테라울프(WULF) 2026년 3분기 실적 발표 콘퍼런스콜: HPC 매출 97% 증가

핵심 요약

• 3분기 매출은 2026년 2분기 4,480만 달러에서 7,620만 달러로 증가했다. HPC 임대 매출은 전 분기 대비 97% 증가한 6,300만 달러를 기록하며 전체 매출의 약 83%를 차지했다.
• 테라울프(TeraWulf) 귀속 GAAP 순이익은 7억 990만 달러로, 2분기 순손실 9억 3,990만 달러 대비 흑자 전환했다. 이러한 실적 개선은 주로 구글 워런트 관련 비현금성 공정가치 조정과 애버내시(Abernathy) 합작투자 지분 매각에 따른 이익이 반영된 결과다.
• 조정 EBITDA는 마이너스 1,280만 달러로 적자를 지속했으나, 2분기 마이너스 1,830만 달러에서 적자 폭을 줄였다. 보고된 HPC 임대 마진율은 17%였으며, 경영진이 산출한 조정 마진율은 장기 목표치인 약 85%에 근접한 약 83%를 기록했다.
• 레이크 매리너(Lake Mariner)는 3분기 중 102메가와트(MW) 규모의 핵심 IT 용량을 가동하며 임대 매출을 창출했다. 경영진은 4분기 말까지 CB-4를 통해 최대 168MW의 핵심 용량이 추가되고, 이어 2027년 1분기에는 CB-5에서 추가로 168MW가 더해질 것으로 예상하고 있다.
• 저스티파이드(Justified) 캠퍼스는 약 190억 달러 규모의 20년 만기 앤트로픽(Anthropic) 임대 계약과 구글의 신용 보강을 바탕으로 추진되고 있다. 테라울프는 핵심 IT 메가와트당 1,000만~1,200만 달러의 예상 비용으로 2027년 후반부터 2028년 2분기까지 단계적으로 서비스를 개시한다는 목표를 유지하고 있다.
• 9월 30일 기준 현금 및 사용제한현금은 총 약 24억 달러를 기록했다. 경영진은 현재의 계획과 가정을 바탕으로 할 때, 기존 유동성과 애버내시 지분 매각 예상 대금으로 공모 주식시장을 통한 자금 조달 없이도 잔여 계약상 개발 약정을 충당할 수 있을 것으로 보고 있다고 밝혔다.

핵심 재무 데이터

HPC 매출에는 임대료 4,600만 달러, 임차인 맞춤형 설비 구축(fit-out) 매출 1,130만 달러, 전력 비용 전가(pass-through) 매출 540만 달러가 포함됐다. 보고된 HPC 임대 부문 마진율은 약 17%였다. 맞춤형 설비 구축 활동, 매출 발생 전 비용 및 비현금성 운영비용, 개발비, 전력 비용 전가를 조정한 후 경영진이 산출한 마진율은 약 83%였다.

사업 및 운영 실적

레이크 매리너

CB-3는 완전 가동 상태에 들어갔으며 7월 13일부터 임대가 개시됐다. 준공과 함께 레이크 매리너 증설과 연계된 6억 달러 규모의 구글 신용 보강도 개시됐다.

CB-4의 경우 첫 두 개 데이터 홀에 임차인 장비가 배치되고 전력 공급이 시작됐다. 나머지 홀은 시운전 중이며, 경영진은 2026년 4분기 말까지 전면적인 단계적 가동을 목표로 하고 있다. CB-5 시운전은 해당 분기 중 시작될 예정으로, 2027년 1분기 4개 데이터 홀 전체의 단계적 서비스 개시를 뒷받침할 계획이다.

테라울프는 울프 컴퓨트(Wulf Compute) 총 프로젝트 비용을 핵심 IT 메가와트당 약 890만 달러로 추정하고 있으며, 이는 기존 가이던스인 800만~1,000만 달러 범위 내에 있다. 프로젝트 자본적 지출(CAPEX) 중 약 29억 달러가 집행 완료됐고 약 10억 달러가 남아 있으며, 잔여 지출의 약 3분의 2는 집행이 확정됐다.

저스티파이드 캠퍼스

하즈빌(Hawesville)에 위치한 저스티파이드 캠퍼스는 현재 말뚝 기초 공사가 진행되는 등 건설 중이다. 플루어(Fluor)의 EPC(설계·조달·시공) 계약 범위에는 3개 데이터센터 건물 전체에 대한 엔지니어링, 조달, 건설, 시험 및 시운전이 포함된다.

경영진은 엔지니어링이 약 90% 완료됐으며 12월 1일까지 100%에 달할 것으로 예상된다고 밝혔다. 조달 기간이 긴 핵심 장비(long-lead equipment)의 80% 이상이 확보됐으며 연말까지 전체 조달 완료를 목표로 하고 있다. 계약상 보호 장치로는 오픈북 원가 공개, 명확한 마일스톤, 최대보증가격(GMP) 구조 등이 포함된다.

테라울프는 9월 30일 기준 저스티파이드에 약 5억 5,000만 달러의 자기자본을 출자했다. 회사는 앤트로픽의 임대 의무에 대한 구글의 신용 지원(백스톱)을 바탕으로 한 프로젝트 수준의 부채 조달을 통해 잔여 자금 소요를 주로 충당할 계획이다.

머스키

머스키(Muskie)는 켄터키 파워(Kentucky Power)의 기존 요금 체계에 따라 500MW 규모의 전력 공급 계약을 체결했으며, 2028년에 첫 공급이 이뤄질 예정이다. 테라울프는 추가로 500MW 확보도 추진 중이며, 해당 용량 일부의 공급 시점을 2029년으로 앞당길 수 있도록 켄터키 파워 및 AEP와 협력하고 있다고 밝혔다.

두 번째 전력 블록은 켄터키 파워가 제안한 760MW 규모의 빅 샌디(Big Sandy) 복합화력 가스터빈 프로젝트를 통해 공급될 예정이다. 경영진은 켄터키 공공서비스위원회(PSC)에 인허가 신청이 접수될 것으로 예상하고 있으며 2027년 중반까지 승인을 목표로 하고 있다. 테라울프는 머스키의 임차인이나 임대 계약 체결 일정을 확정하지 않았다.

경영진 가이던스

• 경영진은 계약된 핵심 IT 용량의 약 절반이 2027년 1분기 말까지 가동되어 약 7억 달러의 연간 환산 런레이트 임대 매출을 창출할 것으로 예상하고 있다.
• CB-4는 2026년 4분기 말까지 최대 168MW의 추가 핵심 IT 용량을 기여할 것으로 예상된다. CB-5는 2027년 1분기에 추가로 168MW를 더하는 것을 목표로 한다.
• 계약 용량이 상용 서비스에 돌입하고 매출 발생 전 비용이 감소함에 따라, 보고된 HPC 임대 마진율은 장기 목표치인 약 85% 수준으로 점차 개선될 것으로 전망된다.
• 저스티파이드의 단계적 상용 서비스 개시 목표 시점은 2027년 후반부터 2028년 2분기 사이로 유지되고 있다. 프로젝트 비용은 핵심 IT 메가와트당 1,000만~1,200만 달러 범위 내를 유지할 것으로 예상된다.
• 테라울프는 기존 사이트 포트폴리오를 활용해 매년 250~500MW 규모의 핵심 IT 용량을 추가 계약한다는 목표를 유지했다.
• 현재의 계획과 가정을 바탕으로 경영진은 기존 유동성과 애버내시 매각 예상 대금을 통해 주식 발행 없이도 레이크 매리너의 잔여 약정액, 저스티파이드의 잔여 지분 출자금, 머스키의 임시 신용장(LC) 요건을 충당할 수 있을 것으로 판단하고 있다.

