Applied Digital Corp. (APLD) Q4 2026 Earnings Report, Transcript and Summary

Good afternoon, and welcome to Applied Digital's Fiscal Fourth Quarter 2026 Conference Call. My name is Pascal Berman, and I will be your operator today. Before this call, Applied Digital issued its financial results for the fiscal fourth quarter ended May 31, 2026, in a press release, a copy of which has been furnished in a report on Form 8-K filed with the Securities and Exchange Commission or SEC, and will be available in the Investor Relations section of the company's website. Joining us on today's call are Applied Digital's Chairman and CEO, Wes Cummins; and CFO, Saidal Mohmand. Following their remarks, we will open the call for questions. Before we begin, Matt Glover from Gateway Group will make a brief introductory statement. Mr. Glover, you may begin.

Thank you, operator. Hello, everyone, and welcome to Applied Digital's Fiscal Fourth Quarter 2026 Conference Call. Before management begins formal remarks, we'd like to remind everyone that some statements we are making today may be considered forward-looking statements under the securities laws and involve a number of risks and uncertainties. As a result, we caution you that there are a number of factors, many of which are beyond our control, which could cause actual results and events to differ materially from those described in the forward-looking statements. For more detailed risks, uncertainties and assumptions relating to our forward-looking statements, please see the disclosures in our earnings release and public filings made with the SEC. We disclaim any obligation or undertaking to update forward-looking statements to reflect circumstances or events that occur after the date the forward-looking statements are made, except as required by law. We also discuss non-GAAP financial metrics and encourage you to read our disclosures and the reconciliation tables to the applicable GAAP measures in our earnings release carefully as you consider these metrics. We refer you to our filings with the SEC for detailed disclosures and descriptions of our business as well as uncertainties and other variable circumstances, including, but not limited to, risks and uncertainties identified under the caption Risk Factors in our annual report on Form 10-K and our quarterly reports on Form 10-Q. You may access Applied Digital's SEC filings for free by visiting the SEC website at www.sec.gov. I would like to also remind everyone that this call is being recorded and made available for replay via link in the IR section of Applied Digital's website. Now I'd like to turn the call over to Applied Digital's Chairman and CEO, Wes Cummins. Wes?

