Fermi Inc. (FRMI) Earnings Call Transcript

Hello, everyone, and thanks for being here on Fermi America's Q2 2026 earnings call. Joining me are our Chairman of the Board, Marius Haas; our Chief Commercial Officer, Anna Bofa; our Chief Operating Officer, Jacobo Ortiz; and our Chief Financial Officer, Rob Masson. Ahead of getting started, please note that forward-looking statements are part of what we discuss today, as defined under the federal securities laws. Those statements represent management's present expectations, and they carry risks and uncertainties that might make actual results differ materially from what we anticipate. If you'd like details on those risks, take a look at our latest annual report on Form 10-K along with our later reports on Form 10-Q. Any non-GAAP measures mentioned today are meant to give you a supplemental view of how the company's ongoing operations are performing. With that, Marius, over to you.

Appreciate that, Barry, and good morning to everyone. Glad you could join us. On our Q1 call, only a quarter ago, I did something unusual for a Chairman. I passed along a list — a disciplined 90-day plan built around five specific commitments — and I asked you to hold us to it. I told you the new leadership team would deliver, and that you should assess us not on our words, but on our results. Several of you wondered why I'd placed such aggressive, self-imposed deadlines on the team while external distractions and internal changes were going on. Absolutely, that came with risk. Even so, here's what I firmly believe: when you've got an exceptionally capable group of people, and they're focused on a single clear goal with priorities they grasp and timelines they can see, they're capable of extraordinary things. Clarity brings alignment, and you get both momentum and heightened focus. That focus eliminates anything irrelevant. The momentum was always about one thing: serving our shareholders well. So here I am, about 90 days later, ready to update you on where we stand — and I won't gloss over it: Fermi is in a much stronger position today than it was then. As of now, all five commitments have been fulfilled, including the one that matters most. Let me walk you through what we've achieved. First, let me briefly revisit the wider economic landscape, since that's the backdrop for the problem Fermi was built to solve. With AI compute, electricity is the true bottleneck — not capital, and not customer demand. Whether you're looking at hyperscalers, neocloud providers, or frontier model developers, the same hurdle surfaces repeatedly: securing reliable power at massive scale, on a timeline that matches the pace at which AI and data centers are being constructed. Globally, announced projects are reportedly falling behind, and the causes are grid interconnection queues plus trouble sourcing equipment. Fermi was created precisely to relieve that strain, and what we got done in the quarter you just saw proves out our strategy. Our 90-day plan's first goal was securing a binding customer agreement. Nothing mattered more, and the team delivered. A binding agreement was executed for up to 650 megawatts. At present we're working to finalize the backstop arrangement behind it. Anna will give you the details shortly. I'll just say that the caliber of those backing our advanced energy and AI ecosystem speaks volumes about the value we've been outlining since the IPO. Goal number two was identifying the right CEO to steer Fermi through its next chapter, spanning 2028 and 2029. So we brought in Lee McIntire, and that appointment takes effect this Tuesday. Lee came onto our Board in September 2025, and he already possesses a deep, practical grasp of the business as well as the project. Beyond that, his entire professional life has revolved around pushing large-scale projects over the finish line, and at this moment our most essential task is getting power delivered to the site. Lee has led some of the toughest engineering and construction firms anywhere — among them Bechtel, McDermott and TerraPower — and at each of those organizations, responsibility for delivering major projects at scale rested with him. So regarding achieving first power at the site in 2027, nobody is better positioned than Lee, in my view. He'll be working side by side with Anna, Jacobo, Rob and myself. Running this business the same way we did across the last three months is our plan. Our approach hasn't changed one bit, and everything will only speed up. Objective number three was advancing strategic partnerships. And just like before, we came through. During the quarter, we publicly announced our EPC partnerships — one with Primoris for balance of plant work on Phase 1, plus another with TSK for early works engineering on Phase 2. Both of those rank among the industry's most capable and well-regarded contractors. And more recently still, we formed a strategic alliance with Hillcore, a top-tier global infrastructure and power provider possessing extensive experience, which will speed our ability to deliver power quickly to large-scale customers. That alliance brings an additional 2.6 gigawatts to the Project Matador campus across the coming 3 years. We know of no other site with the equipment and capability to deliver 4.8 gigawatts in that timeframe. Demand for power delivery is already here, and we have both the team and the equipment required to deliver it. Fourth, we committed to strengthening our liquidity. In July, we closed an upsized $431 million convertible notes offering, which substantially reinforced our balance sheet. Equally important, we accomplished this through a low coupon and a capped-call structure intended to protect our shareholders from dilution. Rob will take you through the specifics. Goal number five, the last item in our 90-day plan, was about moving the site nearer to being powered up. We've achieved significant progress there, and that momentum continues building. This past July, three F-class turbines from Siemens Energy reached the Port of Houston, and by now they've cleared customs. Site development work has carried on as well. Jacobo will soon walk everyone through the complete operational picture. Now let me put those five milestones into the wider context. Consider any single one of them alone — a signed customer contract, a CEO, major partnerships, a stronger balance sheet, turbines sitting on site — and just that one would have made the quarter exceptional. We pushed the team to stretch, and inside one quarter they delivered on every one of them. That's this quarter in a nutshell. Our leadership group kept its attention on the strategy and executed it extremely well. That shows the kind of organizational discipline Fermi 2.0 was built to provide. The Board and I take tremendous pride in the way the team united and, unquestionably, delivered a strong quarter. As the site continues to progress and the commercial landscape grows clearer, I'll hand things over now to Anna. Looking further out, we are also keeping an eye on 2028 and 2029.