리스크 및 주요 관전 포인트

조달 금리 상승과 신용 스프레드 확대는 여전히 핵심 우려 사항이다. 경영진은 신규 임대 계약 체결 시 임대 수익률과 테라울프의 가중평균자본비용(WACC) 간 스프레드를 반드시 확보해야 한다고 강조했다.

전력 시스템, 냉각, 조달, 건설, 고객 하드웨어 통합 및 시운전 전반에 걸친 실행 리스크는 여전히 실질적인 요인이다. 테라울프는 플루어와의 협력, 계약상 이행 보호 조항, 최대보증가격(GMP) 구조를 활용해 비용 및 공기 지연 리스크를 완화하고 있다.

고객의 신용도 역시 프로젝트 파이낸싱에 있어 결정적이다. 경영진은 향후 임대 계약 시 투자적격 등급 임차인이거나 투자적격 등급의 신용 보강을 갖출 것을 요구할 것이라고 밝혔다. 앤트로픽 임대 계약에 대한 구글의 신용 지원은 앤트로픽이 지정된 투자적격 신용등급을 획득하면 종료될 수 있다.

머스키의 두 번째 500MW 전력 블록은 전력회사의 개발 상황, 규제 당국의 승인, 신규 발전소 건설에 달려 있다. 체서피크(Chesapeake) 역시 매도인이 충족해야 하는 미결 규제 승인 및 기타 거래 종결(클로징) 조건의 적용을 받는다.

애널리스트 Q&A 주요 내용

• 자본 구조: 경영진은 잠재적인 자금 조달 수단으로 고객 선수금, 고객 지분 투자, 프로젝트 레버리지, 안정화된 자산의 소수 지분 매각 등을 꼽았다. 이어 매력적이지 않다고 판단되는 밸류에이션에서 주식 발행을 강제하는 대규모 임대 계약은 체결하지 않겠다는 입장을 재확인했다.
• 구글 신용 보강: 테라울프는 현재 앤트로픽이 투자적격 등급이 아니기 때문에 조달 효율성을 높이기 위해 신용 지원(백스톱)이 마련됐다고 설명했다. 임대 개시 이후 채무불이행(디폴트)이 발생할 경우, 해당 지원책을 통해 임대 계약을 의무 승계하거나 부채를 감축하는 해지 정산금을 지급하도록 규정되어 있다.
• 저스티파이드 건설 비용: 경영진은 거의 완료된 엔지니어링, 조달 기간이 긴 핵심 장비의 선행 조달, 플루어의 최대보증가격(GMP) 계약 구조를 메가와트당 1,000만~1,200만 달러 추정치의 근거로 제시했다.
• 머스키 임대 전략: 회사는 하이퍼스케일러와 프론티어 AI 기업 모두를 검토 중이다. 경영진은 고객 선정 시 단순한 다각화보다는 임대료, 신용 보강, 자금 조달 요건, 계약 기간, 위험조정수익률에 따라 결정될 것이라고 밝혔다.
• 자산 재활용: 테라울프의 기본 전략은 인프라를 장기적으로 구축하고 소유하는 것이다. 다만 경영진은 자산이 안정화된 이후 해당 방식이 가치를 입증하고 신규 개발을 뒷받침할 수 있다면 인프라 펀드로부터 프로젝트 수준의 투자를 유치하는 방안도 고려할 수 있다고 말했다.
• 규제 환경: 경영진은 명확한 규제 체계가 지역사회를 보호하고 금융 조달이 가능한 프로젝트와 투기성 개발을 구분해 줄 수 있다고 말했다. 그러면서 뉴욕주에서의 안정적인 운영 실적과 켄터키주 주정부 및 지방정부와의 우호적인 관계를 강조했다.

실적 발표 콘퍼런스콜 녹취록 전문

전체 실적 발표 컨퍼런스 콜 녹취록

경영진 발표

Operator

Greetings, and welcome to the TeraWulf Third Quarter 2026 Earnings Conference Call.

[Operator Instructions]

As a reminder, this conference is being recorded. It is now my pleasure to introduce your host, John Larkin, Senior Vice President and Director of Investor Relations with TeraWulf. Thank you. You may begin.

John Larkin

Good morning, and welcome to TeraWulf's Third Quarter 2026 Earnings Call. Joining me today are our 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'll turn the call over to our Chairman and CEO, Paul Prager.

Paul Prager

Thanks, John, and good morning, everyone. On Wednesday, we announced 2 important milestones at justified. Google's credit support in connection with the financing of our $19 billion Anthropic lease and a definitive EPC agreement with Fluor for the development of the campus. I want to start there because those agreements directly address 2 of the biggest questions facing our sector today. How do you finance infrastructure of this scale efficiently and how do you manage the execution risk associated with building it?

There has obviously been significant volatility across AI infrastructure. Investors are focused on higher financing costs, leverage, the amount of capital being deployed across the sector and whether the returns ultimately justify that investment. Those are legitimate questions, and we are highly aware of the environment. Credit spreads have moved and there is a significant amount of infrastructure financing coming to the market.

At the same time, demand for compute remains extremely strong. Deliverable power remains scarce, and our customers continue to make very substantial long-term commitments. This brings me to our justified data campus in Hawesville. We already have the equity required to fund the project. Google's credit support materially strengthens the financing profile. And we fully expect to fund the remaining capital requirement with project-level debt with contractual protections in place to preserve our returns. We have always believed in raising capital when you can, not when you have to. We did just that.

We have the equity, we have the customer and we have the credit support and a clear path to competitive long-term debt financing. We structured this transaction from the beginning so that financing risk would not get pushed on to TeraWulf's shareholders. We protect our economics that is fundamental to how we operate.

The floor agreement addresses the other side of the equation, execution. We know firsthand how difficult these projects are to build because we are doing it at late Mariner. We understand the labor, procurement, electrical infrastructure, cooling, commissioning and customer coordination involved in delivering capacity at this scale. At Justified, we made a deliberate decision to pair that experience with 1 of engineering excellence and 1 of the strongest construction organizations in the world. The definitive EPC agreement covers all 3 data center buildings and expands Fluor's role from early engineering and preconstruction into engineering, procurement, construction, testing and commissioning. It includes defined delivery milestones, open book cost reporting, procurement oversight and contractual cost and schedule protections with guaranteed maximum pricing being established as engineering and procurement progress.

That gives us substantially greater execution capacity without trying to build a global engineering and construction organization within TeraWulf. Construction is already progressing, pilings are underway, and we continue to target phased, ready-for-service deliveries beginning in late 2027 and continuing through the second quarter of 2028. We also continue to expect project costs to remain within the approximately $10 million to $12 million per critical IT megawatt range, which we have previously discussed on a comparable basis. When you look at Justify today, we have a world-class customer in Anthropic, approximately $19 billion of contracted revenue, investment-grade credit support from Google. The equity required an experienced EPC partner in Fluor and a clear path to project level debt financing of equal import we also enjoy a fantastic regulatory framework in Kentucky.