Thanks, Matt, and good afternoon, everyone. Thank you for joining our fiscal fourth quarter 2026 earnings conference call. This was a defining quarter for Applied Digital, capping a transformational year for the company. We signed leases for 5 campuses, including 3 in just the past 4 months. We created $36 billion of total contracted long-term lease value and approximately $20 billion of that in the last quarter. This represents a 125% increase in contracted lease value underpinned by 1.41 gigawatts of contracted critical IT load for all campuses. As previously mentioned, we recently signed 3 new campuses, Delta Forge 1, Polaris Forge 3 and Delta Forge 2 with the same high investment-grade hyperscaler. These campuses span 3 states across 2 distinct regions of the country. Delta Forge 1 and Polaris Forge 3 are each approximately $7.5 billion in base term contracted lease revenue and Delta Forge 2 adds approximately $5.2 billion, together representing approximately $20 billion in long-term contracted revenue from a single world-class customer. Over the course of the year, we delivered 100 megawatts of Polaris Forge 1 on time and on budget. We also achieved strong financial results, which you will hear about in a few moments. Just as importantly, we deepened our partnerships with local communities, building trust, making lasting investments and helping improve residents' quality of life. We are currently constructing 5 multibillion-dollar AI factory campuses for 3 separate hyperscalers at a scale we believe speaks both to the quality of our platform and to the trust these customers place in our ability to execute. We achieved this kind of scale by leveraging our proprietary data center design and world-class supply chain to efficiently replicate our builds across a diverse set of geographies and climate conditions. At the center of our approach is what we call our franchise model. When we begin development at a new campus, we established the same core group of approximately 15 to 20 leadership positions, each reporting directly to headquarters. This repeatable operating structure, combined with the strength of our supply chain and our status as an approved supplier with every major hyperscaler positions us to deliver a platform that is both differentiated and scalable. We believe this model is one of the key reasons we have emerged as a clear leader in the industry. Together, we believe these capabilities provide a strong foundation for creating significant long-term value to our customers, shareholders and communities we serve. I'm even more excited about our pipeline beyond the 1.41 gigawatts currently under construction, particularly as rental rates have moved higher over the past 6 months. We are actively marketing an additional 1.7 gigawatts across multiple states and expect this new capacity to command higher pricing. We're seeing demand not only for entirely new campuses, but also for additional capacity at our existing locations. We are currently in advanced negotiations with 2 existing investment-grade customers to finalize leases associated with their respective expansion options for approximately 100 megawatts and 150 megawatts at these locations. We expect these expansion leases to be executed on substantially the same terms as the customers' current lease agreements, but at materially higher lease rates than the existing leases and possibly longer duration. If executed, these leases would bring our total capacity to 1.66 gigawatts and over $6 billion of additional contracted revenue based on existing rates and duration. We expect the ultimate amount to be even greater, reflecting the anticipated higher rates and potentially longer duration. Importantly, we believe the opportunity extends beyond simply adding contracted megawatts. As we continue to expand our platform, we see an opportunity to increase operating leverage through premium pricing as well as further diversifying our customer base across both our existing and future campuses. As it relates to our power pipeline, I'm especially excited about our strategy and our work with Base Electron, an independent power producer collaborating with Babcock & Wilcox for regional utilities and regional utilities to develop roughly 1.2 gigawatts of front-of-the-meter natural gas-fired generation in the Dakotas. We're seeking to position our shareholders to benefit from Base Electron's success as we currently own approximately 10% of the company. However, we believe that even more compelling opportunity is the power itself, which is the single most valuable and constrained resource in our industry. This generation is expected to unlock expansion at existing campuses, enable the development of new ones and deepen our access to one of the country's most advantaged energy regions. North Dakota's Bakken Shale is among the most abundant low-cost energy sources in the United States. Combined with the region's natively cool climate and business-friendly environment, we believe the region is exceptionally well suited for data center development. Our core belief is that this combination creates a significant competitive advantage and a barrier to entry that is very difficult to replicate. We believe that if we continue to build the power that hyperscalers will continue to come to our regions. Turning to execution. Last fall, we delivered our first 100 megawatts at Polaris Forge 1 on time and just recently, we delivered 75 additional megawatts at the same campus, again, on schedule. On-time delivery is a meaningful differentiator in the industry, and we strongly believe our track record sets us apart from our competitors. Industry data shows roughly 90% of industry-wide projects costing more than $1 billion are delivered late or over budget. We are proud to be among the remaining 10% category and are committed to maintaining that performance. Today, all of our construction projects are on time and on budget. Turning to our data center hosting business. This segment provides energized space for Bitcoin mining across our 2 sites in North Dakota. It continues to perform well and remains the highest return on asset business in our portfolio. Importantly, we are paid based on the data center capacity provided to our customers. So as long as they are mining, we are paid regardless of where the price of Bitcoin trades, which makes this a steady high-margin source of cash flow. Turning to ChronoScale. During the quarter, we completed the separation of our cloud business, which began trading on NASDAQ under the ticker CHRN in early May. Applied Digital currently holds 96% ownership, so our shareholders continue to participate in the upside of that business as it seeks to scale independently as a dedicated accelerated compute platform. ChronoScale has already made meaningful progress building out its leadership team, most notably the appoint of Raj as Chief Technology Officer. Raj joins after more than 13 years at Tesla, where he served as Vice President reporting directly to Elon Musk and led a broad portfolio spanning AI infrastructure and one of the largest GPU clusters in the world. ChronoScale also named Florence Lamb, who brings more than 20 years of scaling global cloud and AI platforms at companies, including Supermicro as Chief Product Officer. We believe attracting talent of this caliber underscores the scale of the opportunity in front of ChronoScale as an independent company. During the quarter, ChronoScale extended the customer contract at higher pricing for its deployed fleet of GPUs. The company also began demonstrating its secure enterprise environment to select partners. The company's platform allows enterprises to deploy AI in a secure, controlled environment regardless of whether the data resides on-premise or at one of several large cloud providers. The platform supports a multitude of AI models, allowing partners to choose which is best for their enterprise. In addition to the enterprise cloud, ChronoScale is also pursuing multiple large reserve contract opportunities that if secured on favorable terms will allow the company to deploy hundreds of megawatts of compute on a take-or-pay long-term contract. With that, I'll turn the call over to our CFO, Saidal Mohmand, for a detailed review of the financials. Saidal?

