Hội nghị công bố kết quả kinh doanh quý 3/2026 của Keel Infrastructure (KEEL): Thanh khoản 780 triệu USD, mục tiêu cho thuê bị trì hoãn

Điểm nhấn chính

• Doanh thu quý 3/2026 giảm xuống 28 triệu USD so với mức 68 triệu USD cùng kỳ năm trước, chủ yếu do việc dừng hoạt động khai thác Bitcoin tại Mỹ trong quý 2 và giá Bitcoin trung bình thấp hơn.
• EBITDA điều chỉnh âm 27 triệu USD, so với mức dương 19 triệu USD trong quý 3/2025. Lỗ từ hoạt động liên tục mở rộng lên 62 triệu USD, bao gồm khoản lỗ phi tiền mặt chưa thực hiện trị giá 55 triệu USD liên quan đến quyền chọn mua có giới hạn (capped calls).
• Keel đã chấm dứt chiến lược tập trung vào Bitcoin tại Mỹ và bán 2.520 Bitcoin với giá 192 triệu USD trong quý. Tiền mặt không bị hạn chế sử dụng đạt khoảng 780 triệu USD tính đến ngày 4/11/2026.
• Ban lãnh đạo không còn kỳ vọng đạt được mục tiêu đề ra hồi tháng 5 là ký kết ba hợp đồng thuê vào cuối năm. Công ty sẽ không đưa ra định hướng mới về tiến độ ký hợp đồng thuê, với lý do chậm trễ trong cấp phép và tầm quan trọng của việc đảm bảo các điều khoản có khả năng thu xếp vốn.
• Danh mục dự án ngắn hạn công suất 478 MW của Keel đã ghi nhận bước tiến trong quý. Dự án Moses Lake đã khởi công xây dựng, trong khi địa điểm Sharon công suất 110 MW và dự án Panther Creek công suất 350 MW đang hoàn thiện các giấy phép môi trường cuối cùng tại Pennsylvania.
• Keel có 380 triệu USD cam kết vốn trước khi ký hợp đồng thuê cho đến năm 2028. Ban lãnh đạo cho biết thanh khoản hiện tại đủ để chi trả cho các khoản chi tiêu vốn cần thiết trước khi cho thuê và chi phí bán hàng, quản lý doanh nghiệp (SG&A) bằng tiền mặt cho đến năm 2028 mà không cần huy động vốn bên ngoài trước khi hợp đồng thuê được ký kết.

Dữ liệu tài chính quan trọng

Kết quả kinh doanh và vận hành

Moses Lake

Keel đã nhận được tất cả các phê duyệt cần thiết để bắt đầu xây dựng và đã khởi công trong quý 3. Ban lãnh đạo kỳ vọng Moses Lake sẽ là địa điểm đầu tiên của công ty được cấp điện vào năm 2027 và là nơi đầu tiên tạo ra doanh thu từ điện toán hiệu năng cao.

Dự án được tài trợ toàn bộ từ bảng cân đối kế toán của Keel và không yêu cầu hợp đồng thuê hay tài trợ dự án để hoàn thành. Các thiết bị quan trọng đã được mua, ký hợp đồng hoặc đang trong quá trình sản xuất trước khi bắt đầu xây dựng. Công ty cho biết địa điểm này có thể được nâng cấp từ hệ thống NVIDIA GB300 lên Vera Rubin với chi phí phát sinh không đáng kể.

Ban lãnh đạo mô tả Moses Lake đã sẵn sàng cho thuê mà không cần thêm giấy phép hoặc công việc kỹ thuật nào. Nhiều đối tác đang tiến hành thẩm định kỹ thuật và đàm phán thương mại, nhưng Keel từ chối đưa ra lộ trình ký kết hợp đồng thuê.

Sharon và Panther Creek

Dự án Sharon đã đảm bảo được nguồn điện 110 MW từ FirstEnergy. Giấy phép môi trường cấp bang cuối cùng của dự án đang ở dạng dự thảo tại Sở Bảo vệ Môi trường Pennsylvania. Ban lãnh đạo cho biết địa điểm này đang thu hút sự quan tâm từ các nhà cung cấp dịch vụ đám mây quy mô siêu lớn (hyperscaler) và các công ty AI đang phát triển.

Panther Creek đã đảm bảo được 350 MW điện từ PPL và có tiềm năng mở rộng lên 500 MW trở lên. Một trong ba giấy phép môi trường cấp bang còn lại đang ở dạng dự thảo, trong khi hai giấy phép còn lại đang trải qua quá trình đánh giá kỹ thuật tiêu chuẩn. Ban lãnh đạo cho biết các cuộc đàm phán tại Panther Creek có sự tham gia của một số bên mua công suất trung tâm dữ liệu lớn nhất.

Cả hai dự án tại Pennsylvania dự kiến sẽ cần tài trợ dự án. Keel đang ưu tiên các khách thuê có xếp hạng tín nhiệm ở mức đầu tư hoặc các đối tác được bảo lãnh tín dụng đạt mức đầu tư.

Danh mục mở rộng

Keel cho biết danh mục mở rộng của công ty đạt tổng cộng khoảng 2 GW, không bao gồm thương vụ mua lại tiềm năng tại Allentown. Công ty đang đánh giá một hạ tầng nhà xưởng trung tâm dữ liệu đã hoàn thiện cấp điện (powered shell) tại Allentown với 75 MW điện từ PPL, hướng tới mục tiêu năm 2027, tùy thuộc vào kết quả thẩm định và rà soát pháp lý.

Tại Scrubgrass, Keel đang theo đuổi công suất mở rộng lên tới 1,3 GW. Một nghiên cứu phụ tải chi tiết đang được tiến hành và thông tin cập nhật dự kiến sẽ có vào quý 1/2027. Tại Quebec, công ty đã nhận được các phê duyệt của địa phương cho một địa điểm 96 MW và đang chờ phê duyệt cuối cùng cấp bộ.

Hoạt động khai thác Bitcoin còn lại của Keel tại Canada có công suất khoảng 5–6 EH/s và tạo ra vài Bitcoin mỗi ngày. Ban lãnh đạo không có kế hoạch đầu tư thêm vào hoạt động khai thác và dự định tiếp tục vận hành các tài sản này cho đến khi khả năng chuyển đổi mục đích sử dụng được xác nhận.

Định hướng của ban lãnh đạo

• Keel không còn kỳ vọng sẽ ký được ba hợp đồng thuê trước cuối năm 2026 và sẽ không đưa ra định hướng điều chỉnh về mặt thời gian.
• Ban lãnh đạo kỳ vọng các trạm biến áp tại Moses Lake, Sharon và Panther Creek sẽ được đóng điện vào năm 2027.
• Thời điểm sẵn sàng đi vào vận hành tại Sharon và Panther Creek sẽ phụ thuộc vào việc ký kết hợp đồng thuê và thời điểm khởi công xây dựng. Ban lãnh đạo cho biết các khách hàng tiềm năng đang hướng tới việc bắt đầu tạo doanh thu vào đầu năm 2028, sau đó là tăng dần công suất trong suốt năm 2028.
• Thông tin cập nhật về các nghiên cứu phụ tải mở rộng tại Panther Creek và Scrubgrass dự kiến sẽ có vào quý 1/2027, muộn hơn mốc thời gian tháng 12 hoặc tháng 1 trước đó.
• Ban lãnh đạo cho biết Keel có đủ thanh khoản để tài trợ cho các khoản chi tiêu vốn cần thiết trước khi cho thuê và chi phí SG&A bằng tiền mặt cho đến năm 2028.
• Công ty hiện không dự kiến huy động vốn bên ngoài bằng nợ hoặc vốn cổ phần trước khi ký hợp đồng thuê.