Congressman Reck stated, this historic investment in Hancock County is a great example of what's possible with a collaborative community-first approach to data center development. I am confident this project will serve as a national model for communities looking to grow their economies, participate in the industry of the future and improve the well-being of their residents. Justified is not a theoretical development.

It is contracted infrastructure that we are building to deliver. Lake Mariner is where you can already see that model working CB-3 is online and generating revenue. CB-4 is moving into commercial operation with the remaining data halls scheduled to come online sequentially through the end of the year. CB-5 follows beginning in January with phased delivery targeted through the end of the first quarter.

By the end of Q1, we expect approximately half of our contracted critical IT capacity to be operating and generating revenue, and we expect to exit the quarter with approximately $700 million of annualized run rate leasing revenue. That is an important milestone for TeraWulf. For the last several years, Investors have largely been valuing what we were building. Increasingly, our investors will see these contracts convert into operating infrastructure, recurring revenue and cash flow.

And we continue to expect Lake Mariner to be delivered close to the midpoint of our original $8 million to $10 million per critical IT megawatt cost guidance. We told investors what we expected it to cost and we are delivering against that expectation. Nazar will give you more detail on both Lake Mariner and justified in just a moment. But the larger point is simple, contracted megawatts are becoming operating megawatts. Operating megawatts are becoming revenue, and revenue is becoming cash flow. That is execution.

It is also important to understand the business we are building. TeraWulf is not a neocloud. We are an energy infrastructure company. we controlled the land and power, develop and operate the infrastructure and lease that capacity under long-duration contracts. We do not own the GPUs, we do not take technology risk where the chip may win or may not, in the next generation of AI. Our job is to secure scarce power, build high-quality infrastructure around it, contract that capacity with strong customers, finance it efficiently and generate durable cash flow.

We have been developing and financing large-scale energy infrastructure for decades. This is what we know how to do, and we are building a business not collecting megawatts. Growth only matters if it creates cash flow and equity value for our shareholders. That discipline is particularly important in the current capital markets environment.

Higher financing costs matter, and we factor them into every new transaction. As construction costs, tariffs, financing costs have changed, our contract economics have changed with them. We are not going to absorb financing risk that belong somewhere else in the structure, and we are not going to accept a bad return simply to announce another lease. There is another side of that equation, and that's the cost of waiting.

For Frontier AI companies investing billions of dollars in compute, 6 months matter. 12 months matters enormously. When time to market is critical, the value of real deliverable power increases significantly. And that brings me to Musk. We continue to be extremely excited about Muskie. And there is no doubt this site in Kentucky is far more valuable today than it was at the beginning of the week. Muskie has scale investment-grade utility partners in Kentucky Power and AEP, front of the meter grid service, attractive power economics and a highly supportive state and local environment.

Most importantly, we now have a contracted path to the initial gigawatt of electric service. That is real deliverable power. A piece of land next to a transmission line is not a data center and a queue position is not power. The questions we care about are straightforward. When does the power actually show up? Who is responsible for building the infrastructure? What approvals are in place? What does the energy cost? And can the developer actually execute? At Muskie, we are working closely with Kentucky Power and AEP on opportunities to accelerate portions of the development schedule, including potentially pulling forward that second 500 megawatts.

At the same time, our view of the ultimate scale of Muskie continues to increase and the level of customer interest we are seeing reinforces that conviction. We are in active discussions with a number of very high-quality prospective customers. And 1 of the questions we get asked is how we think about customer diversification. Diversification is clearly valuable. We understand the benefit of adding another large, high-quality customer to the portfolio. But you cannot look at diversification in a vacuum. There can be meaningful economic differences between leasing to a hyperscaler and leasing to a frontier model company.

The credit profile may be different, the financing may be different but the rent yield and return on our invested capital can also be materially different. Team and culture matter as well. When you are entering into a long-duration relationship, alignment, speed of decision-making and the ability to work through issues together can have a significant impact on execution and, ultimately, on returns. We weigh all of it. What is the rent? Who is the credit? What does it cost us to finance? How much capital do we have to put in? What protections do we have? What's the duration of the deal? And ultimately, what is the risk-adjusted return to our shareholders?

My job is to make the decision that creates the greatest risk-adjusted value for TeraWulf shareholders. And I take that seriously. Justified is a good example. We have a frontier AI customer with very attractive long-term lease economics paired with investment grade credit support from Google. At Muskie, we have something ably valuable, scale, power and time. That gives us options. It allows us to be patient and disciplined about the selection of a customer. the terms, the credit structure and the economics rather than feeling pressure to just sign the first available tenant.

I am not going to predict when we sign the next lease or who that customer might be. But I am highly confident that Muskie will deliver highly confident in the level of customer interest and highly confident that we could structure a transaction that reflects the value of this important asset. Just as importantly, we are not going to sign a lease that forces us to issue equity at a valuation, we do not believe reflects the value of the company. The commercial structure, customer economics and financing plan, they have to work together. We will bring that home.

Our objective is to structure a financing solution that recognizes the value we've created at Muskie, preserves attractive returns and minimizes dilution to our shareholders. Development will be phased. Capital will be raised and deployed alongside contracted demand. And we will remain intensely disciplined about protecting the balance sheet as we grow.

More broadly, we remain confident in our target of contracting an incremental 250 to 500 megawatts of critical IT capacity annually. And importantly, we believe that we can achieve that target with the portfolio we control today. We do not need to buy sites simply to make our pipeline look bigger, any new opportunity has to have real power and compete for our capital and create value for our shareholders. Do we control the power, can it scale, can it be delivered when the customer needs it? Can we finance it intelligently and can we earn a compelling return. The same discipline applies to regulation and the communities where we operate.

I have said publicly for years that not every data center should be built and not every site is appropriate. We welcome clear and durable frameworks. Good regulation creates certainty, protects communities, protects our project and helps distinguish credible, well-capitalized projects from speculative ones. That is consistent with how we have been developing these many years.

So when I step back, I feel very good about TeraWulf today. At Lake Mariner, we are delivering, CB-4 is coming online this quarter, and CB-5 follows in the first quarter next year. Once the campus is fully ramped, we anticipate more than $700 million of annualized lease revenue. And we're already planning the next opportunity for growth at Lake Mariner with a path to scale to approximately 750 megawatts to accommodate CD-6.

The infrastructure is in place, fully permitted, contracts signed and the cash is flowing. We are also in a professional and established regulatory framework where our management team has been operating for well over a decade. At justified, we have a world-class customer in Anthropic and approximately $19 billion of contracted revenue. We have investment-grade credit support from Google, the equity required under our current plan and a clear path to project level debt financing with Fluor as our EPC partner and guaranteed maximum pricing in place, we have the execution platform to deliver at scale. And we are doing it in Kentucky, where we have found a highly constructive state and local environment that understands the economic and infrastructure opportunity these projects represent.