Thank you, Wes. Before I turn to the quarter, I want to spend a moment on financing because our team did a tremendous amount of work over the past several months to secure lower cost of capital. During and shortly after the quarter, we closed our $2.15 billion of 6.75% senior secured notes to fund our Polaris Forge 2 campus. We closed a $300 million senior secured bridge facility led by Goldman Sachs. We secured a revolving credit facility of up to $550 million and closed our $1.59 billion 7% senior secured notes to fund our fourth building at the Polaris Forge 1 campus. Continuing to drive down our overall cost of capital remains one of my highest priorities. A key driver of that progress has been our work with CoreWeave at Polaris Forge 1. By restructuring the leases at that campus through a special purpose vehicle and by establishing a memorandum of understanding around the credit supporting our debt financing, we were able to place our recent $1.5 billion notes at 7%, 225 basis points inside our first placement, which priced at roughly 9.25%. Just as encouraging, that placement, the notes on our initial 2 HPC buildings at Polaris Forge 1 is now trading at a meaningful tighter spread in the secondary market, which we believe positions us well to refinance that debt at a lower cost in the future. With these transactions, we have now secured the financing needs for the full 400 megawatts at Polaris Forge I and the 200 megawatts of Polaris Forge 2. Looking ahead, we expect the financing for our next 3 campuses to be relatively straightforward. Under our arrangement with Macquarie, they fund 3/4 of the equity. And because these campuses are leased to a high investment-grade hyperscaler, we anticipate favorable rates on our future debt placements. Additionally, signing direct investment-grade hyperscaler leases allows us to maintain a favorable cost of capital through the entire lease term as opposed to indirect or backstop leases, which face uncertainty after the initial 5-year tenor. Taken together, this is a financing model we believe is both repeatable and increasingly efficient as our cost of capital continues to improve. Now let's turn to the quarter. I'll cover the fourth quarter numbers in my comments. Please note that unless otherwise specified, these figures reflect only our continuing operations. In the fourth quarter, total revenues were $258.7 million with $208.2 million of services revenue and $50.6 million of data center rental and other revenue. Overall, total revenues increased 407% from the comparative prior quarter. For the quarter, our HPC hosting business generated $203 million in revenue, consisting of $152.4 million related to tenant fit-out services, $44.1 million related to base rent and $6.5 million related to tenant recoveries. The data center hosting segment, which operates our crypto data centers, had another strong quarter with $37.3 million in revenue, materially consistent year-over-year with stable operating conditions. We are very pleased with this business, which continues to deliver the highest return on assets in the company, generating $12.5 million in segment operating profit in just 1 quarter on $113.8 million in reported assets. Because we own the majority of ChronoScale, we consolidated its $18.8 million of revenues for this quarter. As ChronoScale is pursuing a separate strategy from our core business and now operates as a separate publicly traded company, we have excluded this segment from our non-GAAP results. Services cost of revenues increased by $138.9 million to $208.2 million this quarter. The increase was driven primarily by the $145.6 million in tenant fit-out services performed within our HPC hosting business. Data center rental and other cost of revenues were $25.1 million for the fourth quarter, primarily driven by approximately $14.1 million in depreciation and amortization associated with our first HPC data center at Polaris Forge 1, $6.4 million in expenses, which are reimbursable as tenant recoveries and $4.5 million in personnel and other operating costs supporting our facilities. SG&A expense increased $124.3 million to $165.3 million this quarter. The increase was primarily driven by $116.8 million in stock-based compensation due to accelerated vesting of certain employee stock awards as well as grant activity associated with the separation of the cloud service business and an increase in headcount as well as $7.3 million in personnel expenses also related to the increase in headcount. One item worth calling out this quarter. Our stock-based compensation included $47.9 million tied to onetime awards connected to the ChronoScale transaction and $65.1 million tied to performance stock units. Net loss attributable to common shareholders was $111.6 million or $0.39 per share. Adjusted net income was $12.9 million or $0.04 per diluted share. Depreciation for the quarter was approximately $18.2 million. Adjusted EBITDA was $42.4 million, up from $1 million in the comparative prior quarter. Net operating income, or NOI, was $39.9 million, representing a 91% margin, and we define it as our HBC base rental revenue less our rental property operating expenses, property taxes and insurance expenses. From a balance sheet perspective, we believe we are very well positioned. We ended the quarter with $4.2 billion in cash, $5 billion in debt and approximately $1.7 billion in equity. As you evaluate these results, keep in mind that our current financials on the HPC data center side primarily reflect only the initial 100 megawatts that are online and contributing during the quarter. Looking ahead, as we bring additional capacity online, investors should expect to see a significant step-up in our numbers over the coming quarters and years. We are currently building towards a 1.5 gigawatts of HPC AI infrastructure, and we expect this ramp will drive meaningful growth in revenue, EBITDA and NOI as those megawatts come into service. Now I'll turn over the call to Wes for closing remarks.