Rủi ro và các vấn đề trọng yếu

• Keel vẫn chưa ký hợp đồng thuê trung tâm dữ liệu nào và đã rút lại mục tiêu hoàn tất ba hợp đồng thuê vào cuối năm.
• Khuôn khổ quy định mới về trung tâm dữ liệu của Pennsylvania đã loại bỏ các dự án khỏi chương trình cấp phép nhanh của bang, khiến tình trạng tồn đọng hồ sơ tại Sở Bảo vệ Môi trường (DEP) gia tăng. Ban lãnh đạo không thể đưa ra thời hạn hoàn thành chắc chắn cho các giấy phép môi trường còn lại.
• Lãi suất tăng cùng thị trường nợ trở nên khắt khe hơn có thể làm tăng chi phí tài trợ dự án. Keel dự định kết hợp việc thu xếp tài chính đồng thời với việc ký kết hợp đồng thuê để hạn chế rủi ro chênh lệch (basis risk).
• Chi phí xây dựng có sự khác biệt đáng kể tùy theo địa điểm, quy mô, thiết kế, thiết bị và yêu cầu về thời gian hoạt động liên tục (uptime). Ban lãnh đạo dự định phản ánh biến động chi phí vào hiệu quả tài chính của hợp đồng thuê, nhưng lưu ý rủi ro mất cân đối nếu hợp đồng thuê được ký trước khi các chi phí cơ bản được chốt chắc chắn.
• Các dự án lớn tại Pennsylvania đòi hỏi phải có tài trợ dự án và đối tác có uy tín tín dụng tốt. Do đó, sự hỗ trợ tín dụng từ khách thuê là điều kiện then chốt để có thể thu xếp vốn.
• Sự giám sát về mặt chính trị đối với việc phát triển trung tâm dữ liệu vẫn ở mức cao, dù ban lãnh đạo cho biết các dự án tại Pennsylvania của Keel phù hợp với lưới điện và các tiêu chuẩn cộng đồng của bang.

Nội dung nổi bật phần Hỏi & Đáp với chuyên viên phân tích

• Cấu trúc hợp đồng thuê: Ban lãnh đạo cho biết các cuộc đàm phán đang ngày càng cạnh tranh hơn khi các địa điểm của Keel đạt tiến triển và nguồn cung điện ngắn hạn thắt chặt. Yêu cầu của khách thuê có sự khác biệt theo quy mô, trong đó các công ty AI và đám mây mới nổi phù hợp hơn với Moses Lake, còn các bên mua công suất lớn tập trung vào Sharon và Panther Creek.
• Chi tiêu trước khi ký hợp đồng thuê: Khoảng một nửa trong số 380 triệu USD cam kết đến năm 2028 liên quan đến việc hoàn thiện Moses Lake. Phần lớn số tiền còn lại gắn liền với công tác xây dựng trạm biến áp và đường truyền cáp quang tại Sharon và Panther Creek. Toàn bộ chi phí xây dựng hoàn chỉnh chỉ dành cho các dự án đã ký hợp đồng thuê và đảm bảo được nguồn tài trợ.
• Quy trình phê duyệt mở rộng: Sau khi hoàn thành nghiên cứu phụ tải chi tiết, Keel dự kiến sẽ có khoảng ba tháng đàm phán hợp đồng dịch vụ năng lượng, bao gồm tiến độ cấp điện, tiền đặt cọc và thư tín dụng.
• Hình thức phát triển: Keel cởi mở với các hình thức cho thuê đất có sẵn nguồn điện (powered-land), nhà xưởng trung tâm dữ liệu có sẵn nguồn điện (powered shell) và dịch vụ cho thuê chỗ đặt máy chủ (colocation). Công ty không có kế hoạch theo đuổi mô hình kinh doanh cung cấp toàn bộ năng lực tính toán (full-compute).
• Mua sắm thiết bị: Các thiết bị có thời gian đặt hàng dài (long-lead) phục vụ đấu nối điện đã được đảm bảo tại cả ba địa điểm ngắn hạn. Đối với Sharon và Panther Creek, Keel đang đặt chỗ sản xuất thay vì mua các thiết bị trung tâm dữ liệu đặc thù cho từng khách thuê trước khi biết rõ yêu cầu của hợp đồng thuê.
• Chi phí điện: Ban lãnh đạo cho biết khách thuê hiện ưu tiên tính khả dụng của nguồn điện và tiến độ bàn giao hơn là giá điện. Chi phí tài trợ vốn là trọng tâm lớn hơn trong các cuộc đàm phán hiện tại.
• Tính độc quyền: Keel muốn tránh hoặc hạn chế đàm phán độc quyền để duy trì thế cạnh tranh, mặc dù ban lãnh đạo thừa nhận rằng quyền độc quyền trên phạm vi rộng hơn có thể là điều cần thiết khi làm việc với các đối tác lớn nhất.

Toàn văn biên bản cuộc gọi công bố kết quả kinh doanh

Toàn văn cuộc gọi công bố kết quả kinh doanh

Phần trình bày của ban lãnh đạo

Operator

Good day, and thank you for standing by. Welcome to the Keel Infrastructure Third Quarter 2026 Earnings Conference Call and Webcast. [Operator Instructions] Please be advised that today's conference is being recorded.

I would now like to hand the conference over to your first speaker today, Laine Yonker from Keel Investor Relations. Ma'am, please go ahead.

Laine Yonker

Thank you, and welcome to Keel Infrastructure's Third Quarter 2026 Conference Call. With me on the call today are Director and Chief Executive Officer, Ben Gagnon; and Chief Financial Officer, Jonathan Mir. Before we begin, please note this call is being webcast with an accompanying slide presentation. Today's press release and presentation can be accessed on our website under the Investors section.

Turning to Slide 2. I'd like to remind everyone that certain forward-looking statements will be made during this call and that future results could differ from those implied in this statement. The forward-looking information is based on certain assumptions and is subject to risks and uncertainties. I invite you to consult Keel's 10-Q for a complete list, which is available on our website and the SEC website.

Please note that references will be made to certain non-GAAP financial measures and therefore, may not be comparable to similar measures presented by other companies. We invite listeners to refer to today's press release and our filed 10-Q for definitions of the non-GAAP measures and the reconciliations to GAAP measures. Please note that all financial references are denominated in U.S. dollars, unless otherwise noted.

And now turning to Slide 3. It is my pleasure to turn the call over to Ben Gagnon, Director and Chief Executive Officer. Ben, please go ahead.