Justified is permitted, contracted, financeable and under construction. And at Muskie, we have something increasingly scarce. Scale, attractive power economics and a real path to a very large block of deliverable power. We have an investment-grade utility partner in Kentucky Power and AP. Again, a highly supportive state and local environment and most important, a contracted path to an initial gigawatt of front of the meter service.

That support is real. The Basheer administration has publicly recognized TeraWulf's commitments to pay the full cost of our power, protect ratepayers and the environment contribute significant local and school taxes and create jobs for Kentucky and has described Muskie as a project that will be positive for the community.

We expect to pair that power with a high-quality EPC partner and guaranteed maximum pricing, giving us the same cost and execution discipline we have put in place at justified. That is a very strong hand. But megawatts alone do not create shareholder value. What matters is whether the power is real, whether the capacity is contracted the credit behind those contracts, what it costs us to build and finance and the returns that accrue to our shareholders.

Demand is robust. Deliverable power remains scarce. We have the sites. We have the capital. We have the customers and the management team that has demonstrated that we can execute. We are building generational infrastructure assets designed to stand the test of time. Technology will surely evolve. Customers may change. Capital markets will move. But what we are doing today is fundamentally the same thing we were doing yesterday, and the same thing we will be doing tomorrow, securing scarce power, building critical infrastructure around it contract it intelligently and executing. That focus will never change. Our discipline will never change. And our commitment to protecting the balance sheet in creating long-term shareholder value. That will never change.

With that, I'll turn it over to Nazar.

Nazar Khan

Thank you, Paul. I'll focus on execution across Justified, Lake Mariner and Muskie. Starting with Justified, yesterday's Fluor announcement is an important step in moving the campus from development into full-scale construction. Fluor has already been working with us on early engineering and preconstruction activities. The definitive EPC agreement expands that role across engineering, procurement, construction, testing and commissioning for all 3 data center buildings. For a project of this scale, that matters.

The challenge is not simply constructing the buildings. Electrical systems, cooling infrastructure, utility systems, equipment procurement, customer hardware and commissioning all have to come together against 1 coordinated delivery schedule. Fluor brings deep engineering expertise, global procurement capabilities, access to skilled labor and extensive experience executing large complex infrastructure projects. That complements the development and power expertise we have built internally and significantly expands our ability to execute at scale. The agreement also gives us greater visibility into cost and schedule through defined milestones, open book cost reporting, procurement, oversight and contractual performance protections with guaranteed maximum pricing being established as engineering and procurement in advance.

Construction is progressing today. Pilings are underway and engineering and procurement continues to advance, and we remain on track for phase ready-for-service deliveries beginning in late 2027 and continuing through the second quarter of 2028. We also continue to expect project costs to remain within the approximately $10 million to $12 million per critical IT megawatt range we have previously discussed on a comparable basis.

Turning to Lake Mariner. This is where you can see the execution capability we have built over the last several years. CB-3 is fully online, operating and earning revenue, a milestone for our contract with Fluid stack. At CB-4, tenant equipment is fully deployed and energized in DH4 and DH3, which are the first 2 data halls of 4 data halls in the building. Commercial operation and rent commencement for those first 2 data halls in CB-4 is expected shortly.

For the remaining 2 data halls in CB-4, commissioning has already begun on DH2 and will start on DH1 in the middle of this month. These last 2 data halls are scheduled to come online in the balance of the fourth quarter. For CB-5, commissioning activities are set to start this month to support a Q1 2026 commencement across all 4 data halls. CB-4 and CB-5 are copies of each other. The final commissioning and commencement of operations at CB-4 positions us well to delever CB-5 in a similar manner in Q1 of 2026.

This phased approach is important because these buildings do not move from construction to full operation on a single date. Each data hall progresses independently through installation, startup, commissioning, customer equipment integration and final acceptance. That allows us and our customers to bring capacity on progressively as each data hall is ready. There's also significant alignment around these delivery dates. Lake Mariner represents the next major block of capacity in our end customers' deployment plan. So TeraWulf, Fluid Stack, Google and the broader project team are tightly coordinated around construction, commissioning, hardware delivery and systems integration.

From a cost perspective, we continue to expect the overall Wulf Compute Development to be delivered close to the midpoint of our original $8 million to $10 million per critical IT megawatt guidance range. At this stage, we have substantially greater visibility into the remaining labor, equipment, commissioning and construction scope, which gives us increasing confidence in both the schedule and ultimate cost outcome.

Finally, turning to Muskie. The key point from an engineering and development perspective is that the power plant is real and utility supported. We are working with Kentucky Power and AEP on a defined path to large-scale front-of-the-meter service. The utility infrastructure, service arrangements and development responsibilities are being advanced together rather than a separate work streams. We also continue to work closely with Kentucky Power and on opportunities to optimize and potentially accelerate the development sequence, including the potential to pull forward portions of the second 500 megawatts into 2029.

More broadly, we believe data center demand will increasingly become an important catalyst for investment in transmission and generation infrastructure across the country. Muskie is a good example of where our long history in developing and operating power infrastructure gives us an advantage in working with utilities to develop practical solutions for very large loads.

Across all 3 campuses, the objective is the same: take secured power and contracted customer demand, convert them into operating infrastructure safely, reliably and efficiently. Lake Mariner demonstrates that we can do it, Justified expands that execution capability to a much larger scale and Muskie provides the next major platform behind it.

With that, I'll turn it over to Patrick to review the financial results for the third quarter.

Patrick Fleury

Thank you, Nazar. The third quarter results reflect an increasing contribution from HPC's revenue and disciplined execution at Lake Mariner. Following quarter end, we entered into 3 transactions of significant financial importance. First, we amended the revolving credit facility, increasing the aggregate commitments from $250 million to $400 million, with an incremental $150 million accordion feature for a total size of $550 million. Importantly, we also eliminated the $150 million LC supplement, making the entire revolver capacity available for letter of credit issuance.

Second, we secured a Google credit backstop for project financing in support of $19 billion, 20-year lease with Amope at the justified data center campus. And third, we entered into an EPC agreement with Fluor. Together, the Google credit support and Fluor EPC agreement materially advanced both the financeability and execution certainty of justified and position us well to complete the project level debt financing.

Our Wulf Compute 7-75% senior secured notes issuance was the first financing of their kind in this sector and have since been followed by a number of similar public and private transactions. The bonds have continued to perform well relative to peers, reflecting Google's credit support, TeraWulf substantial equity contribution of approximately 30% of project costs and our growing track record of execution at Lake Mariner.

I'll focus my remarks today on the third quarter results, our updated Lake Mariner capital outlook and the liquidity available to fund our contracted projects and future growth. Revenue for the third quarter was $76.2 million compared with $44.8 million in the second quarter, HPC lease revenue increased 97% quarter-over-quarter to $63.0 million from $31.9 million and represented approximately 83% of total revenue. HPC revenue consisted primarily of $46.0 million of rent, $11.3 million of tenant fit-out revenue and $5.4 million of power pass-through revenue. The CB-3 lease commenced on July 13, and our third quarter results reflect 102 critical megawatts operating and generating lease revenue at Lake Mariner.

