Digi Power X Inc. (DGXX) Earnings Call Transcript

Good morning, everyone, and thanks for being with us today as we report Digi Power X's results for the third quarter of 2026. The third quarter numbers demonstrate that Digi Power X's shift into an AI infrastructure company is starting to take hold. So, highlights for the third quarter of 2026. Revenue rose 85% to $12.25 million versus the second quarter of 2026, and was up 50% against the third quarter of 2025. AI revenue came in at $5.66 million, an increase of more than 400% from the second quarter, and it now accounts for 46% of total revenue, versus no AI revenue at all in the third quarter of 2025. GPU Rental BareMetal via NeoCloudz came in at $2.46 million, a 127% increase over the second quarter, versus zero GPU rental revenue during the third quarter of 2025. Through our power plant, energy sales hit $3.29 million, which is 68% higher than the second quarter and 9% higher year-over-year. Legacy Bitton activities fell to approximately 27% of revenue, compared with 63% a year ago. That decline was more than offset by AI revenue. EBITDA turned positive at $6.5 million, while adjusted EBITDA was positive at $0.6 million, compared with $1.7 million and negative $1 million, respectively, in the third quarter of 2025. We posted positive net income of $1 million, with meaningful support from gains on digital currency and warrant valuations; that compares to $0.1 million of net income in Q3 2025. Now let's look at the balance sheet and liquidity position as of September 30, 2026. Cash and cash equivalents together with digital currencies came to $119.6 million. Working capital was $103 million, versus $17.5 million on September 30, 2025. Our digital currency holdings reached $24.5 million, a 56% increase year-over-year. Net fixed assets were $145.6 million, up 646% year-over-year, which reflects infrastructure investment mainly at the Colombian Alabama facility. As of today, we have no outstanding borrowing. That's the balance sheet and liquidity picture as of today, November 13. We've got cash and equivalents sitting at just over $100 million. So far this year, we've put $168 million of capital expenditures into GPU hardware and data center construction at the Colombian Alabama site. Now for operational highlights and outlook. At the Colombiana, Alabama AI campus, Phase 1 construction is still on track; we anticipate handing our tenant access to the Phase 1 facility on December 15, 2026, which is roughly 4 to 4.5 weeks out. At that point we'll start conditioning the liquid cooling and power systems, and Phase 1 billing should commence in January 2027. As for Phase 2, it kicked off a few months back, and depending on various conditions being met, we're aiming for tenant access to Phase 2 in March 2027. We're on track, and we're also going to be on time to deliver access—much of the investment has already gone toward Phase 2. With our tenant Trias, we have a 10-year take-or-pay colocation agreement. For our AI colocation Internet in Colombia, Phase 1 fees are expected to average roughly $40 million annually across the 10-year term, or about $402 million altogether. Phase 2 should add approximately $58.5 million in fees per year, which would raise the combined average annual fees to around $107 million yearly and take the total contract value for both phases to roughly $1.07 billion. There is also a provision in the agreement for a nation to extend the relationship for one more term, which could raise the total potential contract value to roughly $2.5 billion. The Neocate token factory is now live. This quarter we launched our token factory out of Berlin, which moves Digi Power X from being an AI infrastructure provider to selling AI tokens directly. Operating on our own power, data centers and GPUs, the platform offers AI infance directly to customers. The big advantage here is margin; we're seeing margins in the range of 85% to 90% actual margins, so this is a business that will generate substantial positive cash flow as we continue investing in it. Since we own the entire stack, we retain the value of each token we serve rather than renting out compute at wholesale rates. We anticipate the whole factory will open up a new, fast-growing revenue stream with meaningfully higher margins, a major step a major step in moving our infrastructure into a full-scale AI business. Now let's discuss the $200 million project level debt financing. As we announced on October 13, 2026, we got a commitment letter from a major lending institution for $200 million in debt financing indicated by Goldman Sachs to support development of our 40-megawatt AI data center in Colombia, Alabama and to strengthen our balance sheet. The financing is anticipated to be arranged at the project special purpose vehicle level, backed by an unsecured company guarantee that runs through project completion and certain other limited matters. Those proceeds are intended to help fund completion of the