Appreciate that, Marius, and hello everybody. Let me open with what matters most regarding this quarter. Our project wasn't merely proven viable in theory by the market — the market committed to it. We're just getting started here. What customers require is dependable power at massive scale delivered on faster schedules, and Fermi was designed precisely to provide that. Lately we've been in active talks with about twelve potential clients and strategic allies. Those conversations have dug into real specifics — capacity planning, the order in which things get delivered, and the commercial structures needed to turn interest into signed agreements. Amid that wider effort, TensorWave emerged as a notably strategic initial client for Fermi. Our thinking here was an ecosystem entry play. First off, speed is essential. TensorWave is a very entrepreneurial outfit, competing in one of the quickest-expanding segments of AI infrastructure; they can decide fast and decisively, and importantly, their demand trajectory lines up with how we bring power online. Second, we see TensorWave as giving Fermi a distinctive spot within the wider AI ecosystem. Having a direct tie to a hyperscaler is enormously valuable, and we're cultivating those as well, though that's essentially a connection to just one customer. TensorWave occupies the space between the semiconductor world and the end users of that compute, which hands Fermi exposure spanning several layers of the market. Compute underpins the entire AI stack. And because TensorWave is so close to that layer, we get earlier insight into where demand is building, what infrastructure those workloads need, and which customers will likely require capacity afterward. Via a single anchor relationship, we obtain connectivity not just to TensorWave, but additionally to the chip ecosystem and to numerous hyperscalers, enterprises and AI firms that might eventually use that capacity. To us, that carries strategic weight; our aim is for Fermi to turn into a hub where the wider AI infrastructure ecosystem converges, rather than just a campus dedicated to one client. Third, the fact that this relationship unfolds in multiple phases strengthens that opportunity. The arrangement could eventually bring various workloads and end users onto the campus, which gives additional customers and partners firsthand exposure to what Fermi can do. Phasing also counts from an execution perspective. By developing in stages, we're able to align capital spending more tightly with delivery schedules, cutting down on complexity and keeping our capacity to advance rapidly toward first revenue. And by teaming TensorWave with a highly skilled data center partner, we secure direct insight into who can engineer and build these facilities swiftly and efficiently — abilities that can be redeployed for upcoming Fermi customers. Fourth, this deal confirms how distinctive what we're constructing truly is. Fermi brings together sizable power blocks, secured infrastructure, faster delivery and land with permits in place. TensorWave presented the chance to set commercial terms that, in our view, firmly capture the worth of that package while still being competitive for the customer; so choosing to work with TensorWave was a calculated call. Our search was for an anchor customer capable of matching our pace, drawing power according to our delivery schedule, backing an economic structure that properly captures what this differentiated asset is worth, and — crucially — broadening Fermi's connections throughout the wider AI ecosystem. Now, on to the agreement in question. We've executed a binding contract with TensorWave covering as much as 650 megawatts of capacity, assuming every expansion option gets exercised. Three phases make up the structure of this agreement. The first phase consists of a 15-year commitment covering 222 megawatts of gross power via a turnkey data center solution, and it amounts to roughly $6.5 billion in revenue for Fermi across that initial term. We're in the process of finalizing the backstop agreement that supports certain lease obligations. That counterparty will be named once the agreement is finished, something we expect within the next few days. A commitment at this level comes after thorough diligence covering our power strategy, permitting, secured equipment, land control, construction plan, delivery schedule and execution capabilities. So when we speak of validation, we're referring to something quite specific: a sophisticated customer along with its partners has assessed what we're building, accepted commercial terms we consider competitive and reflective of our offering's worth, and committed to the campus over the long term; and crucially, we think this first commitment reinforces the ecosystem surrounding Fermi and builds a sturdier base from which to scale as talks continue with several other customers. With that, I'll hand things over to Jacobo to explain how this plays out on the ground.