Completion of CB-3 also satisfies the applicable conditions for $600 million of Google's credit support to become effective, further strengthening the contracted revenue profile of the late manner build-out. Turning to operating results. Cost of revenue exclusive of depreciation increased to $25.2 million from $12.4 million in the second quarter, primarily reflecting higher power costs and $10.1 million of tenant fit-out costs.

Demand response proceeds, which are reported as a reduction in cost of revenue, increased to $3.2 million from $2.8 million in the second quarter. Operating expenses increased to $39.5 million from $23.4 million as we continued scaling the platform ahead of additional HPC capacity entering service.

The increase primarily reflected $4.7 million of additional site level expenses, including security, labor and pre-revenue operating costs. million of site preparation and demolition costs associated with future development and $1.3 million of amortization of deferred charges partially offset by $2.6 million of lower minor equipment repair costs. Reported HPC Leasing segment profit margin was approximately 17% during the quarter compared with our long-term target of approximately 85%.

Importantly, the reported margin includes $11.3 million of tenant fit-out revenue and associated costs, $11.1 million of free revenue and noncash operating costs at Wolf Compute and $18.0 million of development costs across our portfolio. Adjusting for those items and excluding approximately $5.3 million of power costs from both revenue and expense, HPC Leasing segment profit margin was approximately 83%. We expect reported margins to move toward our long-term target as additional contracted capacity enter service, pre-revenue cost decline and mature HPC lease revenue becomes a larger portion of the overall mix.

SG&A expense decreased to $57.1 million from $126.9 million in the second quarter. Excluding stock-based compensation, adjusted SG&A was $27.1 million compared with $28.6 million in the prior quarter.

Depreciation increased to $29.6 million from $21.2 million in the second quarter, primarily reflecting additional HPC infrastructure placed into service at Lake Mariner. Interest expense was $53.6 million compared with $56.4 million in the second quarter, and we recognized interest income of $28.6 million in Q3 compared to $28.9 million in Q2.

Cash interest paid was $5.0 million compared with $125.7 million in the second quarter. GAAP net income attributable to TeraWulf was $709.9 million compared with a net loss of $939.9 million in the second quarter. The increase was primarily driven by the noncash fair value adjustment associated with the Google warrants and the gain on sale of equity interest in the Abernathy joint venture. Non-GAAP adjusted EBITDA was negative $12.8 million compared with negative $18.3 million in the second quarter.

Turning to the balance sheet. Cash and restricted cash totaled approximately $2.4 billion as of September 30. At the parent level, we held approximately $1.1 billion of cash and restricted cash at quarter end with approximately $300 million of this cash used to collateralize LCs for the Muskie data campus.

We expect this cash collateral to be replaced with LCs from our revolving credit facility in December when the CB-4 lease commences and our revolving credit facility capacity increases from $150 million to $400 million. We also expect to receive an incremental $150 million from the sale of the Abernathy JV on or before December 31, 2026, and another approximately $130 million on or before April 30, 2027, subject to the terms of the transaction.

At Wolf Compute, we had approximately $1.2 billion of gross cash at quarter end or approximately $800 million after accounting for debt service reserves and interest during construction accounts. Approximately $2.9 billion of project capital expenditures have been completed with approximately $1 billion remaining.

Approximately 2/3 of the remaining capital expenditures are committed, providing increased visibility into the remaining capital requirements. Our original cost guidance for Wulf compute was $8 million to $10 million per critical IT megawatts. We currently estimate total project cost at approximately $8.9 million per megawatt, well within that original guidance range. At the justified data campus, we have contributed approximately $550 million of equity as of September 30, representing a substantial portion of the equity contribution contemplated under our current project financing plan.

Our near-term liquidity remains robust, over $1 billion of cash at TeraWulf parent and up to $400 million of revolving credit facility and LC capacity. Based on our current plans and assumptions, our existing liquidity and expected Abernathy sale proceeds provide the capacity to fund our remaining Lake Mariner commitments, our remaining equity investment at justified interim letter of credit requirements at Muskie, all without accessing the equity capital markets.

And as all contracted leases at Wulf Compute commenced through the first quarter of 2027, we expect the project to generate significant cash flow, rapidly repay debt and operate at moderate leverage. That creates another important source of capital for TeraWulf. Unlocking equity from stabilized assets and recycling that capital back to the parent for future contracted growth will be a priority in 2027. Our financial approach for existing and future projects remains simple. Sign a lease with an investment-grade tenant or backstop provider, walk in our build cost and finance the spread. This is how large scale long-duration energy infrastructure has been contracted, built and financed for decades. As the cost of capital increases, preserving that spread becomes even more important.

We will remain disciplined about the relationship between our weighted average cost of capital and the yield on our leases with the objective of ensuring that each incremental megawatt we contract creates value for shareholders. In summary, the third quarter results reflect a financial profile increasingly driven by long-term contracted HPC revenue.

The fourth quarter will accelerate that trend as CB-4 begins phased lease commencement adding up to 168 critical megawatts of contracted capacity at Lake Mariner, followed by CD-5 in the first quarter. Cash flow is ramping. Approximately 3/4 of Wulf compute capital expenditures were already behind us as of September 30. We -- we have substantially greater visibility into the remaining capital requirement, and we have the liquidity to complete our contracted developments and fund the next phase of growth while remaining disciplined about the balance sheet. With that, operator, we are ready to take questions.

Operator

[Operator Instructions]

Our first question comes from Mike Grondahl with Northland Securities.

질의응답

Mike Grondahl

A competitor earlier this week announced kind of a capital-light lease, very robust development yield in kind of a modular construction approach, is that something that might be on the table or possibility for the Muskie site?

Paul Prager

Mike, listen, it looks like a great deal on a return basis, but I don't really have the full picture on all the details. In our deals, we believe we are a strategic partner to our end user. We like to work closely with the end user and deliver creative ideas around delivery speed, design, financing, uptime flexibility. Given capital requirements, we would fully expect to see an evolution of our deliverable. We've spoken to hundreds of our investors in the past 3 months, and I consistently hear about terminal value. We have a very strong view that our sites have strong terminal value, given a roughly 5-year upgrade cycle, attractive power prices and locations that allow for reference.

Certainly, the structure that Ryan did is interesting, but it's not currently our primary focus.

Patrick Fleury

And if I can add -- this is another here, Mike. If I can add a few things. If you just -- if we look at the space, right? It's kind of a function of what are you spending and what are you earning a return on. And so just from the outside looking in on that structure, -- they're spending $2 million on kind of a powered land structure for 910 megawatts since they're getting that. It's not clear on how much of the next kind of $4 million bucket kind of their earning. And then if you compare that to a colo model, as we mentioned earlier, we're in that $10 million to $12 million range per critical IT megawatt for delivering kind of a turnkey building at Muskie.

So we're looking to kind of earn a return on that full $10 million to $12 million. So I think it's a function of both how much capital are you putting up what is the cost of that capital? And what's the spread you're earning on it. We remain of the view that the model that we have kind of value -- creates the most value per megawatt of power that we have available. And so kind of providing that full stack. And on a risk-adjusted basis, we think we kind of generate kind of the most equity value per megawatt with the turnkey colo solutions that we're going after.