Colombian project along with related project costs, and closing is presently anticipated during the fourth quarter of 2026, contingent on the relevant documentation and on meeting various conditions, including the lenders finishing due diligence. Now for the 2027 outlook and capital plan. By the close of third quarter 2027, we anticipate reaching an annualized revenue run rate of roughly $250 million to $300 million, which works out to about $21 million to $25 million per month, assuming adequate financing, market conditions and various order conditions. Roughly $143 million of that run rate comes from existing operations plus the Colombian tenant agreement. We expect the rest to arrive via further AI colocation and GPU capacity commentary that's in development and under customer negotiation. On CapEx, we're looking at about $350 million for 2027, mainly for roughly 10 megawatts of AI in France—or taken factory at 85%, 90% margins—along with capacity and finishing our Phase 2 Colombiana. We plan to fund this mostly through customer deposits and prepayments and contract debt financing, including the proposed $200 million facility, together with cash on hand. This adverse ties capital deployment to contracted revenue and did to minimize shareholder dilution. Let me remind all shareholders that back in 2016, I put in $8 million of my own cash, and to this day I haven't sold a single share. I'd also like to state that my long-term vision entails a very substantial return on that investment. Turning to an update on U.S. data centers: Digi Power X owns approximately 48% of U.S. Data Centers, Inc., a business that designs and deploys the arms modular AI data center system and is developing ERP atop a robotic platform for automated deployment and maintenance. USBC is raising capital independently, which will further dilute the stake we hold, while our capital contribution will come from Digi Power X with no issuance of Digi Power X shares. Therefore, in order to finance our expansion—which appears highly encouraging based on the U.S. data center figures—we intend to bring in capital. We've experienced dilution at Digi Power X, or revas receiving any allocation on the Digi Power X. And I believe it's crucial that everybody grasp that Digi Power X at present is sharply concentrated on supplying power to our GPU Bermel under the colocation arrangement and lacks the ability to set aside additional funds for growing robotics or modular systems. That, then, explains why we've established a separate structure for bringing in capital. Which growth and expansion opportunities exist? Our New York infrastructure assets are still being readied so they can be converted into AI data center operations, and for these sites we're aiming at a transition in the early part of Q3 and Q4 of 2027.

These sites have been powered—one has had power for the last 10 years, and another one where we have a total of 120 megawatts, including our power generation that has been powered for 30 years in terms of generation. And we received an allocation from the utility, which is great since it's all green power, hydropower, and we got an allocation of 63 megawatts there. Because we've been careful with our finances, we're moving quickly to build it alongside the Alabama Columbiana AI center, which lets us keep our CapEx in check. We're not looking to hurry and end up short on cash. So as we shift from a project into execution—and we've been sharing videos of our progress on that—I was actually there a few days ago, and it's really—we're very close to completing Phase 1 in a few weeks. So once we go from project to execution to live, we can continue with our other assets. In North Carolina, we now hold 2 sites, roughly 40 acres next to 1 of the state's largest electrical sword. Under the plan we are working with today, we are looking at 275 megawatts, with roughly 150 megawatts falling in 2029 and 2030. That is what will fuel our expansion in the years ahead. Across New York and North Carolina, we have room to scale beyond 400 megawatts of capacity, and how much revenue each megawatt generates depends on how we optimize it—the upside is enormous. So we clearly have—right now we hold the assets needed to double or triple our business annually for the coming several years. Revenue rose 85% versus the prior quarter, and AI already represents close to half of our business, all of this ahead of the scheduled January 2027 opening of the Colombia colocation building. Right now, what we're focused on is carrying out our plans with financial discipline, and after that we'll grow our procurement capacity and turn our Pall assets into revenues locked in through long-term contracts. At the close of the quarter, we held roughly $120 million in cash and cash equivalents, which provides a solid base for the phase ahead. We're building this company with a sharp emphasis on value per share. Our method is to lock in bankable customer commitments, put capital to work where