Thank you, Sid. We're seeing the AI infrastructure build-out enter a powerful new phase. Hyperscalers are no longer just investing in AI infrastructure. They're accelerating their commitments at an unprecedented scale. U.S. technology companies have now committed to approximately $850 billion of data center lease obligations over the next several years, an increase of roughly $570 billion year-over-year, more than triple prior levels. These are not forecasts. These are long-term contractual commitments backed by many of the world's largest technology companies with the largest and strongest balance sheets and credit ratings. AI infrastructure spending at the major hyperscalers is projected to reach approximately 3.2% of U.S. GDP in 2027, surpassing projected U.S. national defense spending for the first time. Taken together, these long-term commitments reinforce our conviction that we remain in the early stages of what we believe could be one of the largest infrastructure investment cycles in the modern economic history, and we believe Applied Digital is well positioned to capitalize on that opportunity. As I wrap up, I would like to leave you with a few final thoughts. Building AI infrastructure at scale is incredibly complex. Balancing aggressive construction schedules, customer expectations, power infrastructure and community partnerships is no small task. Through it all, our guiding principles have remained remarkably simple, do it the right way. For our customers, that means delivering high-quality GPU-ready data center capacity on time. That commitment is reflected in both the customer relationships we're building and the industry recognition we've received. For our communities, it means being a trusted partner who creates lasting economic value. From the beginning, we believed in building with our communities, not simply in them. We engaged early, listened often and strive to ensure every project leaves a lasting positive impact. The jobs, tax revenue and long-term investment we bring help strengthen local infrastructure, support schools, first responders and create opportunities that can be truly transformational for these communities. Our operations are delivering measurable benefits today. At Polaris Forge 1, our use of excess regional grid capacity has already returned more than $45 million in electricity credits to local ratepayers. If you'd like to see our approach firsthand, I encourage you to watch our behind-the-build docuseries, where we share the town hall meetings and community conversations and an important part of every project we undertake. Finally, a year ago, we set a goal for $1 billion of net operating income within 5 years. We now expect to achieve that run rate goal a year from now or 3 years ahead of schedule. Our platform is now supported by approximately $36 billion of long-term contracted lease revenue. More importantly, we believe the structure of our contracts, the majority being directly with investment-grade customers establish a durable earnings and cash flow foundation from which we can continue to expand as our customer demand continues to grow. We cannot overstate the competitive advantage our established footprint provides. Our current campuses have the ability to expand and in some cases, expand dramatically. We believe expansion on current campuses alone provides us visibility to expand to over 5 gigawatts of critical IT load through 2032. Utilizing existing campus infrastructure will not only shorten economies -- not only shorten development timelines, but will improve returns via economies of scale and establishing regions of excellence. Our collaboration with Base Electron is intended to augment the power to support this expansion. We do expect to continue to add new campuses this year and next year, compounding the growth potential and diversification. Since our call this time last year, we've expanded from 1 campus to 5, increased contracted revenue from $7 billion to $36 billion and added over 1 gigawatt of capacity with investment-grade customers with over 80% of that leased to a high investment-grade customer. We have delivered significant capacity on time and on budget. We have also dramatically lowered our cost of capital. I'm extremely proud of our team and their accomplishments. Our opportunity is significant, but our focus remains unchanged: Execute with discipline, deliver for our customers and our communities and create long-term value for our shareholders. With that, operator, we're happy to open the call for questions.

Your first question is from the line of Mike Grondahl with Northland Securities.

Some of the questions we've been getting lately really revolve around the 3 recent leases for 810 megawatts and the lower yields or yield to development cost kind of in relation to your peers. Can you talk about Applied's strategy and signing those 3 leases?