Mike Grondahl

Got it. That's helpful, guys. And then maybe just as a follow-up, how did you decide at the end of the day to use the Google backstop versus maybe just going direct with Anthropic. Can you talk about kind of what you think you're going to benefit there from the Google backstop?

Patrick Fleury

Yes. Mike, it's Patrick. So as you know, credit has been from day 1 and remains our #1 focus. Obviously, Entropic is a terrific client, and we're working across multiple sites in partnership with them. but they are not public, and they're not investment grade yet. And as you know, that is -- has been our #1 focus. We have to have investment-grade support so that the cost of capital is efficient for us and that we create durable equity value when we sign leases. So that will be always first and foremost in our minds as we create value here.

Mike Grondahl

Perfect. Good luck this fall and winter, guys.

Operator

Our next question comes from Stephen Galgoe with KBW.

Stephen Glagola

Nazar, you hit on this a bit in the prepared remarks, but can you just walk us through more what gives you confidence in the $10 million to $12 million per megawatt CapEx estimate for Hawesville particularly given recent cost overruns across the industry? And how far along is the design for and how much of the cost is fixed today and who ultimately bears the risk of any overruns?

Nazar Khan

Sure. So great question. On CB-3 buildings at Justified. And so -- and this is kind on the Justified side or the Muskie side, Stephen.

Stephen Glagola

This is -- I mean this was on the Justified side, but Yes, yes.

Nazar Khan

So on the justified side, we're kind of at 90% on the engineering and the drawings, and we expect to be at 100% by December 1. So that then leads into the procurement. We are at over 80% today kind of on the long lead equipment and we'll be at 100% by year-end.

So between the engineering, which we'll have locked down here, by the first of December on the long lead equipment which we will have locked down by the end of the year. We feel like we're in pretty good shape. Overall, kind of 80% of the scope under the floor agreement, so that would be all the construction and the long lead equipment will also kind of be buttoned down kind of by that pipe early December.

So that's what gives us kind of good visibility into both from what are we building and what is the cost of what we're building. So that's where we sit today. And if you look at just where we are on total cost overall compared to Lake Mariner, you can see kind of a fairly meaningful change in what our cost per megawatt was there versus what we're seeing kind of real time on justified here. So hopefully that gives a sense of the confidence around both kind of what we're building through kind of completion of the design engineering as well as kind of procurement of all the long-lead equipment required to support the project. In terms of -- and then just kind of the last piece you had there in terms of just kind of who kind of bears cost overruns? There -- we mentioned in the prepared remarks, -- this is kind of a GMP gross max price structure. And so there are components that either subcontractors bear with respect to work that they're doing or kind of the combination of all that floor bears together. So have worked closely with Florida ensure that we are very well defined on what it is we're building so that there's clarity on the responsibility for the cost. And so through that G&P structure, that's another layer of protection that we have as we roll forward.

Operator

Our next question comes from Eric Luebchow with Wells Fargo.

Eric Luebchow

Patrick, I just wanted to explore, you talked about kind of protecting your spreads if you see a change in the cost of capital after you sign a lease. So maybe you can talk a little bit about mechanisms in the contracts to potentially rentalize if you do have to raise at a higher cost of capital what protections you put in place both for the hospital site and then maybe on a go-forward basis?

Patrick Fleury

Yes. Look, the terms of our leases are confidential, so I can't give you the details, but what I can tell you is a couple of times. First, similar to our structure of Wulf compute, Google has entered into agreements to back stop Anthropic's obligations under the lease. Upon the lease commencing, Google's credit support then protects against Anthropic default through, one, a mandatory lease assumption or to a debt reducing termination payment. So that's the same structure as well compute,

Unlike our structure at Wulf compute and that evident in all of our disclosures, TeraWulf is not advancing Google any incremental upfront equity here. and the cost of Google's credit support is recoverable from the tenant under the lease. In addition, Google's credit support falls away when entropic achieved certain investment-grade credit ratings. So those are all the details I can give you today. The rest of the details will be public when we kick off the financing. But I think, as Paul has said publicly, when we negotiate these leases, we seek to preserve our economics. And so there are a number of ways to effectively do that -- and I can assure you that we will keep doing that in future leases as well.

Eric Luebchow

Great. And just a follow-up. As you look at the range of options for Muskie, so far, you guys have the deals with noninvestment-grade companies with backstops. I mean does the attractiveness of a direct IG lease is currently in the current market. Is that more attractive today? Or how are you kind of thinking about the range outcomes for kind of credit for Muskie?

Patrick Fleury

Yes. Again, I think reiterating what I said particularly in a market like this, but I want to go back because we have put this front and center since day 1. Right? We are a capital markets business. We cannot be dependent on another company that is a capital markets dependent business without an investment-grade credit rating. So all of our leases will be either backed by investment-grade tenants, or have investment-grade tenants as the primary lessee because that is the only way to effectively finance these projects. In addition, our projects will have a healthy equity layer. I would point out, again, Wulf Compute was the first of its kind in the space. And there's about 25 bonds on my Bloomberg screen and Wulf compute other than 1 bond trades the absolute best.

And that's a combination of the fact that we have an investment-grade tenant and that TeraWulf has put in a significant -- very significant close to 30% equity check into that business. and that we are executing. So again, I think those are the themes that you will see time and time again from us. We are not going to deviate from that.

Operator

Our next question comes from Jed Dorsheimer with William Blair.

Jonathan Dorsheimer

I guess first one, maybe a complement. I mean you guys have kind of forged a pathway that historically actually looks like 1 of the few that's kind of narrowly avoided some of the pitfalls of or challenges in the market. And Meta actually put out an interesting post last night that kind of differentiated the difference between -- or differentiated between campaigning and governing.

And so my first question to Paul, I'm just curious, there's a lot of fear and rhetoric in the markets right now. I'm just curious how you see things playing in terms of governing functions associated with around data center and a lot of the noise that's in the market? And then I have a follow-up.

Paul Prager

If you take a look at the markets where we're active, right? In New York, Lake Mariner, again, fully permitted. We're familiar with the regulatory framework there. We operate a call-fred power plant there for 20 years. We shut that down, mitigated it in partnership with the state, put up a data center that's very well received, brought a lot of jobs, tax revenue to the state. What the governor did there, I thought was sensible. She didn't look at sort of at coming up with a ban on data center. She looked up with coming with the framework on how we could work with future data centers.

And we weren't impacted by that, as you know, we're already permitted, and we weren't interacting with the DEC for that site. But she put it in the hands of her professionals, and I think they'll come back with something that's very sensible. As I said in my prepared remarks, good regulation is a very good thing and protect our investment. And it shows the projects that make sense from the ones that probably don't. In Kentucky, there's a state where, again, it's a democratic state. Governor Basheer has got real national prominence. Loves the facility in my prepared remarks. We quoted this office on how he thinks about Terra Wolf. He thinks we're the model for how data centers ought to be developed on a nationwide basis.

Listen, I think there are some states that haven't thought about how data center should proceed to the level that people in, for instance, Kentucky and what Virginia have. And I think that it may take a little bit of time, but data centers can be developed really in concert in partnership with the community. They are big drivers for the economy. They are of utmost importance from a national security perspective. And it affects everyday quality of but a really positive and constructive way.