Benjamin Gagnon

Thank you, Laine, and good morning, everyone. I want to start today by addressing the macro elephant in the room. The last few months have been challenging for everyone. Rates are surging and the 10-year treasury just hit its highest level in more than 2 decades. Debt markets that were wide open 90 days ago are still functioning, but they are more selective and more expensive today. Equity markets have also been unforgiving and many names in our space have fallen more than 50% from their highs earlier this year. Data centers have become a political lightning rod, even though nearly every American relies on them every single day. And geopolitical tensions have added another layer of uncertainty to global and domestic energy markets. None of this is new to anyone on this call, and none of this is unique to Keel. We have nothing to hide. So let's face it directly.

When we named this company Keel, we said it keel is the backbone of a ship, unseen but essential. Its main purpose is to keep the ship upright and moving forward even in rough water. It's the reason a vessel can hold its course when the wind is blowing hard in the wrong direction. This quarter, the wind blew hard. The seas were rough, yet we held our course. And holding our course means doing what we said we would do. So let me hold us to that.

Last quarter, we told you Moses Lake would finish permitting and start building. It did and is under construction today. We told you we were highly confident in our remaining Pennsylvania permits, and that hasn't changed. At both sites, we have zoning and land development in place, and we're down to our final state environmental permits. They've taken longer than we expected, but to date, not a single permit application has been denied at any of our sites.

We also told you to expect an update on our expansion load studies at Panther Creek and Scrubgrass as early as December or January. These applications continue progressing positively, albeit at a slightly slower speed, and we now expect to give you that update in Q1 2027. If this time line changes, we will tell you directly. None of that is luck. It's the strength of our team keeping us steady and moving us forward even while we're sailing up wind.

Where we haven't delivered on what we told you is leasing. In May, we set a goal to sign 3 leases by year-end. And while we no longer expect to sign 3 this year, demand has never been higher, and we are not wavering on our strategy. All of the recent headlines have one key factor in common. They all further restrict supply of the primary bottleneck on growth across the industry, power. A project blocked here, a moratorium there, a permit paused, a financing pulled, a power plant delayed. In the first half of this year alone, it is estimated that roughly $200 billion of data center projects across the country were blocked or delayed. Every one of those headlines takes '27 and '28 power off the market and makes what is left like our current projects even more scarce and valuable. That's the supply side.

Now let me show you the demand side, in the words of the biggest buyers in the world. Turning to Slide 4. While supply keeps shrinking, demand hasn't blinked. In the last few weeks, the 4 largest buyers of data center capacity in the world, Microsoft, Alphabet, Meta and Amazon all reported. And I'd encourage every investor on this call to read those filings because they reinforce the demand thesis better than we can. Together, they are on track to spend more than $800 billion on CapEx this year, and all 4 have indicated that these investments aren't slowing down anytime soon.

Microsoft, Alphabet and Amazon combined contract backlog has grown by about 2/3 since January to roughly $1.8 trillion. That is demand customers have already signed, waiting on capacity. And all 4 have told their investors in their own words, they don't have enough capacity. Amazon said, we will still not have enough capacity to meet all the demand we have in '26, and I believe this demand will also be true in '27. In fact, the demand we already have for '28 is striking.

Microsoft said customer demand continues to exceed available supply. Last quarter, Microsoft actually spent less than it planned because third-party capacity showed up late. Meta said there's just nowhere near enough compute for all the demand. And Alphabet has started renting third-party compute as what it calls a bridging strategy while it builds up more of its own. NVIDIA says it sees at least $1 trillion of demand for its Blackwells and Rubin platforms through '27, and Morgan Stanley estimates a 70 gigawatt shortfall in U.S. data center power through '28. We are not afraid of this market. To the contrary, it is exactly the market we laid out in our thesis nearly 2 years ago and have been preparing for.

Here's what I think matters most. Between them, the hyperscalers have signed roughly $1 trillion of data center leases that haven't even started yet with much of the power coming online in '28 or later. We continue to see demand exceed supply for ready for service dates through the end of '28 because hyperscalers, neoclouds and many enterprises have already committed to the chips for '27 and '28. And what they are short of is power that is available on the same time line.

So ask yourself this, if the biggest, best capitalized builders in the world don't have enough '27 and '28 power, who does? Keel. While the supply of near-term power is shrinking across the industry, at Keel, our '27 and '28 power is further advanced, more secure and closer to delivery than it was 90 days ago, and our continuous inbound customer demand shows it.

Now turning to Slide 5. On August 18, 1 week after our last call, Governor Shapiro signed an executive order, making Pennsylvania's grid standards for data center development mandatory. The grid standards require data center developers in Pennsylvania to pay their own way for new power capacity and engage communities early and publicly, deliver community and workforce benefits beyond what the law requires and commit to environmental protections with ongoing monitoring. Put simply, pay your own way, be open with and get back to communities, hire locally where possible and protect the land, water and air.

These standards benefit everyone, and we fully endorse the order and the need for responsible data center development. And this is an easy endorsement because the grid standards are largely in line with Keel standards and the work we've been doing since we established our presence in Pennsylvania. So we were able to adapt. But the implementation of the executive order also still brought a lot of new real work for the team. It took a lot of engagement with every relevant regulator and numerous government offices and agencies to first fully understand the executive order and its impact across Keel projects, and then additional engineering by our team and by our potential customers' teams to design solutions. While this took time, this is largely behind us, and we now have clarity in engineered solutions that work.

For investors, here's what the implementation of the order means in practice. One, the permits we have in front of the DEP today are not subject to the order's new requirements, but the order also took data centers out of the state's fast-track permitting program. In plain English, data center permits are no longer eligible for priority review. They move through the DEP standard process. While this does not directly impact our sites as the review of our remaining permits is not dependent on the fast-track program, it does create a larger backlog for the DEP.

Two, backup power. Every one of our Pennsylvania sites was always going to need a backup solution. This was always in the budget, and this was always in the plan. What change is the type of backup. Over the past 3 months, we have engineered a solution, a component of which is centered on battery storage that works for our customers, works for Keel and complies with the order. And most importantly, we don't expect it to meaningfully change the economics or the timing of these projects for tenants.

Three, our planned power delivery schedules under our ESAs have not changed. We expect both Sharon and Panther Creek substations to be energized in 2027. At this stage, RFS dates are directly linked to lease signings and will be a function of when we sign and break ground. Customers understand this and want to get megawatts online as soon as possible with revenues beginning in early 2028 and each site's capacity to ramp through 2028.

But this all misses the spirit of the executive order. The spirit of the order was about communities, and this is an area where Keel leads the pack. Community pushback against data centers comes down to 3 things: a lack of transparency, a lack of safe for the people who live there and no real commitments once the permits are in hand. Across our industry, most developers answer that with a benefits package, and that's where they stop.

We call our answer the Keel standard, and we're backing it with 3 concrete commitments. First, a community benefit agreement or CBA. We're negotiating our first now with the Nesquehoning Borough Council for Panther Creek, and we started that work before Pennsylvania required it. Sharon and Scrubgrass are up next. Our CBA is market-leading and is designed to scale with increasing power on site, which lets the community benefit alongside our success.