Sure, Mike. So when we go back to the goal of what we have been seeking to achieve for the past 12 months, it was to put a solid foundation in place for Applied Digital, and we've walked through this for many quarters where we signed CoreWeave first with an investor-grade hyperscaler. And after we signed CoreWeave, we made a commitment to get to 70% of our contracted lease revenue to investment-grade hyperscalers. And so that's been a big accomplishment for us. We signed 2 investment-grade hyperscalers. We're over 70% soon to be, I think we'll be close to 80%, 76% of that on investment grade. So we made that achievement. And what I would say about the yields I think from a cost perspective, we're fairly conservative. We want to set expectations that we can at least meet, if not beat. So from a margin and cost perspective, I think we're fairly conservative from a pricing perspective. These discussions were going on for a significant amount of time. I do think the prices have increased, as I mentioned in the script, and I think you'll see that flow through for us on what we have in front of us. But I do think we're hitting on our lease return rates. I think that you comped the entire industry. Now there's a very small portion of the industry that's public company. But if you comp the entire industry and you look at companies that were doing leases of this scale with these types of customers directly, we would be right in the band, if not at the middle towards the higher end of that band of contracted lease rates again for these types of customers with this kind of duration and this kind of scale, and I think we'll be able to drive our financing costs significantly lower. But I think the important part, Mike, is we have a very solid foundation to get -- you can do the math and kind of walk through that we have about $2-plus billion of net operating income contracted on an annual basis at this point. So we'll be able to grow from there. But with very solid customers. And when you look at how those returns go through, and this is kind of how our team -- myself and our team has thought about this from the beginning when we entered this business, these businesses, whether they're public or in the private markets when they transact kind of go from 20x to 25x that NOI number. And so that's really what the goal was for us is to get into very solid leases and really good SLAs. I talked about that a significant amount because lease yields over time really depend on your SLAs once you have the building operating. And I think we have those locked-in from a contract perspective. So I'm really proud of the achievement there.

Got it. And then maybe a question for Saidal. Can you talk about your strategy around managing the cost of capital really on both the debt and the equity side?

Yes. Great question. So one, I think as Wes alluded to, the type of hyperscaler that we contract with, going direct with the end high-grade -- investment-grade hyperscaler particularly at -- rates that are in the high end of the band where we see these types of leases transact and our peers comp to, that acts as a function that you can actually lower your cost through the initial if it's 15 years or even longer. So I think versus -- and I talked to you in the transcript, some of these longer duration or shorter duration backstop leases where the guarantees run off after 5 years. So that's 1 portion of mitigating and creating a consistent return. But then if you think about it, we talked about our flywheel. And our flywheel really has 3 components. So the first component, as Macquarie funds 3/4 of the equity, the first component comes from the Applied's balance sheet. So between our corporate cash flows, our low-cost revolver, which were recently secured, we can fund the initial portion at a very attractive cost of revolver, currently is SOFR plus 225. But the MAM equity, I think, is 1 part where I think people tend to overlook and how that's an attractive form of capital. So Macquarie funds through their $5 billion JV with us, 3/4 of the equity that's required. If you look at the 1.8x MOIC in that -- for that transaction, it's roughly a mid-teens IRR throughout. Now if you compare that versus just a common stock issuance, right, which can be highly dilutive if you think the current and past prices are extremely undervalued or even a more attractive option versus large comparable debt offerings, which is 1 popular form to plug in the equity. And I think looking at the convert, for instance, despite the convert being portrayed as a cheap cost of debt, but there is a disadvantage of having it in a large scale. One, it's negative from a credit ratings perspective, which can negatively impact your cost of capital and then also can create an overhang on the equity if you ever hit times of turbulence when you need to refinance it and it's truly treated as debt. So that's 1 thing. And if you look at the cap call math too, right, the cost of capital on issuing equity when it's severely undervalued, or going for convert can be well in excess of 20%. So we think the MAM is a programmatic consistent approach for funding the majority of the equity. And then the third portion of the flywheel, which we have is effectively the site-specific debt. So the first form is through construction. We tap it predominantly through the project bond markets as of today. The project finance markets have also been available, which is through the banks. Generally low cost. And then what you'll see 2 years for the bond or if it's within a -- as RFS occurs for the project finance market, you can roll it into more of a permanent financing, be it CMBS, ABS or still the 144A IG market. And with that, your construction cost and your cost of financing decreases, right, as construction risk is taken off the table. So we approach it from, I think, 3 different forms and all through managing leverage at a very conservative level. So if you take an 80% LTC and what we need to build out for 1.4 gigs, that's contracted and take our average annualized NOI of $2 billion, your sub 7x levered, which is well below comps in the -- particularly in the private markets, which operate in excess of 10 turns of leverage. So that's how we handle financing, and we try to do in a conservative programmatic stable manner.

Your next question comes from the line of Nick Giles with B. Riley Securities.

I wanted to ask about the cadence of CapEx spend to the balance of the calendar year. It seems like quarter-over-quarter spend took a significant step up as expected. But curious if you have any sort of run rate or kind of where that should go quarter-over-quarter as we try and model out '26 and '27?