So I think there's -- as CEO, I'm not supposed to comment on the markets, but Certainly, there's a lot of concern and fear out there. And -- but I think when you really get down to it, you got to be located in the right places. They have to have clarity on what permitting is like they have clarity on -- they have to have a skilled labor force. They have to have good energy policies.

And that's been the focus for us. We didn't focus on Kentucky by accident. It was all a really thoughtful and well considered development effort. And that's what you'll continue to see in our development efforts in our pipeline.

Jonathan Dorsheimer

That's helpful. And then Patrick, just as my follow-up. I look at the same Bloomberg screen in terms of the debt in you and 1 other are the only 2 green on that. I am curious as you do have I know you guys are not at Neo cloud. But as you look at the debt that's trading of some of the NeoCloubs and others in the stack, they've gone into that junk range. And I'm curious whether or not you think that a clearing event would be -- is a necessary evil to kind of reset here or what your thoughts are on the credit side?

Patrick Fleury

Yes. Look, I think there's a lot of clearing events in the market right now, right? The election was 1 of them, Anthropic's IPO was another one. So yes, I think -- but as a 20-year credit distressed analysts on the buy side, I mean this is natural for the markets, right? They -- and not only that, but you have a lot of issuance all competing, whether it's the hyperscalers, the U.S. treasury. So I think what you're seeing in the credit markets is pretty natural. They haven't frozen up, they have just moved spreads higher because you have to induce new capital to come into that market.

And to do that, you got to move spreads higher. So that's, again, why we come back to it is absolutely critical to have an investment-grade tenant so that we preserve that spread between our WACC and our yield on cost. And again, that is something that is fundamental to our business. If you look again, as I said in my remarks, every single piece of long-lived, long-term energy infrastructure in this country has been built -- that's how it's been done, whether it's nukes, refineries, gas plants, LNG facilities, and so we will continue to kind of prosecute that as we move forward.

Operator

Next question comes from Stephen Byrd with Morgan Stanley.

Stephen Byrd

Lots have been addressed. I just wanted to hit on, I think, a misconception that we sometimes hear, which relates to equity overhang in the event of signing really attractive deals and I know you all have thought about many options, and I think maybe the market doesn't fully appreciate just how many options there are to think about funding your overall capital needs. So I just wanted to go back to that topic and make sure we really address that.

Patrick Fleury

Yes. Stephen, it's Patrick. I'll take a cut at that and then Paul or Nazar can bat cleanup here. But Look, there's a lot of options in this playbook for equity, right? There is a prepayment from a customer. There is customers buying equity at higher prices. There is leverage in the system, which I mentioned in my remarks, right? I mean, Wolf compute when it comes online is significantly underlevered vis-a-vis other public and private companies out there. there -- so there's also, as I think you pointed out in a recent research report, right? There are infrastructure funds that are chomping at the bit to buy equity in stabilized assets at the project level at significantly higher multiples than where any of us trade today. So there are a lot of tools in our toolbox.

And as Paul said in his remarks, we will not sign 1 of the biggest leases that the company has ever signed and be forced to dilute our shareholders at these levels. We just will not do it. And so my job is to make sure that every single tool in the toolbox is available to us that we are looking at all of them and that's what we're doing.

Paul and Nazar I don't know if you want to add anything.

Paul Prager

I would only want to add. And our customers are entirely sensitive to that. and they work with us when we're negotiating with them to ensure we could preserve our economics and preserve the spread for shareholder value. It's really that simple.

Nazar Khan

Yes. Just -- and Stephen, just to add to what Paul just said. If you look at the customers, right, there -- and you can see this in the public companies, and you can see it now in Anthropic interest on they are making tremendous amounts of profit. So it's easy for us to have a logical discussion with them on an open book basis as to -- here is our cost of capital. Here's my cost of equity. Here's my cost of debt depending upon what the lease looks like and who the investment-grade support is on the lease. And then the simple discussion we have is, look, we're just trying to keep a spread between the yield on cost on the lease and our WACC.

And I think when you go with that open book to any hyperscaler or any frontier AI lab, it's a very easy logical discussion to explain to them that you want to preserve that, that you don't want that to be at risk. So that is the approach we take with every single 1 of our customers. whether it's hyperscalers that we're talking with or Frontier labs or otherwise, that's our goal. That's how we create double equity value, and we will look to preserve that, and we will get lease protections that allow us to preserve that.

Operator

Our next question comes from Nick Giles with B. Riley Securities.

Nick Giles

You talked about working with Kentucky Power to potentially pull forward the second 500 megawatts. So I was just curious if there's any upfront investment that may be required on Terawatt what's ultimately being contemplated there? Is it something from a generating perspective or infrastructure upgrades? Just more color would be great.

Nazar Khan

It's Nazar here. So we have a 500-megawatt ESA signed under an existing tariff of Kentucky Power. So the first 500 megawatts starts delivering in 2028. We then went back to Kentucky Power to say, "Hey, we're building substation here that has the capacity for a gigawatt, how do we get our hands on the next 500 and that next 500 megawatts requires new generation." And as time goes on here, I think more broadly, if you look across the country, power that's going to be available at 29, 30 and beyond will likely require some sort of new generation and transmission to support that.

And so they have -- Kentucky Power has a project called Big Sandy. It's a 760-megawatt CCGT that they've been working to try to figure out how to get into their system. So we work an arrangement with them where we're going -- we will be filing here with the Kentucky PSC in the next few weeks to get approval for the construction of that 760-megawatt CCGT. The load that we will have will be a gigawatt will be greater than the total capacity of that plant. And so that's kind of the second 500 megawatts. That will be a filing in front of the PSC here shortly. We're expecting approval on that by mid-2027. And importantly, within that structure, it is a clear line between what Kentucky Power is going to be spending and what we're going to be paying to use that power.

And so included within that structure are kind of 2 important components. One is, is there an incremental $100 million residential credit embedded within the agreement that will kind of be paid out over 10 years of the contract. And the second is that the interest during construction for the power plant, which will kind of go from 2027 once the approval comes in through the delivery of the power generation facility in 2031, there's kind of interest that accrues during construction. And so that will be factored into our rates.

And if you think about our load profile, we're ramping up in 2029 for that second 500 megawatts, 2028 for the first 500 megawatts, so that -- there's an overlap between us taking power mix from the grid and then kind of bridging to the 760-megawatt CCGT on the back end.

So I think it, again, it illustrates our ability to understand how to be catalyst to bring this generation on. We think we've structured a very reasonable and fair agreement with Kentucky Power that allows Kentucky Power to draw a straight line from the costs that they're incurring versus what we are paying. And therefore, that's going to give us this full gigawatt of power prior to the end of 2029 in the market that's tight on power availability.

Nick Giles

That's super clear. Just a quick follow-up separately. What should we be modeling just from a number of megawatts billing kind of as we look to Q4 here and then as we kind of exit the first quarter.

Patrick Fleury

Yes. So Nick, it's Patrick. I think what you're asking is with regard to CB-4 and CB-5. So I think exiting fourth quarter, we'll have CB-4 online. So that's an incremental 168 megawatts. We have 102 critical megawatts running there today, earning rent, so exiting fourth quarter, another 168, exiting first quarter next year, CB-5 online, so another $168 million and then Hawesville starts to come online late '27 into '28. And I think you can kind of think about that as ratable megawatts kind of CB-1, CB-2, CB-3, again, late Q4 and then CB-2, CB-3 first half of '28.