Second, a formal way for residents to raise concerns. We're building an agreement process that gives every resident a local point of contact, defined response times and a route to appeal with oversight at the Board level. It launches in Nesquehoning first then at every one of our sites. This is not a hotline or generic e-mail. This is opening of all avenues of communication to the community to tell us what they're seeing, what it is that they're hearing and how they're feeling. With a company mandate behind it to not simply be a good corporate citizen, but to be a good member of the community.

Third, measuring before we build. We've engaged third-party consultants to survey Nesquehoning's residents through a community baseline assessment. Before Panther Creek is built, listening and documenting the community's views on housing, employment, education, income, health, local services and a number of other areas. In November, they're sitting down with residents and in December, every household in the Borough will receive a confidential survey. We'll share the highlights with the Borough Council in '27, repeat the assessment over time and do the same in Sharon and Scrubgrass. That way, we can track what our projects actually change, how we're impacting the community and ultimately make decisions based on evidence rather than assumptions.

Across our industry, most community commitments stop at a benefits package. Few pair it was a formal way for residents to raise concerns. Fewer still measure conditions before and after they build. We're doing all 3. That is what staying accountable after approval looks like in practice. And you can already see it on the ground. In Nesquehoning, through our investment in Panther Valley School District, we fund a full-time community school coordinator, support the after-school programs, fund the school bus system and fund and staff food pantries, which together serve about 500 families and 2,000 individuals monthly. Our team is out there in the community showing up because there are people.

We have also held more than 20 town halls or informational sessions across Nesquehoning, Sharon and Scrubgrass. These sessions have been publicized, socialized and the towns have shown up. More than 500 people attended the events, and they heard directly from our senior team how they explain data centers, explain Keel and walk through what our projects mean for them. This isn't charity and it isn't public relations. It's being a good neighbor and being a good neighbor is good business.

Infrastructure projects backed by strong community support get built and survive the test of time. And so because of these standards, both Panther Creek and Sharon secured their zoning approvals before the executive order was even discussed. And when the order was signed, we didn't actually have to change a single thing about how we engage with the community. Frankly, this is one of the most rewarding parts of the job, and we did not need an executive order to tell us to do the right thing. We have been doing it since Day 1. And our team regularly receives feedback from all levels of government about our community engagement and the benefits of conducting ourselves the way we do. What we did have to do, though, was work through the new requirements with regulators, our engineers and our customers.

And that brings me to leasing because I know it's top of mind. In May, we set a goal to sign 3 leases by year-end. The new regulatory framework in Pennsylvania has moved our permitting and commercial time lines back by months, and we no longer expect to sign 3 leases this year. Yet we have not wavered on our commercial strategy. We're seeing unprecedented demand for our sites and in an industry where some of the largest developers are facing delays of a year or more, a delay measured in months is one our customers understand. To date, no potential customer has walked away for this, and new parties keep coming to the table. We strongly believe that securing strong leases is not a question of if, but when. And as I've said many times before, a lease is not a trophy for a press release. It's a 15-year commitment, and we will not sign the wrong lease at the wrong price just to hit a date.

Conditions in the financial markets have been far more volatile than conditions on the ground. And because we haven't signed a lease yet, every lease we sign will be priced at prevailing market terms and price to the tenant-specific requirements. What matters is that these projects are priced and judged on a yield on cost. Cost for newer technology, materials and labor may rise or fall and so may the cost of capital. When they do, we plan to reflect these in the lease economics so that to minimize the risk to our shareholders.

So here is our commitment going forward. We won't give you a date for a lease. We will give you clear, transparent updates on every site as development progresses. And when we sign, it will be the right lease with the right customer at the right economics, including the cost of financing it. That is how we stay the course and maximize long-term value for shareholders.

Turning to Slide 6. Now to the sites starting at Moses Lake. During Q3, we received every approval we needed to begin construction, and we broke ground. Moses Lake remains on track to be our first site energized in '27 and our first site to generate HPC revenues. It's modular, it's built to suit and the path to energization is shorter than a traditional build. Everything critical was already bought, contracted or in manufacturing before we even broke ground, and we have engineered a path for Moses Lake to move from GB300s to Vera Rubin at a marginal incremental cost, giving tenants more flexibility and making Keel an even more valuable solution.

Moses Lake is the only site fully funded from our balance sheet. Our goal there, as everywhere, would be to announce the lease and its financing together. However, we don't need a lease or project financing to finish Moses Lake, which means we are under no pressure to finance it on disadvantageous terms. Near-term power in the Pacific Northwest is as scarce as ever, and we have multiple parties engaged at Moses Lake that need megawatts now. Moses Lake is our smallest site, less than 4% of our near-term megawatts and a site this size better fits a fast-growing AI or cloud company rather than a hyperscaler. We will hold it to the same standard as every other site, a structure that protects our returns, credit support that matches the customer and financing that matches the lease. Smaller on, we know a Moses Lake lease will be read as a signal of how we execute, and we are treating it that way.

Turning to Slide 7. As I just mentioned, commercial tension continues to build across our Pennsylvania sites. Over the past 3 months, we have completed a significant amount of technical diligence with multiple prospective tenants, advanced LOI negotiations with multiple parties at both sites and most importantly, kept our power delivery schedules under the ESAs intact despite a new regulatory framework.

Let me start with Sharon. Sharon is 110 megawatts of secured power FirstEnergy. It is our most advanced Pennsylvania site down to its final state environmental permit, which the DEP has already issued and draft. At 110 megawatts, Sharon's scale fits a wide range of customers from hyperscalers to fast-growing AI companies, and that's the mix we're seeing at the table. It's also why credit structure matters so much there. Pairing a fast-growing AI company with investment-grade credit support is what turns a strong customer into a financeable long-term lease.

Panther Creek is our flagship. It has 350 megawatts of secured power of PPL with potential room to grow to 500 megawatts or more in one of the most sought-after AI corridors in the country. That scale is why the conversations at Panther Creek are with the very largest buyers of capacity. Zoning and land development are in place, and we're working through our final state environmental permits, which 1 of the 3 is already in draft and the other 2 are in standard technical review with the DEP. For a campus this size, the counterparty matters more than anything. The bigger the customer, the more they value certainty on power, permits and delivery. And with every permit that lands, Panther Creek becomes something only a handful of sites in America can offer, '27 and '28 power at scale in PJM with room to grow for years.

Unlike Moses Lake, both Pennsylvania sites are large enough to require project financing. And that is why at both Sharon and Panther Creek, we have been disciplined and will continue to be selective in all of our potential tenant discussions. We are focusing on signing with an investment-grade tenant or a counterparty backed by an investment-grade wrap.

Turning to Slide 8. Now I want to step back because I think this is the most important slide in the deck. There are a handful of things I believe this company should be judged by today. How many megawatts do we have? How secure are they? How confident are we in their time lines? How far along are they in permitting, design and construction? And how much demand is there for them. By every one of those measures, Keel is better off today than it was 90 days ago.