returns are appealing, and finance expansion using customer contributions along with contract-backed financing. Our goals are ambitious, and our discipline is just as unambiguous: increase recurring revenue and cash generation while keeping shareholder dilution to a minimum. Thanks for being with us today. We'll now open the floor to questions. Okay, so this is really a remarkable number one for us. Consider that a year ago, we were essentially a coal company on its way out. Coming into January 2025, our cash position was tiny — I believe $1.7 million — and our assets were minimal, with six of them worth around $30 million. Over the course of the year, we moved from having almost no cash and few assets, plus a partially combined business, to holding $117 million in cash and $140 million in net fixed assets. On capital expenditure, we've already put $160 million in this year, and AI revenues now account for nearly half of our overall revenue. So in our view, this has been a year and a quarter of genuine transformation, as we continue expanding at our own pace. We began the process of their metal rental in May, and it went very well; many companies visit our site to observe how we execute, and everyone comes away very impressed. Colombia Alabama gives us our strongest long-term contractual revenue visibility, through a 10-year take-or-pay agreement. Phase 1 comes to roughly $402 million in contractual fees, rising to about $1.07 billion once Phase 2 is included. NeoCloudz and our other operating business also bring in ongoing revenue, though contract duration and utilization range anywhere from 2 to 5 years. We're concentrated on increasing the portion covered by long-term agreements—five years for NeoCloudz and ten to fifteen years for co-location. That makes December 15 an important figure. Considering we closed the deal with Tilibra, our tenant, in early May—only a handful of months back—it's genuinely remarkable that we could design, finance, and complete Phase 1 in roughly seven months. Being capable of delivering in such a brief window makes us, I believe, unique. Come December 15, tenants will be able to access the data role, and we'll start commissioning the power along with all the liquid cooling systems that are essentially down. This delivery milestone matters a great deal to us as the landlord, since commissioning is what lets billing begin. As mentioned a few minutes ago, the first phase of 150 megawatts is worth roughly $402 million across 10 years, while the 25-megawatt second phase contributes slightly more than $600 million, bringing the total to $1.07 billion, or about $107 million per year on average. Work on Phase 1 is already underway; Phase 1 and Phase 2 share numerous common denominators, and both phases include all the fire-suppressing requirements as shown in the data. The chillers for both Phase 1 and Phase 2 are already in place. For both Phase 1 and Phase 2, every piece of equipment has already been bought. We structured that contract so it could handle two phases, which lets us fund it financially regardless of circumstances, and we're quite far along on Phase 2 too. So granting access to Phase 2 in March 2027 doesn't present any problems in our view. Commissioning will be simpler since the liquid cooling commissioning will all be finished — that is, this month and early December. We therefore think that kind of commissioning will go further in March, and we anticipate having the full total monthly contracts in April. Turning to the $1.1 billion figure for the initial term — we have the ability to extend that up to $2.5 billion through an option. In our view, the tenants are making an enormous investment. They'll be placing data on equipment, servers and GPUs valued in the billions of dollars, so relocating becomes extremely costly for a tenant. Beyond the expense, remaining in place makes strong economic sense for as long as possible — comparable to the power shell of various cars nowadays. From where we sit, ten years out the value is going to be much bigger. What's interesting is that earlier we had a real financing offer that I said no to back in July from one of the largest institutions globally — but I wasn't happy with the terms. After that, we decided to turn down an approved financing that would have been moved and syndicated via Goldman Sachs, and doing so saved us roughly 2 points on this financing, which is amazing. Even with interest rates going up, we managed to save 2 points on the financing, and we're very close to closing it. We have the loan documents. So we anticipate closing soon, definitely in Q4 ’26, but we are pushing to close sooner; I think we are 1 of the best candidates for a lender, since we invested our own capital, we hold cash, and we are executing on time; so we are a pretty good candidate for a lender. We succeeded in being in single-digit interest, which is where things stand these days. Our funding strategy combines customer deposits and prepayments.