Thanks, Anna, and good morning to everyone. What I want to cover today is straightforward. All of what Marius and Anna laid out is supported by actual steel that's already in the ground, plus a seasoned team prepared to get it done. During the past 12 weeks, we've sharpened how we'll execute and deepened our EPC ties with top-tier contractors, and once project finance closes, our team can break ground immediately. Speed to power remains our priority as customers sign leases, and we won't lose the capital discipline that ties big spending to commercial milestones. Our execution group has partnered with TensorWave to bring its turnkey project in on schedule. To reduce the risk of execution problems, Fermi is teaming up with a recognized data center contractor that can access essential components and has extensive multi-project know-how in reliably building projects like these. We'll share news of this partnership soon. Let me first walk you through power delivery, then cover our strategic alliance with Hillcore. By the fourth quarter of 2027, Fermi intends to bring online 640 megawatts of nameplate capacity, and the initial 210 megawatts should be available July 1, 2027. Beyond that, Phase 1 covers 6 Siemens SGT-800 turbines under Primoris's lead, along with 3 GE 6B units and 7 TM2500s under Relevant Power Solutions's lead, all running in simple cycle, plus 200 megawatts of utility power supplied by Xcel. We're still advancing toward our power delivery targets for '28 and '29. Three Siemens F-class units, each rated above 242 megawatts and capable of up to 728 megawatts in simple cycle mode, reached the Port of Houston several weeks back. TSK, the biggest power-focused EPC company in Spain, is handling fast start engineering for those turbines. Among heavy-duty turbines on the market, the F-class units rank among the most commonly deployed. They can ramp at 40 megawatts per minute, hit full speed in roughly 5 minutes, and deliver global fleet reliability of almost 99%; that mix of high output, quick starts, and dependable performance is precisely what AI and advanced computing demand. We have those units in our project, and we're prepared to execute. Now let me shift topics, since speed to power sits at the heart of our strategy, I'm glad to take you deeper into our strategic alliance with Hillcore Energy Partners. Hillcore and its affiliates, JV Driver among them, carry over 35 years of experience in heavy industrial work and power generation, covering natural gas, steam, biomass, and power boiler facilities, from project development and construction all the way through commissioning and operations. Through this alliance between our 2 companies, roughly 2.6 gigawatts of added power comes to the site under a build, own, operate, transfer structure, which doubles Project Matador's overall power output to 4.8 gigawatts. You can think of Hillcore as a tenant paying rent while supplying extra power to Fermi's private grid. This speeds up and expands our revenue by growing total power supply so we can satisfy rising customer demand faster than a self-build approach alone could. And Fermi retains 100% of the income generated from base rent. Per this arrangement, Hillcore is the one funding, constructing, holding ownership of, and running that facility of theirs. That means Fermi brings in purpose-built, behind-the-meter electricity without any capital spent up front. Fermi serves as the anchor off-taker, yet our commitments only happen once tenant leases get signed, and every power block gets activated by demand under contract from a tenant; each PPA aligns with the duration of whichever lease sets it off, plus fuel expenses flow through to that tenant too. Such alignment eliminates stranded capacity exposure, while construction, operating and performance risk remain with Hillcore. On top of that, we keep an option to buy the facility at fair market value once year 10 passes. Initial power is aimed for delivery inside 24 months from the notice to proceed, and crucially the Hillcore build-out proceeds simultaneously with our own, taking our capacity up twofold to 4.8 gigawatts of total power available. We possess a fantastic site, a 6-gig air permit, power equipment and a seasoned construction team along with trusted strategic partners prepared to carry out Project Matador's mission. For AI, power is the gating asset, and Fermi stands ready to provide what our customer requires, more power now. Fermi is where AI gets power; having said that, Rob will walk you through our strengthened financial position. Thank you.