Start with the near term. Our 478 megawatts at Moses Lake, Sharon and Panther Creek advanced on every front. Moses Lake went from approved under construction. Pennsylvania is down to its final environmental permits and our designs are updated for the new grid framework. And we have advanced leasing discussions across all 3 with new potential tenants still coming to the table.

Then look at what's coming behind them. Our expansion capacity represents the opportunity for us to quadruple our currently secured capacity. At Panther Creek, we are working with PPL and Blackstone Infrastructure joint venture to support the build-out of natural gas generation and expand the campus beyond 350 megawatts towards 500 megawatts or more. This scale is a key selling point to the world's most discerning tenants.

At Scrubgrass, we are progressing well on securing our expansion capacity of up to 1.3 gigawatts, and we did it the Keel way. Land secured a detailed load study underway, horizontal lateral planning progressing with our development partner and 5 community sessions held before filing a single permit. And in Quebec, we have received the local approvals required in order to advance our 96-megawatt site and are now awaiting final ministerial approval.

You may have also seen recent press reports about a site in Allentown, Pennsylvania. As we have previously discussed, we continue looking for opportunities to grow our portfolio and create value for shareholders and customers. This site in Allentown is a potential opportunity that we are exploring that would allow us to acquire a fully constructed power shell and 75 megawatts of 27 PPL power. We are conducting detailed diligence with the current owners and local and state regulators, and we'll keep our investors updated as the process progresses.

Capital discipline guides how we grow. We are pursuing this new site only as customer demand is clear and we understand the regulatory requirements and the time line to ready for service. Not including Allentown, our expansion portfolio is about 2 gigawatts total. And here's what excites me. Every quarter, more of these megawatts climb the ladder from potential to secured, from secured to permitted, from permitted to designed, from design to under construction and next, from leased to energized. That is how real value gets built in this business, not by announcing megawatts, but by maturing them. And every project in our industry that slips takes more '27 and '28 power off the table, and that makes the power that does get delivered more valuable. While supply keeps shrinking, demand keeps climbing, and we intend to be one of the companies that delivers.

Turning to Slide 9. Let me bring this all together. This was a tough quarter for our sector. I said it at the top, and I'm not going to walk it back now. The water is rough and it's rough for everyone. But here's what actually happened inside Keel this quarter. We broke ground at Moses Lake. We got Pennsylvania down to its final environmental permits under what the governor calls the strictest guardrails in the nation without a single application denied. We set the industry standard for how we engage with our communities as demonstrated by our community benefits agreement, and we've kept every potential customer at the table while new ones keep coming. And our pipeline is more secure and further along than it was 90 days ago.

Things took longer, but nothing was canceled and demand is higher today than there's ever been with increasingly fewer options for customers to grow in '27 and '28. The 4 largest buyers of compute in the world just told their investors, they can't get enough capacity, and we are building exactly what they're missing.

I'll tell you this, the market is proving out the exact thesis our strategic pivot was built on 2 years ago. We anticipated the supply-demand imbalance. We prepared for it. Now it is here. And when supply keeps getting scarcer and demand keeps climbing, the companies holding real near-term power in the right places and developing it in the right way are the ones that will win. That is Keel. We've built this company for exactly this moment, and we are not going to waste it, and we have the balance sheet to see it through.

Jonathan will now take us through it. Turning to Slide 10.

Jonathan Mir

Thank you, Ben, and good morning, everyone. Before talking about our capital allocation and financing strategy, I'll review our third quarter financial results. Turning to Slide 11. For the third quarter of 2026, revenue was $28 million compared to $68 million in the third quarter of 2025. The decrease was largely due to the decommissioning of U.S. Bitcoin mining in the second quarter of this year as well as a decline in average Bitcoin price.

Operating loss for the quarter was $8 million compared to an operating loss of $18 million in the prior year period. And loss from continuing operations was $62 million or $0.10 per share compared to a loss from continuing operations of $11 million in the third quarter of 2025. This change in loss from continuing operations was driven primarily by a $55 million noncash unrealized loss on the fair value of the company's capped calls associated with the 2025 convertible notes offering.

Adjusted EBITDA for the quarter was negative $27 million compared to $19 million in the prior year period, reflecting the previously mentioned wind down of Bitcoin mining operations as well as a decline in average Bitcoin price. For the first 3 quarters of 2026, our cash SG&A has averaged $24 million per quarter. On Bitcoin, we have met our commitments to both substantively exit Bitcoin mining and sell our Bitcoin holdings. During the third quarter, we sold 2,520 Bitcoin for $192 million in proceeds, executing the sales during periods of strength. As of September 30, our Bitcoin holdings were nominal.

In Q2 2026, we decommissioned all of our U.S. mining operations. Only our Canadian Bitcoin mining operations remain active, generating a few Bitcoin a day. We sell production on a regular basis to support our operations and investments. We continue to exclude our small remaining Bitcoin operations in Canada from any liquidity planning. As of November 4, 2026, our liquidity was approximately $780 million, comprised of unrestricted cash. We have an additional $53 million in restricted cash acting as collateral for various letters of credit that we have posted for expansion capacity.

Turning to Slide 12. As we discuss each quarter, the 3 pillars of our financial strategy are capital allocation, capital formation and capital structure. This strategy informs important operating principles as we look to maximize long-term risk-adjusted shareholder value. First, we do not build speculatively. Second, we value strong liquidity reserves. And third, we will use prudent amounts of leverage in financing our projects and lock in financing terms as closely as possible to lease signing in order to avoid basis risk from trying to time markets. These are first principles of infrastructure investing, and Keel is an infrastructure company.

As detailed in our 10-Q, our current pre-lease capital commitments are $380 million in total through the end of 2028. This spending includes items such as engineering agreements, procurement and manufacturing and site management. We will selectively invest to ensure we can move at the speed our customers want. As a reminder, these pre-lease investments taking place now will be part of our equity contribution in support of our eventual project financings. The dollars we spent now represent avoided holdco level debt and/or equity needs at a later time.

Additionally, proceeds from a project financing at Moses Lake will allow us to recycle capital into the equity required for Sharon and Panther Creek. Every dollar we raise at Moses Lake eliminates the need for dollar of outside equity. This type of capital rotation is a core part of our strategy. We will continually look across our assets for opportunities to use their value to both reduce or avoid external equity financing and ensure balance sheet strength.

As has been the case, we have more than enough liquidity to fund all of our capital expenditures required before leasing and construction as well as our cash SG&A through 2028. We will be selective in respect of all spending ahead of leasing and construction to ensure we maintain the benefits of a strong liquidity position. We currently do not see the need for any external debt or equity financing in advance of signing a lease. In highly volatile markets, it is incumbent on us to act prudently and raise capital when we see an alignment of attractive use of proceeds and reasonable financial terms as was the case this summer and then use that capital in a transparent manner for the benefit of shareholders.

In this regard, one question we receive from shareholders is how we govern and prioritize our pre-lease spending. We do not fund projects based on speculation without predefined plans, limits and without confidence in value creation. We fund projects across what we call 3 stage gates. Each gate has a defined scope, a defined spend, and a defined set of conditions that must be met before we move to the next one. That structure is what allows us to advance and derisk our sites at the pace our customers require without committing shareholder capital ahead of the information that justifies it.