Yes, Nick. So CapEx, so you should expect it around $600-ish million for the upcoming quarter, and that will take a step up as we enter more advanced stages of construction at the new campuses. I would note, though, as you see, what we have done is we've tapped the financing markets for these sites. Earlier on in the construction versus the first 2 Ellendale buildings, which were well more advanced versus where we're at with PF2, for instance.

Got it. And maybe just on the restricted cash balance, over $2 billion. Can you just remind us what the split is between debt service reserve, letters of credit, and then kind of what we should expect on the release of that cash?

Yes. Let me -- so vast majority of the restricted cash was the Polaris Forge 2 bond. So that was in held in escrow until the ESA was a release which we released that, I believe, in June. So that cash is something has been unrestricted. And our 10-K will have greater detail of the actual other accounts.

Your next line of questioning is from Rob Brown with Lake Street Capital Markets.

Congrats on all the progress. I just wanted to dive in a little bit on the 2 customers you mentioned that you were looking at expanding, I think, 100 and 150 megawatts. Could you give us a little color on just sort of how that would play out? Would these be the existing sites? And maybe some color on the comments about the rate increases.

Sure, Rob. So on the expansions, when we started building Polaris Forge 2, we started building 2 buildings 300 megawatts, we contracted 200 megawatts. We expect in the near term to contract that additional 100 megawatts with the same tenant at that campus. And then on 1 of our Delta Forge campuses, we were negotiating and we mentioned, in advance stage negotiation with the tenant there for a third building on 1 of those campuses. And as I mentioned in my prepared remarks, we do expect materially higher pricing on both of those expansions as well as new campuses and new capacity that we signed in the future. It's been -- we've seen other contracts out. It's great to see pricing moving up in the industry -- just gives a really strong indicator of the demand that's out there.

Okay. Great. And then on Base Electron, how do you sort of see that driving incremental kind of customer demand as that you have the power available in North Dakota that will allow you to sort of add capacity there just in terms of what Base Electron sort of does for you?

Yes. So as we mentioned, we've announced that there's 1.2 gigawatts being built in North Dakota. This is all front of the meter on grid capacity. We're working on another project there in a different part of the state as well. But these will go with the utilities that we work with already. They'll deliver that capacity for our data center campuses but also for other rate payers on the regional system. But that's a start for Base Electron and we expect that to expand significantly. And we talked about our campuses when we first started with the -- we signed initial capacity with physical infrastructure transmission, infrastructure is there to significantly expand the campuses, you need to add some additional electrons. And it won't just be us or it won't just be sorry, Base Electron adds additional power generation to the network. We expect a lot of other power generation projects in the region as well. But as that additional generation comes online over the next few years to be able to expand all of those samples, they all go north of the gigawatt and as I mentioned 1 of those campuses are significantly north of gigawatt and so we're excited about that entire region. We started there. We had great success building up in North Dakota, and we look forward to just continuing to build that. And Rob, lastly, as I mentioned in the prepared remarks, we see clear line of sight with our existing campuses to over 5 gigawatts of critical IT load capacity, and a big part of that is the Base Electron generation additions.

Your next question is from Derrick Whitfield with Texas Capital.

Congrats on your commercial progress over the last year. I want to start first with just what you see on the demand side, maybe with regard to the high investment-grade hyperscalers and the next lower tier. Are you guys sensing any change in demand based on inflationary pressures?

We haven't seen that. As you see -- as we mentioned or as I mentioned in my prepared remarks, we've seen pricing moving higher in the market. I think that's a good indicator of demand versus supply. But we still see extraordinarily robust demands in the market across both of the categories that you mentioned.

Great. And then just based on your prepared comments, I mean it appears your projects are still tracking in that $11 million to $13 million per megawatt range. With that said, I mean, what are the general conditions that would lead you to the lower end versus higher end of that range? And are you expecting any regional differences based on labor conditions in those areas?