Operator

Our next question comes from Michael Rollins with Citi.

Michael Rollins

Curious if you could speak more to the current pricing environment? And as there's all these cross currents on constraints in certain markets, what are you seeing in terms of just the overall yields you can achieve, how sensitive those yields or spreads are to changes in the rate environment. And if there's anything else in terms of escalators or other pricing dynamics that are evolving that we should all be mindful of?

Paul Prager

I appreciate try not to focus on where the current market pricing is and focus on building shareholder value. we're doing just that. And in fact, as Patrick said, we're able to, as we move forward, protect the economics of our deals in our contract with our customer so that the only risk we're really taking is on execution. I talked about it in my remarks, how we mitigate that risk by being on top of it. Separately in the new project, Hawesville, we signed for as our EPC with the GMP. We would expect to do the same thing at Muskie we'll have an EPC contract with the GMP, highly credible folks. And then on the debt side, as Patrick has indicated, we absolutely require investment-grade support. Otherwise, that customer is not going to be a real interest to us. This is a business about financing efficiently and so between the terms of the deal where we protect our economic spread and our execution skill set and the investment-grade support on the debt we think we have a lock on creating shareholder value. And that's what we do. It's really that simple.

We do it every single day that is our exclusive focus.

Operator

Our next question comes from Michael Funk with Bank of America.

Michael Funk

Thank you for the question. So your comment on the call that unlocking value from stabilized assets that come online, I think I understand your point there, but there's also a lot of vehicles in market today looking to buy stabilized assets. Would you consider selling stabilized assets to recycle that capital back in a new development?

Patrick Fleury

Yes. Look, that's -- it's Patrick. Thanks for the question. Look, our business is to build and own these long term. But as I've seen in other infrastructure assets throughout the country. I mean, for example, 1 that resonates with me was when Blackstone bought into FirstEnergy's transmission business. the classic example, right, of an infrastructure fund, buying into an operating business at a significant multiple that proved out equity value at the parent.

So that has been done time and time again throughout different energy infrastructure businesses in the U.S., both public and private. So yes, is that a possibility, right, if the market continues to trade, all of us that are building data centers for NVIDIA, AMD, Amazon, Google, Microsoft others to put their chips in, Yes. But I think, first and foremost, our #1 priority, again, is getting -- executing, getting the buildings up and running, getting them online, converting that into cash flow and then repeating that process in a financially efficient way.

Michael Funk

And then 1 more if I could, quickly, going back to the sort of previous question about cost and build costs. Presumably, there is greater demand for developers, hyperscalers to contribute to cost of incremental power contribute back to communities, either through tax base or other areas. Who do you see bearing that cost? And how does it affect development yields in the future in your deal?

Paul Prager

I mean we're fortunate to have Anthropic and such a great customer. They're very socially conscious. They want to participate and contribute to the community. We do to in many respects. These are communities where our employees live and work and will be living and working in operating facilities for the next 20 years.

So I think the answer is both parties should look at this. We have always looked at it. We have historically looked at it. We have focused on teachers, kids, education, first responders. I think Anthropic as a customer because it's a very profound focus for them. I think it is for many of the hyperscalers as well. So I think you'll see the opportunity to come up with contribute post arrangements or post agreements throughout, but I know when we went to Kentucky, without having to sort of commit to anything in particular, we were already there in Hawesville, we got we should post it. We got a great -- we gave the fire department great new engine.

Biotie called me personally, was all excited about how much that helped his community and his team. And so I think the answer is we should all be responsible to participate in the communities where we live, our employees live, and we're establishing businesses that we think are going to be legacy assets for generations. So that's the way we approach it.

Operator

Our next question comes from Brett Knoblauch with Cantor Fitzgerald.

Brett Knoblauch

Kind of on the backstop nature of this market. Congrats on getting the Google back stop. I'm curious what your guys' sense is for -- it feels like every data center project out there needs to backstop by 1 of the hyperscalers. What is their appetite to continuing to do this? And at what point does that appetite slowed down? And how are you guys thinking about a market in which an IG backstop won't be there?

Patrick Fleury

Yes. Brett, it's Patrick. Good question. So look, I mean, as you hear from everybody, not just TeraWulf, but semiconductor companies, hyperscale demand is robust, right? And adoption continues -- so look, I think what you're seeing from us again is version 2.0 of the wrap shows that hyperscalers are willing to put their balance sheet to work. But at some point, it's like whether you're learning to ride a bike with training wheels or you're raising your kids. When they get 18, they got to leave the nest and fly, right? And so that's -- you look at version 2.0 of our wrap, right, in the details, Google is there. And then once Anthropic achieves a certain investment-grade rating, that wrap falls away.

So I think you'll see that more commonly through all of the hypers because again, their job is to help the ecosystem grow. But then once it gets to a point where those companies can stand on their own and their strong investment grade, then that support should fall away and their balance sheet is left for their own use because they all need it.

So I think that's the way at least we're seeing the space mature, which is very logical and is something that we welcome.

Operator

Our next question comes from Darren Aftahi with Lucid Capital Markets.

Darren Paul Aftahi

First, just could we give an update on Chesapeake and where the regulatory approvals stand?

Paul Prager

.

Sure. Chesapeake, as you know, is under our purchase and sale agreement. We have most of the regulatory approvals in hand. I think there are a couple of outstanding, 1 of which is the Seco. But there are conditions precedent to close there that have yet to be fulfilled by the seller, and those conditions are important to us. And if those conditions are met, great. And if not, that would be a bummer. But we're not going to bring in any anything into our pipeline that we don't think will make a lot of sense, be able to be absolutely accretive and be a promising data center development.

So we love the state Governor Westmoore has put out some pretty, I think, progressive notions on how data centers can contribute. It's also in the middle of a transmission super highway. -- but there are conditions precedent to close. And when they're fulfilled, we'll be ready to roll. And if they are not fulfilled, well, we'll have to rethink that.

Darren Paul Aftahi

Great. And then just 1 more. Just kind of given the changes in the political landscape and who knows what assets were available and maybe a bit frozen in the future? Like what's your kind of view on the shape of lease rental rates over the next coming, call it, 12, 24 months?

Paul Prager

Listen, I got to focus on what's in my portfolio. And as I said in my prepared remarks, what we have today is a lot more valuable than it was last week. And I think 1 of the reasons for that is the tremendous demand there is in the market for partner now, for prompt power. So I could speak to the fact that the phone has been ringing off the hook with respect to what we have at Muskie. And so I think I don't know if leverage is the right word, but I feel really good about our side of the table in negotiations with potential customers with respect to where we will end up on that facility.

Equally, I think, 1 has to be thoughtful about where they develop. We always have, as you know, we've built energy infrastructure for the last 25 years. So we've always gone to places where we thought we could navigate the politics and the regulatory framework. That hasn't changed. And so I think you'll continue to see us do just that in the areas as we build out our pipeline.

Operator

We have reached the end of our question-and-answer session, which now concludes today's teleconference. You may disconnect your lines at this time. Thank you for your participation.