Stage 1 is securing power and permits. It's the smallest spend, and it creates the most value per dollar because it turns an energy position into a permitted buildable asset. Stage 2 is locking in long lead time equipment that sets the delivery date, such as transformers, substations and fiber as permitting and approvals near completion. These investments make our ready for service dates real to customers. Stage 3 is the full build, which is where the bulk of the capital is spent. We only commit that capital against a signed lease when the economics are contracted and we have secured project financing that locks in an attractive return for our shareholders.

We fund Stages 1 and 2 from our balance sheet, and we fund Stage 3 with project financing when a lease is signed. Capital goes in as risk comes out. In respect of project financing, we believe the best outcome is to lock in financing economics at the time of lease signing and minimize basis risk. We will not attempt to time the market. Our objective is to lock in a spread for our investors that delivers attractive returns. We are keenly aware of the sensitivities in the economic model of our business and the current financing environment. We speak almost daily with our financial advisers to ensure we understand real time the likely financing terms associated with any specific customer's potential lease, and this information is a key component of our evaluative process.

Moses Lake has been the one deliberate exception to this strategy. It's small, it's modular, and it puts our first megawatts on the grid earlier in 2027. Again, once the site is leased, proceeds from its project financing will be recycled into the equity requirements for Sharon and Panther Creek. Because of its small size, we expect that Moses Lake will be leased to a customer that is smaller than those we are discussing with Sharon and Panther Creek. And again, because of size, we expect to raise financing via private credit solution rather than via the high-yield bond market. We believe that our lease pricing at Moses Lake will provide reasonable risk-adjusted returns even in the current financing market.

We reiterate the commitment made on every earnings call. Whenever we raise external capital, we will be clear on the uses of that capital and why we believe the associated long-term risk-adjusted returns create value for shareholders. The stage-gate structure is the process we use to hold ourselves to that commitment. Capital goes in as risk comes out and not before. We do not anticipate the need for any debt or equity financing before signing a lease.

As we have demonstrated over hundreds of investor calls, when investors ask for additional explanation or clarity on any aspect of our business, we'll provide that in our discussions and continue our practice of having Ben and me readily accessible to our shareholders.

I'll leave you with the following thoughts. Keel has many paths to success. We have multiple sites, we have multiple potential customers. We have multiple ways to structure leases and financing. We are not dependent on any single project. We operate in geographies that are supportive of responsible data center development. We have kept our infrastructure-only business model simple and easy for investors to understand, secure power, advance it through permitting and construction, lease it to creditworthy customers and finance it on the strength of those leases. And the time of volatility, we're focused on the fundamentals and those matters within our control with the consistency of a strong balance sheet, a disciplined strategy and plans that are easy for investors to understand.

And with that, operator, let's open the line for questions.

Operator

[Operator Instructions] And wow we're going to take our first question. And it comes from the line of Mike Grondahl from Northland.

Phần hỏi đáp

Mike Grondahl

In your discussion or discussions with tenants, can you kind of talk about how you're seeing lease structures and terms evolving over the last 90 or 180 days?

Benjamin Gagnon

Mike, yes, absolutely. It's a great question. The terms that we're looking at across the sites do vary quite a bit based Moses Lake, Sharon versus Panther Creek. I think generally speaking, the tension tends to be building across all 3 sites because we continue to advance the sites and bring them closer and closer to the time line. The sensitivities I think that people have right now are increasingly focused on permits, community and financing, obviously. The permits and the community side of things, I think we've given our potential tenants great confidence in our path and our success and our continued path forward.

And in terms of the financing, that's something that we're working through with all the various counterparties. And we're quite selective in terms of the various counterparties that we are working with and speaking with because we're only focused on good quality counterparties that are going to be financeable. But generally speaking, I think the terms are getting a little bit more competitive and a little bit more tense in the negotiations as we get closer and closer to our commissioning dates and most of the market seems to be -- most of the market supply seems to be shrinking.

Mike Grondahl

Got it. And then maybe just as a follow-up to that. On a relative basis, can you kind of rank your 3 sites just in terms of demand at each one?

Benjamin Gagnon

So the way you think about kind of relative demand amongst the customers per site, it really depends on the scale. And so what we spoke to in the earnings call a few months ago is that the kind of counterparties at Moses Lake and Panther Creek are wholly different. While at Moses Lake, it's going to be more of a -- probably more of an emerging company. At Panther Creek, there's only a few buyers in the world who could afford that scale and are really interested in that capacity. And so when we're speaking with the biggest developers in the world, the biggest buyers of compute, their focus is Panther and Sharon. And then when we're looking at more of the neoclouds or emerging companies, language labs, those sorts of things, their priority is either going to be on Moses or Sharon depending on their scale and secondarily, their time to energy requirement.

Operator

Now we're going to take our next question. And it comes from the line of Greg Lewis from BTIG.

Gregory Lewis

It's been a long week, guys. I did have a question. I really appreciated the way you laid out kind of how you advance the project with some initial CapEx and then we kind of wait to kind of wait for the lease before we start spending. I guess as I think about other existing sites that are -- that we're working through developing. Any kind of sense for what that CapEx to get it from to Stage 2? Is there any way to kind of think about what that CapEx is going to look like over the next 12 months?

Jonathan Mir

Yes. So over the next -- so between now and the end of 2028, we have capital commitments totaling about $380 million. About half of that is related to finishing Moses Creek. And then most of the rest is related to substation work and fiber work at Sharon and Panther Creek, both. In terms of the Stage 1 development costs, those are very modest. Those are -- it tends to be people and a key space, folks in conference rooms. So by the time we get to the end of a Stage Gate 2, setting aside the decision to purchase long lead time equipment, it would take quite a bit of effort to spend $1 million a megawatt to have a site ready to go. So the dollars in Stages 1 and 2 are quite small compared to the overall project construction cost. And we are very careful in how we think about allocating those dollars across longer-term projects like Scrubgrass, for example.

Gregory Lewis

Okay. Super helpful for that. And then just as I think about the -- I mean, obviously, we're looking to expand power at all the site -- well not all the sites, a few of the sites. Could you walk us through the process? Obviously, we're doing the detailed load study that's ongoing at Panther Creek. But kind of like post the detailed load study, then what does that process look like for final approval of that extra 150?

Benjamin Gagnon

Yes, happy to go through that. So once the detailed load study is complete, that basically confirms all the engineering requirements, the financial requirements, the timing for delivery of the power by what estimated kind of date. Then what happens is there's going to be a negotiation period usually around maybe 3 months to negotiate the ESA. And within the ESA, we'll work through the final negotiations on power delivery schedule, upfront requirements in terms of initial deposits, letter of credits and everything else that's needed to fund the infrastructure and that'll be about 3 months.

Operator

And the next question comes from the line of Michael Colonnese from H.C. Wainwright.

Michael Colonnese

First one for me, I was hoping to get a little bit more color around the expected timing of securing the remaining permits at Sharon and Panther Creek. And is it safe to say that those milestones could be considered the closing conditions for you securing contracts with prospective tenants?