With the crisis that played out across the past four to five weeks — that was the culture around our share price, but for Mysterylax company, the entire market fell. It also created an opportunity for us to lock in larger deposits on new contracts. And at this point, with several talks in progress, we're wrapping up new contracts, and we're receiving much more than was the case six months back. So six months ago, a tenant or bare metal rent was prepared to give you 15%; at present they are prepared to give you 30% and occasionally up to 50%. During negotiations at present, we can reach as much as 70%, which means from the buyer's standpoint, obtaining financing is quite straightforward, enabling our expansion, since we accomplished something remarkable. We have credibility; we are not an organization that merely discusses capacity and possibly pledged capacity we settled on in '28 and '29. We deliver. And numerous truly major corporations visited our facilities. Extremely large firms tour our sites and recognize that we are building something real, that we are trustworthy. So at this point they are prepared to support us, provide larger deposits, and assist us in funding our expansion. So the main win we're seeing come through execution is that roughly $143 million of that contract is already in hand. The other $107 million to $150 million should come from building out our oversight via colocation or GPU capacity. If we had a blank checkbook, we could have wrapped it up in Q1 or Q2, but with where the market stands and our approach to financial discipline, going fast means moving deliberately and staying cautious. We made it through every bitcoin cycle, including 2016 — we've always come through — and we don't want to risk a financial problem by expanding too quickly. There's still plenty of time and lots of market share available for us to capture. And we think that me will land in the run rate that we announced. So let's turn to Q3. In Q3, we produced $4.7 million in positive cash flow. Second, that came with the help of a $12 million nonrefundable tenant infrastructure contribution. And we believe that in the next quarter, we will be cash flow positive as well, between the business and an additional nonrefundable contribution — a bigger 1 this time, given the relationship going from 15 to 25 megawatts. So at that point we are live, and then we have cash flow coming in. We think we’re well positioned right now to generate positive cash flow. We began with Permit GPU rental, and as NeoCloudz expanded, you saw a 127% sequential rise to $2.46 million. Bare metal, GPU as a service, and taken factory give us the chance to raise revenue per megawatt and margins per megawatt. That matters given the environment—the political environment. And by the way, being in Alabama is a real advantage for us. Last Tuesday, at election night, I was with the governor elect, the Coach to Tebodin, and we have an excellent relationship. He told me directly, again, that Alabama welcomes data centers, welcomes what we bring, welcomes the jobs we create, and welcomes the remarkable science and technology that will save lives, cure cancer, and cure most incurable diseases. So it is very pro technology and welcomes growth in Alabama. That makes us very excited, because we can certainly expand in Alabama, contribute to the communities, and create value for those communities by generating a great many jobs. Right now, at the campus we operate in that quite small city, roughly 250 construction workers are on site — well-compensated positions that are spending money at restaurants and hotels, so it's pretty clearly an economic boom for the area. That makes it genuinely interesting, and we're quite enthusiastic about our presence in Alabama. Returning to the topic of getting the most out of and optimizing each megawatt, given how scarce power is and the regulations coming down the road — where, state by state, they'll essentially begin capping power below 50 megawatts — I just don't foresee a situation where you'll see a campus of 500 megawatts or one of 1 gigawatt. My view is that the approach leans toward campuses of 50 megawatts and 25 megawatts, and those are quite large when you're running GPU dermatan — the revenue potential is very large. With bare metal, as things stand today, we're able to book near-term capacity at $7.75 per hour per method. On the token side, that works out to roughly $15 to $20 per hour equivalent per token. So pricing for GPU as a service is surging, and that stands to reason, since the bottleneck — the difficulty, the limiting factor — is permitted available power. The reason this opens a door for a business of our size is that we actually hold the power assets and grasp every layer, from generating electricity