Thanks, Jacobo. What stands out most to me since the previous earnings call is a much healthier balance sheet, and that comes down to the convertible notes we brought to market a month ago. We completed an upsized offering exceeding $430 million in 5% convertible senior notes maturing in 2031; that amount reflects the initial purchasers' option being fully exercised for another $56 million, which clearly shows how much demand this paper drew. After accounting for the capped call, net proceeds came to roughly $382 million. I'll take you through the reasoning in 4 parts. To start, timing; our attention is on running the business, not on trying to time the market. We brought in capital with caution as the guiding principle, so we'd have operational flexibility and financial runway to pursue strategic objectives. Next, negotiating position. A healthier balance sheet allows us to push customer and partnership talks forward from strength rather than while under financing pressure. Third, structure; we intentionally built the structure with shareholders in mind. Our view is that pairing the convertible note with a capped call shields existing shareholders from dilution before catalysts hit, up until the share price exceeds double its current level; and fourth, cost of capital. A 5% coupon across a 5-year term is an appealing rate, and that time line aligns with our development and power delivery plan. Connecting this to the 90-day plan, it satisfies the liquidity goal, and it's one source out of several. As we enter the next phase, what we say about funding the project and capital stays the same: careful sourcing, disciplined deployment, aligned with commercial progress. With that, I'll hand things back to Marius for closing remarks.

Thanks, Rob. Fermi 2.0 has consistently centered on two elements coming together: a physical asset foundation exceeding $1.5 billion that's already constructed in the ground, plus the organizational capacity now in place to unlock its complete worth. During this quarter, you see those two elements converge. Within just one quarter, we secured a binding customer contract with a top-tier AI compute player, established a fresh strategic partnership alongside premier contractors, improved liquidity under shareholder-friendly conditions, brought in the appropriate CEO to steer Fermi's next growth phase, and achieved significant on-site project milestones that safeguard our speed-to-power edge. Yet I want to close where I started, with the team, since the key aspect of this quarter isn't any individual milestone. It's that we established bold targets and achieved enormous progress. That captures the whole essence of Fermi 2.0. This is an enterprise that doesn't waver when news cycles shift. The engineers laying our foundations, the groups handling our supply chain and permitting, the commercial staff finalizing agreements, they demonstrated this quarter what institutional discipline and outstanding execution look like. We think we hold the market's single most valuable power and compute site. And as one team, we're united behind a single overarching aim, maximizing long-term value for our shareholders. The path forward is challenging. We must scale this remarkable business, finalize the existing customer agreement, bring in additional ones already well advanced in the pipeline and concentrate on flawless power delivery to the site; but we started this year's second half with a signed anchor customer, an even stronger leadership team that has demonstrated it can deliver, improved liquidity, power progressing toward the site and we start with momentum. We're happy with our position, yet never completely content. There's always room to improve, and we will. We anticipate sharing updates on our ongoing progress as we carry out the Fermi 2.0 vision. Operator, we're ready for questions now.

Congratulations on everything you've accomplished so far. First, I'd like to ask about TensorWave and that lease arrangement. In what way should we consider the portion of that related to passing through power costs? What makes up the $6.5 billion figure? And perhaps how are you approaching the hedging of power that you would supply straight from the standpoint of your own private grid, since that represents a distinctive element of your project?

Operator, we're able to hear you. Operator, are you present?

I think we're ready for the next question.

I just wanted to dig into something — you all laid out, when the TensorWave deal was announced, that $1.5 billion had already been spent. So when we consider the ramp in spending toward that first 222 megawatts, plus the additional 640 megawatts, could we get a sense of how much spending still lies ahead?