Benjamin Gagnon

Mike, so in terms of the remaining schedule for the permits, at Sharon, we just are down to our last final environmental permit, as I said on the call, and that's in the draft form. And draft form effectively means that they've done extensive review. They effectively think that this is likely going to be approved and they intend to approve it, but there's still a review process and a window that they need to work before it's final and before it's actually done. And they could always revise and change and could extend that a little bit should they need to. But we would expect that once things are progressing at this stage that they're likely going to come. It's just a matter of time.

When it comes to Panther Creek, we have 3 remaining, one in draft and 2 in technical review. So everything has been submitted, Everything is being processed. Everything is moving forward. We've given no red flags or indication that any permits are going to have issues or faults, but it's a little hard to guide on the time line now given the backlog at the DEP, and that's kind of a changing condition in real time. I think that the positive thing here is that we were so far advanced on our permits before the executive order came out, that the impact on us is pretty minimal because we were down to our final permits already, and most of those were already in process.

Michael Colonnese

Very helpful color, Ben. Appreciate that. And just a quick one on the mining business. Just remind us where the hash rate stands today and an update around the expected wind down of mining from here?

Benjamin Gagnon

It's about 5 or 6 exahash, Mike. It's not very meaningful. It's all in Quebec, and that generates a couple of Bitcoins a day. It still generates positive free cash flow, but it's pretty marginal, and we don't include it in our liquidity forecast planning, as Jonathan mentioned on the call. We would expect that that's going to continue to get smaller as we get the approvals from Quebec that we're hoping to get and work through over the coming months and quarters. And until we have a clear path forward on the conversion of the Quebec assets, we intend to keep running the Bitcoin mining, but we do not plan to invest further in it. No more dollars are going to be spent to buy more miners or upgrade. We're just going to continue running it until we have confirmed the conversion or the convertibility of those sites.

Operator

Now we're going to take our next question. And it comes from the line of Michael Donovan from Compass Point.

Michael Donovan

I see you increased contracted power at Baie-Comeau to 30 megawatts from '22. How are you thinking about developing the site for HPC? What additional approvals would be needed? And how might it delivery time line compare with Sherbrooke?

Benjamin Gagnon

Yes. So the path that we have in Quebec is we're following our Sherbrooke playbook. So that was our lowest identified hanging fruit. It's the community where we have the best relations, the most scale and the best potential location for an HPC site. So that's our first project that we're moving forward with. And we -- as we have communicated previously, we've got all the local approvals, and we're just waiting on the final ministerial sign-off.

When that happens, we're waiting for that to happen. And then what we do is we'd replicate the same playbook from Sherbrooke across Baie-Comeau. And we had an opportunity to effectively acquire a small expansion for no cost, and we could acquire the right to get an additional 8 megawatts at the Baie-Comeau site for effectively no dollars out the door. So it was a very interesting option for us, assuming that we get the Sherbrooke playbook done. We have a slightly more compelling site at Baie-Comeau to develop.

Michael Donovan

Appreciate that. And then to follow up on Mike's question, as deal structures continue to evolve, would you consider a powered land lease at any of your sites? Or are you focused on delivering the full facility?

Benjamin Gagnon

Yes. We'd be open to doing a powered land lease. We have looked at powered land, powered shell, full colo, as well as the cloud model. And we think that there's a fit for each one of these business models for different companies. We specifically have chosen not to pursue the highest rung of that ladder, which is the full compute. But going powered shell, powered land or colo, all of those are interesting options that are on the table.

Operator

[Operator Instructions] And now we're going to take our next question. And the question comes from the line of Brian Kinstlinger from Alliance Global Partners.

Brian Kinstlinger

As it relates to Moses Lake, with permitting done in September and unprecedented demand, what has held back moving forward with the lease there? And should we expect that before year-end? Or is that more of a first quarter 2027 event?

Benjamin Gagnon

Brian, thanks for the question. As we said on the call, we're not going to provide any sort of guidance on lease timings for any of the sites from here on out. But the site has a tremendous amount of interest and demand and the competitive tension there at that site. And one of the things that's been very helpful for us at Moses Lake is that because we are funding this site off of our balance sheet, we've been able to continue moving forward with the site, continue to develop every day we move forward, every beam of steel that goes up continues to derisk the project and create more value for the end customers. And so that's helping us to maintain a very high level of tension in Moses Lake, and we're highly confident that we're going to get a very good deal done there. But we don't want to provide or we're not going to be providing any sort of expectations as to the timing of that.

Brian Kinstlinger

My follow-up, just to be clear, it's lease ready. There's nothing you need to do to get it prepared for lease at this point. Is that right?

Benjamin Gagnon

We do not require any other permits or any other engineering work to get a lease done. And we're under very extensive technical diligence and negotiations with multiple counterparties at this point during the commercial process.

Operator

Now we're going to take our next question. And the question comes from the line of Bill Papanastasiou from Chardan Capital Markets.

Bill Papanastasiou

Apologies in advance. I was disconnected earlier when I raised my hand. First, maybe you can provide an update on how the team is navigating the supply chain environment for long lead time equipment, just given the level of demand for data centers and restricted supply.

Benjamin Gagnon

Bill, it's a great question. So what we've done is kind of a few things. First, we secured all the long lead time equipment that we need for the power connection, right? So everything that we need for the substation transformers and building out the actual powered land piece of it, we have all of those long lead time items secured across all 3 of our sites. At Moses Lake, we obviously have everything secured and contracted and on the way, a lot of it already received and in hand.

And then for the longer lead time item equipment for building out the data center at Sharon and Panther Creek, what we don't want to do is actually lock ourselves into long lead time item equipment purchases that may not ultimately be what the end tenant wants. If you think about kind of time line to RFS is a very simplified Gantt chart with 3 buckets. There's obviously permitting and planning. There's another bucket, which would be the leasing. And then the third bucket would be the long lead time item equipment.

And given this stage in the permitting process across Panther Creek, the critical path is going to be leasing. And the way that we can keep the critical path on leasing is because instead of committing to long lead time item equipment, what we try and do is we try and commit to manufacturing slots, which help us to minimize the CapEx or minimize the capital that we commit in terms of deposits to secure time lines and also preserve the flexibility for us to modify the actual equipment that we're purchasing for the end tenant based on their ending needs. So this is a way for us to try and be very capital efficient while we maintain our time lines and maintain our flexibility through the negotiations.

Bill Papanastasiou

Great. Appreciate that. And then secondly, we've heard that there's been a few major developments in the Canadian data center market, particularly with Anthropic entering the country and what is rumored to be a second major deployment. Curious if you've seen any uptick in demand in this geography and how you think Frontier AI lab exposure in Canada will develop over time?

Benjamin Gagnon

So in Canada, it's a very, very different market than the United States. I don't think we've seen as many frontier labs in Canada just because it has been harder to develop new data center capacity in Canada. When you look at actually how many GPUs are plugged in around the country, there are numerous data centers in the United States that have more GPUs plugged in than the entirety of Canada for HPC and AI purposes. So it's relatively small scale at this point.