all the way to the token factory — something quite uncommon, since few firms both produce power and can sell tokens; and with each megawatt today, we know the way to get the most revenue and value out of it. Each megawatt can bring in roughly $2 million to $3 million every month, so merely switching 50 megawatts over works out to $100 million monthly, okay — and by the way, those carry extremely high margins, in the 85% to 90% range; that means what we need is to execute, and we're 95% of the way there, plus staying disciplined financially, and we're an amazing company; the market right now has no — demand is enormous and it's far bigger today than it was 6 months ago, since we moved from < > into inference, which is what's driving that demand; we currently have a lot of possible contracts that we could sign. We’re taking a very conservative approach and making certain we take advantage of how scarce our powered assets are — for example, New York is permitted and operating the Bonstill legacy business as a data center; it’s already there. So all we need to do is move it into a redundant data center. And our goal is to put together the right deal for the company. We weren’t in a hurry, because of financial discipline, and we want to complete our debt financing, and once that’s done we’ll build out our New York assets, which will be an incredible revenue asset. Demand is abundant—finding customers isn’t an issue, with a broad pipeline that includes several IG investment-grade customers. So New York has us very enthusiastic, along with the value and protection of DG Power capital. Naturally, Digi Power would like to manufacture everything—can formers, switch gears—but that would leave no capital for the focused business. Hence, by serving as a major shareholder—the largest shareholder of US data center, which is now planning to produce pad systems, which are quite adequate by the way—the timing is incredible, because you’ll see—we’re getting a lot of action on the U.S. data center side with smaller sites, since those aren’t available, and that’s what stays under the political radar. There’s a great deal happening on our end across ’27 and ’28. Incidentally, this is going to be an amazing company.

So using this modular design, we can accommodate any GPU brand—MDDR included, plus the B-300, the Verbora L4 and LA, the CerebraCSI,/CS 5, AMD, and the GB 300—everything's ready. And through the U.S. data center, we're gathering deposits from customers to support growth, together with debt financing later on. That small dilution from raising capital doesn't affect Digi Power X at all. So at this point, with the U.S. data center, our goal is to create genuine value for that business while we're also building out robotics. And we've actually got industrial robots going live this quarter of 2026, and they—those robots will be tied to industrial sites for maintenance, for monitoring temperature or humidity variation, all the basic data you need in a data center, and we're going to begin leasing these robots out to various industrial sites; so by building value there, Digi Power gets a free claim on a substantial share of the data center growth upside without taking on any of the risks that, frankly, it can't afford to take. We don't have capital allocated to go invest in new ventures. At Digi Power X, our priority is remaining centered on the core business and delivering on it; we have a three- to five-year plan, though it takes us to a very high level, and that is what we ought to do. The—it's fairly straightforward; it comes down to pure execution. Phase 1 tenant access in essentially five weeks, four to five weeks, with commissioning starting December 15 or earlier, billing starting in January, and closing of our purpose financing within the coming weeks, though in quarter 2026. We're making headway on Phase 2, with access coming in March 2027, then commissioning, going live, and billing in April 2027. We're also bringing on a handful of AI customers who'll put down very large deposits—that way we avoid dilution and can secure the strongest financing terms, including debt if we need it. This milestone will show just how well we turn our investment into recurring revenue. We think this is the optimal use of our capital and delivers the strongest ROI, given that we operate at margins of 85% to 90% with power already hooked up. That represents the optimal deployment of your capital, and we don't anticipate any deceleration across the coming three to five years; in fact, we are in contract negotiations at this very moment—a massive five-year agreement, and with very little power. So execution is possible for us. What we're seeing, then, is ongoing expansion in that sector; we simply have to ensure the financing is