we have $1.5 billion of cash on the balance sheet today. That's plenty of runway and we don't need to rush to the market. We will prepare and go to the market when we're ready and when we think the markets are open.
On the pipeline and the three to five pursuit sites, the same principle of regional diversity drives those discussions. Our team, led by Kerri Langlais, looks at dozens of sites in real time. One of the pursuit sites is likely our first international data center effort. We have a history developing power plants outside the U.S., and we believe we can deliver data centers for customers in Northern Europe, though we are not there yet. We are also looking at the Midwest and Southeast. We talk to customers about their needs and where they want to be. For us it's all about available power and our ability to quickly contract that power. The development process is a large diligence effort that includes talking to the local community, political leadership and customers before advancing. Because of recent developments in places like Texas and New York, a lot of projects will fall away that are not credible or lack experience or cash to post collateral. That will create opportunities for well-positioned companies like TeraWulf.
The next question comes from Chris Brendler of Rosenblatt Securities.
Congrats on the results. On the FluidStack lease modifications, what drove the changes? Was it client-driven increased costs or modifications of design? And the increase in equity — was that a result of these changes, or was it intentional to increase equity to make the project more palatable for debt investors? Just looking for clarification.
Chris, it's Patrick. These projects are complex and designs evolve. We developed a reference design that has changed over time as clients get more experience running the hardware. There's a true partnership with the tenant and many design-change requests, referred to as tenant fit-out requests. At CB-3, we had about 43 tenant fit-out requests that each had cost and schedule impacts. The impacts at Lake Mariner were about $150 million that TeraWulf agreed to fund, which is being recovered over the 10-year lease term. In return, WULF Compute expects to receive over $300 million of incremental rent related to those items, essentially generating a mid-teens return on those costs. Combined with the move from 162 to 168 critical megawatts, these changes yield over $500 million of additional revenue over the initial lease term. We could have debt-financed those costs but chose not to in order to improve the project's cash flow and deleveraging profile. That approach is intentional; leverage is useful, but we want balance.
One quick follow-up: we've seen governors take action in New York and Texas recently. How do you feel about the political situation in West Virginia, Maryland and Kentucky at this point?
I don't think we've seen anything surprising. We've said all along that data centers are important but need to be done right. We expect more formal standards and frameworks. We develop projects proactively, so we can be in front of regulatory concerns and show regulators and local communities that we are responsible and credible. In Maryland, we are working on an industrial site that used to be a larger coal facility; we'll work with the state to clean it up, bring jobs and provide surplus generation to a power-constrained region, especially important to the Washington, D.C. corridor. Kentucky has been exceptional in its support; working with Governor Beshear and local communities has been very positive. West Virginia is an area we study and like; we think it wants to win business and create opportunities. We design power and data centers to be around for decades, so we try not to react to momentary political noise. Instead, we focus on what local communities and states need and how we can create win-win outcomes. We have not been materially deterred in any of our active or pipeline sites.
Our last question comes from Michael Funk of Bank of America.
You touched on constraints with labor, particularly electrical labor, and highlighted the estimate for cost per megawatt. What protections do you have in your contracts for rising costs? How are those costs impacting development yields or projected returns?
This is Nazar. When negotiating a lease, two large cost components are labor and equipment. On the equipment side, we usually have 12-month rolling forecasts with vendors and often have deposits down or have firmed up prices, so we have decent visibility. On the labor side, two levers are the number of labor hours required and the cost per hour. The cost per hour has been generally moving up as more projects move forward. It's difficult to pin down exactly for a construction cycle that's 12 to 15 months long. Working with Fluor and other vendors, we put parameters in place and award to subcontractors to lock in ranges for price. As we deliver more projects, our ability to estimate required hours improves. The yields we've been targeting are in the mid-teens range. With Anthropic and the FluidStack changes, we've maintained that range. Equipment pricing visibility at signing and labor controls with partners help manage cost risk, but per-hour labor cost is the area with the most uncertainty.
Let me address yield. The most important part is getting an adequate return on capital. The changes we've processed are generating over $500 million of revenue over the lease term, so we are recovering incremental costs. We are open with customers that we need a mid-teens return. That is necessary because if your weighted average cost of capital exceeds your yield on cost, there is no equity value. For example, if debt costs 6.5% and equity costs 25%, that results in a WACC near 10%. For a project to be accretive, the yield on cost must be greater than the WACC. I've seen other market deals where WACC exceeds the lease yield; that doesn't create value. We're creating tangible equity value for our shareholders on day one.
This concludes our question-and-answer session. The conference has now concluded. Thank you for attending today's presentation, and you may now disconnect.