I think the really interesting opportunity with Canada, something we've talked about a few different times, is that there is this legislative drive for data sovereignty and compute sovereignty in Canada, both nationally and provincially. And because there is such a stricter bottleneck on growth and supply of data centers and GPUs in Canada and you have the legislative capture to keep that compute demand local, there tends to be significantly higher premiums associated with compute in Canada than for the same compute in the United States. And so it provides a great opportunity for many companies to be coming into Canada. I don't think it's necessarily limited to or maybe even focused on language labs. I think there's probably some really interesting -- maybe more interesting opportunities with the major Canadian corporations and the government than there may be with the language labs who have the flexibility to go anywhere.

Operator

[Operator Instructions] And now we're going to take our next question. And the question comes from the line of Nick Giles from B. Riley Securities.

Nick Giles

I just wanted to ask about your appetite to enter into exclusivity, if that's something you'd consider? Or should we expect that you really would just want to sign a lease? And then if you prefer to just sign a lease and not go into exclusivity, would you consider some sort of expansion option or ROFR on additional capacity Panther Creek comes to mind?

Benjamin Gagnon

Thanks, Nick. Yes, it's a great question. At the various sites, we do our best to maintain as much competitive tension in the process by not signing exclusivities or by signing limited exclusivities that give us the ability to continue negotiations and marketing the site. That's our preferred strategy because obviously, it gives us the best probably adjusted outcome, right? It keeps as many people at the table. It keeps everyone focused and it keeps everyone kind of on their toes that if they don't lock up the lease, then somebody else will.

This is something that, for example, at Moses Lake has really helped us to drive the progress forward with various tenants because they know that they're in a competition to get that power and to get those time lines. So as much as possible, that's going to be our intent. Now that's not always going to be possible, Nick, especially with some of the biggest players, but we're going to do our best to keep exclusivity limited, tension as high as possible and to keep as many people at the table.

Nick Giles

Makes a lot of sense. I appreciate that, Ben. And maybe just a follow-up on electricity costs. You don't hear as much about electricity costs as we did in the Bitcoin mining days. But I was wondering if you could just touch on kind of where your sites sit from an electricity cost perspective and to what extent tenants really have sensitivity to electricity costs in your discussions?

Benjamin Gagnon

Thanks, Nick. Frankly, it really does not come up very much. The electricity costs are really quite small relative to the operating margins of the GPUs themselves. And everyone is so focused on time line to power that there's very -- there's way less sensitivity around the price of the electricity itself. It's always around the time line and the risk of deployment and the risk to deploy on time that everyone is focusing on first and foremost.

We probably will see energy cost considerations come in, in the future. But I would think that the bottleneck on growth would need to be alleviated. I would think that supply would need to overtake demand for those economic sensitivities to start being the prevailing thing that people are focusing on. I think if people are focused on cost, they are focused on the financing costs, not on the energy costs.

Operator

Now we're going to take out next question, and the question comes from the line of Brian Dobson from Clear Street LLC.

Brian Dobson

Thanks for the great color on the lead in and the call. As you're speaking to potential clients and understanding that -- I don't want specifics, just in general, how are these potential customers reacting to rising building costs and rising financing costs? Do you think ultimately, that gets passed through and the yield on the contract remains relatively similar to what we've been seeing or might, given the potential for regulation might see better pricing?

Benjamin Gagnon

Brian, thanks for the question. When it comes to changing costs, I think that there's a couple of different things going on. And I can understand why investors in the market is a little bit confused around this because the numbers that you'll see out there in the market are pretty wide ranging. You might see solutions as low as $5 million a megawatt. You might see solutions maybe upwards of $20 million a megawatt. And there is no real defined rule of thumb that I think you can apply to all sites. Scale, location, timing, design, equipment, uptime requirements, everything is going to factor into the cost.

What we think and what we mentioned on the call is that these projects are judged by a yield on cost, both by the tenant and by us as the developer. As costs for the construction do go up, we do reflect those in the lease economics so that the yield costs or the yield economics remain competitive. And when the costs went down, the lease economics would also be reflected in the same way. So there is a little bit of a natural inbuilt kind of hedge, you might say, in the way that these contracts are negotiated and structured. Where you get into, I think, a big problem is when you sign a lease without having everything contracted on the back end. And then there is a gap between the price that you've signed and then the price that you're actually fulfilling at.

And what we've been able to do so far, we've been very conservative in how we're moving forward with these sites and what we're committing to so that we don't have that problem, and we don't have that mismatch between signing a lease and actually pricing out all the individual components within it. So that's our best hedge against rising prices is just trying to best time the lease and everything together, which is the strategy we've been executing on.

Operator

Now we're going to take our next question. And it comes from the line of Martin Toner from ATB Cormark.

Martin Toner

Do your comments today preclude signing a lease before year-end?

Benjamin Gagnon

As I said on the call, we don't want to provide any sort of guidance as to timing, and that's positive or negative. We don't want to provide any sort of guiding on that. What we will say is that the tension is as high as it's ever been. We continue to progress the sites. We had good progress even this morning. So we're continuing to advance. It's on a daily basis. These are all being developed and negotiated in real time. And so we'll provide updates to the market when appropriate and when we're able to do so, but we're not going to provide any sort of guiding expectations on time.

Martin Toner

Fair enough. How about a couple of questions about politics. The state Senate in PA has flipped. Do you think that will result in legislation around data centers that could create -- and then what would that do to time line?

Benjamin Gagnon

So it's hard for us, and we generally don't like to comment on politics, especially when the midterms were so recent. I think what we will say is that the Governor Shapiro, he won in a land slide as he was expected to. And he's remained very supportive of building data centers, but under what he has put in his own words as the strictest guardrails in the nation.

I think when you look around at the states that were impacted by midterms, Pennsylvania is probably better off than many, many other states over the last couple of days. And I think the reality of the situation is that data centers will continue to be a political subject for some time. I don't think it ends at the midterms. But what we do think is that we have been operating effectively under the Keel standard, which is basically very similar to the grid standard since we started in Pennsylvania. And we think that we're going to be very, very well positioned for whatever is coming up in the next year or 2. We have received great positive feedback from permitting offices, regulatory agencies, government offices, the local state senators and politicians. We received positive feedback up and down, and we expect to continue to progress on our good work in Pennsylvania.

Martin Toner

Last one. Can politicians take action to prevent you from getting the remaining permit that Panther Creek or Sharon?

Benjamin Gagnon

I don't know what a politician could do to stop that process because these are very technical permits, and they're really not political. I think what Governor Shapiro has already put in place with the strictest guardrails in the nation are very, very strong principles. And I don't think that we should expect that there needs to be further action in the immediate coming weeks. But we believe we have high confidence -- we still have high confidence in closing out our permits. I don't know -- I don't believe that there's anything that anyone could do at this time to stop that.

Operator

Thank you. Dear speakers, there are no further questions for today. I would now like to hand the conference over to your speaker, Ben Gagnon, for any closing remarks.

Benjamin Gagnon

Thank you all for joining us today, and thank you to the entire Keel team. This was a quarter that tested everyone. We will see you next quarter, further along. Thank you.

Operator

That does conclude our conference for today. Thank you for participation. You may now disconnect. Have a nice day.