Sure, glad to address that. Looking at cash for the quarter, it dropped roughly 50%, which is right in line with the general preparation we described. Regarding the deal we recently signed, there are particular cash flows tied to it; and we've previously discussed deploying capital in a disciplined way; so our outflows will be aligned with actual signed deals and incoming funds. Jacobo also mentioned this same point. The site, as we discussed, is finished with Phase 0 and moving into Phase 1, and that lines up nicely with the new TensorWave deal. Did that cover your question? Or is there something else you'd like to ask? I want to be sure I addressed everything.

No, that seems reasonable. I believe that wraps things up. Excellent. And then, when we consider as well, I suppose, clearly, at this point the complete management group appears to be set. And when we think about this sort of a move toward a turnkey offering on the data center side rather than simply leasing something like access to the power. When we sort of consider that, I suppose, what makes you folks confident about the capacity to sort of carry out that broader, more comprehensive sort of build-out, so to speak. < >

What we intend to do is collaborate with a top-tier data center partner to provide that portion of the commitment. And we have already held discussions to secure that opportunity, and we feel entirely confident that we are teamed with the correct partner to provide that data center.

Nick, it's Marius here. From a context standpoint, I want to build a bit on what you said regarding the team. Without question, we will keep growing and expanding the essential, key parts of this organization so that we can deliver flawlessly on the commitments we've made to our tenants and our customers. Our executive compensation for officers has been deployed. As you saw, we've also rolled out both the long-term incentive program and the short-term incentive program for our employees. So everything is now in place for us to scale this organization, and we clearly recognize that certain functions will need to grow so we can deliver the full scope of what we're describing. Having our first offering be a turnkey solution definitely positions us to wrap our arms around what a total offering could look like, and it will give us the experience and the opportunity to keep doing this on an ongoing basis. That said, as Anna pointed out, that won't be the pervasive model going forward. Does that make sense?

Great. That is quite straightforward.

Congratulations on reaching all those milestones. To begin, I have a few follow-ups regarding the TensorWave transaction. Could you provide a total CapEx figure that you believe is needed to satisfy the initial phase of that contract? And perhaps whether any money has already been spent toward fulfilling that agreement? Also, is there any way you could help us understand what margins are anticipated on the $6.5 billion? And finally, when do you anticipate something like full commencement of the first roughly 200 megawatts? Then, if I may ask a follow-up, perhaps a question for Anna. You mentioned certain discussions with other tenants. Is there any additional detail you can offer regarding how advanced you might be with a second tenant? I suppose, continuing from that last question, are the neoclouds the sort of category we should expect going forward among possible tenants? Or any other detail about what we might expect later on, or the timing or quality of potential tenants you could be speaking with?

Yes, I would be glad to begin. We will discuss the market broadly. In general, if you consider it as 2 phases, a standard battery CapEx project falls within the range, or the power side falls within the range, of $3 million to $4 million for each megawatt of gross power capacity, and this one can be considered within that range. As for the data center itself, typical CapEx for a turnkey solution, as you will find in the market, is roughly $10 million to $12 million for each megawatt; and once again, this project will fall within that range; we are not revealing an overall total cost figure for the project; and we are providing a precise number now, not to establish expectations, but I believe if you consider those market comps, that is what we are looking at for this project; so essentially right in the middle of what you would anticipate.

Absolutely, and I can address the latter portion of what you asked. Regarding when things get underway, our expectation is that construction along with the beginning of revenue will proceed in stages. We're looking at roughly the close of Q3 or the start of Q4 in '27; and once more, delivery happens in phases. As soon as the initial portion of the build is finished, the customer begins paying rent, and that continues right through until the entire 220-megawatt build is done. On your separate question about providing some detail on other customers, we are definitely in active talks with additional customers, and some of those discussions are quite advanced. When it comes to neoclouds, our clear emphasis is on partnering with customers who are prepared and able to absorb the capacity we have, though we are certainly also in discussions with hyperscalers. Something that stood out to us about this particular deal is that, through our ongoing talks with hyperscalers about various opportunities, there's a promising route here where they might secure capacity earlier by teaming up with TensorWave, given that TensorWave has these 3 possible buildings; as a result, we're pursuing both types of conversations — hyperscalers could deal with us directly to arrive at the site, which interests them a great deal, and they might also seize the chance presented by the multiple buildings TensorWave now has at its disposal, which could get them on-site somewhat sooner. Still, what remains abundantly evident is that demand is enormous. Opportunities comparable to Fermi are extremely scarce. Sites free of permitting problems that still offer capacity are extremely scarce; that's precisely what has drawn every one of the hyperscalers to talk with us, and we're quite happy with the direction those discussions are taking.