arranged correctly; many shareholders ring me up asking, but are you certain about Sebas, whose stock stood at $360 and fell to $100? Shouldn't that concern you? My concern will be about one alternative to NVIDIA; and the results they posted today demonstrated that they are performing excellently. Their year-over-year growth is 55%, cloud services revenue sits at 255%, and cash on hand is nearly $8 billion. Our partnership with Cirrus is fantastic — from the CEO and management all the way through the organization, they are extremely pleased with us, and we feel the same about them. There's room for us to expand further together. The sole factor holding back growth is what I cited before: we must proceed incrementally and stay financially disciplined. Still, as you're aware, we've announced 700 megawatts under contract, and 700 megawatts is difficult to secure these days. So without question, they will take additional co-location from us — 100%. Our goal is simply to build a varied customer base, which lets us command a higher valuation; and one key point needs to be clear. In March, once our data center is finished — the yard's condition, incidentally, laid out by a top U.S. designer — when that data center is complete, what a liquid cooling data center is worth today, in my view, runs roughly $30 million to $35 million, and megawatts of IT, and I'll explain the basis for my valuation. Before the election, Blackstone paid $27 million per megawatt for a data center in West Virginia. Following the election, that asset commands roughly a 20% to 30% premium, driven by scarcity and the potential for regulation in Texas and elsewhere. So a completed data center, looking at its value aside from the contract, runs $30 million to $35 million per megawatt times 40 — that gives you a sense of the worth. Subtract the $200 million of debt financing from the $1.2 billion to $1.4 billion, and you're left with about $1.2 billion in net equity. That's the value. Then there's the contracted value, which generates monthly EBITDA and guaranteed recurring revenue, and that adds another $200 million. So we have a lot of confidence that our partnership with Cerebral has really boosted our credibility team—I think Telebras GPUs are extremely strong when it comes to tokens per second, TPS inference—and given all the issues NVIDIA has, Telebras is in a very good position to expand and gain share, especially since NVIDIA holds 90% of the market; it wouldn’t shock me if NVIDIA begins losing share and Cerebral starts picking up share. And they’ll need solid partners they can rely on, and we’re one of those partners; so we’re really enthusiastic about our relationship with Saba. Any further questions? So thanks, everyone. Thank you, shareholders, for your patience and your support. I realize that watching the stock price swing up and down through these cycles can be really frustrating. Still, our commitment is strong. Our team is remarkable, and I truly want to express my appreciation to them. Take Alex in particular — he isn't one to say much, yet he puts in enormous effort, and he's the person who set the Celera deal in motion and has been handling negotiations on it since last year. He also has a number of truly impressive deals in the works at the lab at this moment, and we'll be discussing those shortly. On top of that, we have a fantastic team on the West Coast. We bought a building of 18,000 square feet in Sunnyvale, extremely close to Cerebras, extremely close to Supermicro and Google, a lovely building that we’re renovating and turning into beautiful state-of-the-art offices, where we’re going to have rates of D300, Vera, and we’ll be able to use it as a showroom and a demonstration of what we can do. We’re going to have a small token factory in there. And we’re assembling an incredible team with Jag and various engineers. Of course, again, we make sure to grow in relation to our revenue growth as a ratio of revenues. One thing I also noted — and I'll wrap up my remarks here — is that our SG&A costs dropped 30% between Q2 and Q3. We're working hard to stay disciplined in how we spend. As everybody is aware, our style of doing business is entrepreneurial rather than heavily corporate. The lawyers and accountants can handle the corporate side; when it comes to moving fast, though, our team is tiny, so we can talk things through, act, and settle on choices extremely quickly. That's how we stay competitive.

Even so, there's a solid chance for us to pick up market share in this area. Thanks a lot, everybody, and here's hoping today turns out to be a great one. Thanks, folks.

Thanks. That brings us to the end of today's conference call. At this point you can go ahead and disconnect your lines. We appreciate everyone for joining us.