Labor is a big issue that we have been solving and I expect us to continue to solve in almost every region. But one of the things that we do on labor -- so let me back up and start with kind of the high and low end of the band. The high end of the band, there's some site-specific things that typically go into that. So when we start a new campus, you'll typically see our builds at the higher end of that band because we include all campus costs in the first building, first 2 buildings. So that typically includes new substations, some transmission on campus, the land and the power side itself. So that tends to drive it towards the high end as we try to work it down over time as we add additional buildings on those campuses. And you mentioned that kind of the economies of scale as we go at each individual campus. And then there's other things, depending on -- for example, 1 of our sites in the South will be slightly higher because of additional dirt work and site prep because of the type of soil that we're dealing with. So there's some very specific things that happened from a location by location basis. But the remainder of what we do is really dialed in at this point from supply chain, from construction process and construction, labor rates have been fairly steady over the past 6 months. But we do a lot of work in 2 ways on labor. One, we try to stay out of the most crowded markets and so you see we're the only ones that are really building right now in North Dakota. We have a few other markets that are not as crowded as, say like, Texas or West Texas or some of the other markets that you have 100-plus projects happening. So we try to stay in less competitive markets from a labor perspective. And then also, we do a lot of education in those markets. We work with local technical college implicational schools. We do that right at the beginning. We've been doing that for a while in the Dakotas. And when we move into other states, one of the first things we do is work with the local vocational technical colleges, we even work with them to set curriculum. We typically make some donations but we'd want to train people up there to work construction and operations of our facilities.

Your next question comes from the line of George Sutton with Craig-Hallum Capital Group.

Wes, I wondered if you could just talk about the governors to your growth. Obviously, demand does not seem to be 1 of those governors. But when we think through power and supply chain and the number of teams you could handle at any 1 time, where do you see the governor being?

Yes. So I think, George, that's a great question. So hours definitely top of the list as far as when power is available and how much as we contracted, I think, 2.1 gigawatts of utility power over the past year. And so that's definitely one. And it's when is power available and the timing or building to start to match when power becomes available at the location. So that's one. Supply chain is another as we've talked about many times in the past, we worked really well on supply chain a few years ago, locking in a lot of capacity sort of electrical, for all of the energy, the mechanical, electrical and plumbing. But it always does have limits. I think at one point I mentioned we had about 700 megawatts per year, that's critical IT load, and so we have contracted to build over the next couple of years of 1.5 gigawatts. So we're definitely exceeding that a little bit. But those are 2 definitely the biggest governors. We have great process from a construction perspective in place. Our first building on Polaris Forge 1 took us about 24 months from start of construction to RFS. Our second building on that site was under 12 months. We've really dialed in and we're just getting better. The team is getting better with every iteration that we do and just how we sequence things, how we just make everything much more efficient from a construction process. So I feel really good about that piece. Managing supply chain and power are probably the 2 biggest constraints at the moment.

So I wondered, we're obviously in a market that's gotten very cautious relative to AI. Your stock has gotten brought into that. And none of what you're talking about on this call represents some of the concerns out there relative to the NIMBY and the open model concern. I'm just wondering if you can give us a bigger picture AI thesis as you see it today relative to what the market is thinking.

Sure. Just from the demand side, this is -- I think this might sound a little strange, but I think I'm a fairly conservative person. And so that's why, over the past year, we really focused on high-quality customers, durable contracts. That was really our mantra was durable contracts, both from an ability for us to deliver, an ability for our customers to cancel and then also from an SLA perspective because you don't operate the sites well then in almost every instance that I've seen in to people in the industry, customers who have the right to cancel. So that's really been the focus to make sure we get those types of contracts, rebuild the right type of buildings. So we try to make it as absolutely as efficient as possible but we don't want to skip over things just because it could lower cost because remember, we need to operate these buildings for at least 15 years on the contract, but we think they're 30-plus year assets. So we build buildings that we can operate, and we think will meet all those SLAs for a really long period of time. And so that was the focus was making sure that we have that type of a platform. And then as far as -- we focused on high investment-grade hyperscalers, investment-grade hyperscalers and those types of companies. So what did we purposely avoided signing leases with the very large model companies. Those are great companies, but I don't think that -- I don't know how that plays out over time. we see this volatility result with DeepSeek in 2025. We see it with Kimi in 2026 now. And so we see this volatility. And then outside of those models, you see the volatility just with the U.S.-based companies. So if we were speaking this time last year, OpenAI was absolutely the best day to introduce GPT 4.0 in April of '25, and they were doing extraordinarily well and then it rolled to Google and now to Anthropic. And so we just really had a focus on the highest quality companies that have a high investment grade rating. But what appears, George is happening, whether it's open source or closed model, that's a totally different debate, but it seems to all need a significant amount of compute. And they seem to use the same amount on inference or a little bit more on inference. So compute still is the foundational layer. I feel really good about our positioning in that market and the band for compute, regardless of which way kind of the world goes from a technology perspective over the next few years.