Nice work on such a packed 90 days. For me, I've got two things — the second one builds on what was asked before. Earlier on, our assumption was that you generate roughly $700 million to $900 million in EBITDA for each gigawatt you deploy annually; I'd just like to know whether that still seems like a sensible thing to expect moving forward? Then my other item: regarding the TensorWave contract and similar ones, could you walk us through the sorts of risk factors involved, plus the main obstacles we ought to track when it comes to time to power and what liability you might face depending on how the schedule plays out?

I'll begin with risk and then get into the earlier part of your question. Regarding the risk factors, when you're working through the process of getting to a signed agreement, a key step is figuring out ways to lower your exposure; so our capacity to bring a data center partner into the mix definitely cuts risk in how the deal is structured; that's the reason we ran diligence to figure out whom we viewed as the strongest possible partner. And above all, we found a partner capable of hitting the timeline. So what you do is make sure you've aligned the power coming online, the data center being delivered, and in the end, where the agreement lands regarding when those 2 elements converge, so that risk goes down. That's exactly the strategy we used for the TensorWave deal, so we'd have visibility into confirming that what we committed to was something we could actually deliver, and we have full confidence in our capacity to do so together with the data center partner.

Now, regarding the EBITDA piece — I can address that. We aren't able to discuss it in detail, nor do we plan to, but based on the total contract value and the nature of this agreement, it closely mirrors other market deals that have been made public, and the profit expectations align with those. There's a modified net structure in place here, which makes for a strong setup, and as our first contract out of the gate, we're happy with it.

Steve, this is Marius. I want to build on that with a few items you can anticipate. What's coming very soon is the reveal of the backstop involving a top-tier IT infrastructure provider. Next in line, clearly, would be the project financing piece. Once that's in place, we can shift everything over to the site and prepare to deliver on the power. So those important milestones are definitely where the team's attention will be, yet all of them are already progressing well and approaching quickly.

Right. And to build on what was just said, let me go a bit further. Once more, our goal is to be completely transparent about the way we bring the risk down; and naturally, with a transaction of this kind, an investment-grade backstop plays an essential role in that, especially when you move into project finance; so a great deal of effort has gone into making sure it is arranged and prepared to move forward. And as was noted before, we expect to be in a position to reveal that partner within the next few days.

Let me just tack on one more quick item. Earlier you suggested — or at least that's how I took it — that on the CapEx side, what comes in and what goes out are going to line up. So does that imply not much in the way of Fermi cash is required to get this built, with the customer covering a portion of the working capital? How should we frame that?

Yeah. What I believe we were getting at is that we've been doing this in stages, and we said we'd hold off on the following development phase until a contract was in hand and we were ready to move. So Fermi moved quickly at the start. We locked up assets, and that's what handed us the edge on speed to power. The way we built out the site lets our schedules align nicely, which means power stops being the constraint on standing up and serving customers; that's the point we were making. Then I'd say every individual project carries its own need for capital deployment. We'll consider a range of funding options for that, the usual ones. So really we're saying that when we kick off projects, we'll place all the capital behind them to construct that particular project and keep Project Matador on track.

I wanted to give us one more shot at raising that same issue — specifically, how power pass-through from the front of the queue gets handled within the TensorWave deal, and in what way are you offsetting fuel exposure, given the private grid piece at this location?

Yeah. And my apologies for the technical issue from before. To put it plainly, the structure of this kind of deal works like this: we charge a base rent, we charge a set amount for power, and after that, the variable energy charge — which does involve risk, since it can rise — gets passed through entirely to the customer. And that's something we intend to keep doing across every one of our deals. That variable energy charge will always be taken by us and passed through in full to the customer, which brings that risk down.