Your next question comes from the line of John Todaro with Needham & Company.

I guess just going back to the earlier question on the growth governor. Wes, I think you've mentioned in the past you can work on 7 campuses simultaneously, you're at 5 now. I imagine some get completed, you had other ones. Is that number push higher even above 7? And is the currently marketed 1.7 gigawatts, will that be kind of captured within what you could do simultaneously?

Yes. So great question. So yes, I've mentioned 7 before -- I will say I think we can go higher than that. Let us get there and see how it's working -- One thing, John, that we don't want to do is overextend ourselves. We want to make sure that we always are in the right position to execute and deliver. That's number one. When I go through the risks were initially signing contracts, then second is delivery. We've got those 2 dialed in. And then the third is operations. We've been operating in the first building for roughly 8 months now. That's going extremely well. So we're getting all of those things dialed in really well. I just want to make sure that we continue to execute and deliver on time for our customers because it's such a key thing. So I'd rather -- if we need to stop at 7, we'll stop at 7, but we'll see when we get there and how far we think we can go with managing the construction process. And then again, as I mentioned, there's definitely some limits on supply chain. We continue to be able -- we've been able to continue to keep expanding that limit. So I don't know where that is for certain, but it definitely does exist out there as far as that supply chain limitation. But I still see if we get to the 7 campuses, we're obviously marketing to that now. But we feel comfortable with that. And then once we get there, we'll see if we feel comfortable going above it.

Understood. And then, Saidal, I think you had mentioned the NOI margin at 91%. Just wondering if that's kind of more so the target number we should go with for all the contracted capacity or if like the mid-80s, which I think we're at, at least, is more fair.

So as Wes mentioned earlier, right, we try to aim conservative for margin targets and then obviously deliver to the high end. There's a mix of it, right? One, you'll see us as we get a site up and running post the initial RFS date, you should see margins continue to increase, just one, we get better operating the site. And then secondly, there's also -- you should see some economies of scale as we have more buildings on within a specific campus. So that's definitely the goal. The timeline of how we get there for every campus, right, could differ depending on the -- differ depending on the region, but that's where we're marching towards as well.

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

Two, if I may. So on Base Electron, can you just give us a general sense for when that capacity might come online? And then, Wes, your comments about your North Dakota properties being able to expand to various levels, how much of that is contingent on Base Electron versus just utility growth? And then second question, beside from the 250 megawatts you guys talked about, the marketing of the -- I guess, the remaining 1.5 how would you kind of characterize that between existing customers that have taken down capacity versus maybe some folks have been lesser to alter negotiations?

Sure. So on Base Electron, the timing, so that's in '29 and '30 for that initial capacity and then continue to ramp from there. In North Dakota, there's some other projects, there's some transmission projects, plus some generation projects that will fill in nicely. So we've got a lot to build for '26, '27 and some '28 and we'll have additional power capacity, we think, coming online at that time to just continue to expand those campuses. But they're both important to meeting that goal is both Base Electron and then additional power projects that are going on in the region and transmission projects. There's a JETx line that runs between Ellendale and Jamestown that will come online and expand the Ellendale power capacity for us as well. Darren, remind me on your last question.

It was more around the power you're the marketing right now -- relative to existing customers versus others, yes?

Right now, Darren, just given what we have going on, I would expect that to be new customers.

Your final question comes from the line of Michael Donovan with Compass Point Research.

Congrats on the execution. So your AI factors are designed to support flexibility in the type and density of comp you deployed. What changes are you seeing in recent customer requirements beyond GPUs, particularly for CPUs, memory and networking?

Yes. So just -- you made a good point about we've made this very flexible architecture that will handle GPUs, TPUs, CPUs. When you get down to that level of detail, though, it really goes customer by customer. And we don't see like that type of granularity all the time in what we do. So we get specs and how to do -- just because we're doing the fit out, how we do fit out, and so we have kind of a general idea of what they're doing, but not enough that I want to give you insights that you should rely on as far as trends of CPUs versus GPUs and storage amounts. But we do build very flexible facilities, and we've worked through that on the design perspective to where you can put almost anything that you want to, even if you were back to a standard cloud format with much lower power density, our facilities would still work for that as well.

There are no further questions at this time. I will now turn the call back to Wes Cummins for some closing remarks.

Thanks, everyone, for joining our Q4 call and look forward to speaking with you in October. Thanks.

This concludes today's call. Thank you for attending. You may now disconnect.