Axe Compute Inc. (AGPU) Earnings Call Transcript
November 17, 2026
Earnings Call Speaker Segments
Good morning, and welcome to the earnings call for Axe Compute's third quarter of 2026. I'm Erin McMahon, and I serve as Chief Marketing Officer and Head of Investor Relations at Axe Compute. With me today are Christopher Miglino, our Chair and Chief Executive Officer, Jeremy Yaukey-Witter, our Chief Financial Officer, and Kyle Okamoto, our President. This call is being delivered as audio and slides. After the call ends, a replay along with the presentation will be posted on our investor relations site. Before we get started, today's remarks contain forward-looking statements as described on slide two. Important disclosures. Before we get started, a reminder that forward-looking statements are part of today's remarks. These cover statements regarding signed contract value, customer prepayments, deployment timing, annualized run rate, margins, and financing. They carry risks and uncertainties that might cause actual results to differ materially. Please refer to the risk factors in our most recent Form 10-K and Form 10-Q, our other filings with the SEC, and the full statements on this slide; we undertake no obligation to update these statements except as required by law. We will also refer to operating metrics such as total contract value and annualized run rate, and we will refer to certain non-GAAP measures, including Adjusted EBITDA. Definitions and reconciliations appear in this presentation and in our earnings release. With that, I will turn it over to Chris.
Hello, everyone, and thank you, Erin, and thank you for joining us this morning. I'm Christopher Miglino, the CEO of Axe Compute. During the previous quarter, we said this call might originate from our Sweden deployment. As it happens, there isn't sufficient snow there at present, so we're coming to you from warm and pleasant Miami. The Sweden installation is progressing very well, and we anticipate being there at some point down the road. We had an excellent quarter and achieved a great deal. I'll start with the figures that matter most this quarter. Revenue was $7.2 million, a 125% increase from $3.2 million in the second quarter. We hit that milestone while recording our inaugural quarter of positive operating income at $6.8 million, along with our first profitable quarter on a net basis at $11.5 million. Cash stood at $147.7 million when the quarter ended. This is the first time we're reporting results as a pure GPU-as-a-Service company, following the September divestiture of our legacy medical business. The revenue increase came from a growing customer roster, disciplined spending, and expansion of existing operations, plus shedding prior obligations tied to the legacy business; no corporate debt was added to support that incremental growth. Jeremy will take you through the specifics later in the call. The bottom line is straightforward. Revenue more than doubled, we turned a profit for the quarter, and this is only the beginning. For those just getting familiar with us, this is our business. We design, construct, and run purpose-built AI infrastructure as our primary operation. Each cluster gets co-engineered with the client, the hardware belongs to us, and our staff operates it 24/7. We work with data center owners, leasing their power and space. Every facility we build is backed by a multi-year take-or-pay agreement with an enterprise-grade customer, and that customer pays upfront. Our builds are located in the U.S. and Sweden, while we procure capacity across the globe. Our people are in conversations with data center developers worldwide. We install the newest NVIDIA GPUs, collaborating with partners such as NVIDIA, Supermicro, and Penguin Solutions, and engage with those data centers we identify around the world. A lot of our customer relationships get their start with Axe Compute Access, which is a shared-capacity product, and from there they usually expand into a dedicated build. This portion of our business functions, in effect, as a lead-generation and client-development offering. Some clients who began as small compute users have since become considerably bigger. I'd like to walk through how we fund growth, since that's been a big topic across the industry lately and it works differently at each neocloud company. It's also what lies behind the figures Jeremy will present. Each build stands alone as a project, beginning with a multi-year, take-or-pay contract that's signed. At signing, the customer pays upfront, and that prepayment covers a meaningful share of the hardware. If a project needs financing, the financing sits inside the project entity and is secured by that project's own contract and hardware. Axe Compute holds ownership of the cluster and operates it throughout the entire contract term. After that, we run the same playbook for the following build. Practically speaking, each build is under contract and the prepayment has been deposited before any hardware is ordered. Every project is designed to stand on its own, and signed contracts plus customer cash determine how fast we grow; we've talked about the big contracts signed within eight months of starting business, and these market statistics demonstrate the demand we're seeing. Spending on AI infrastructure is expected to reach $1.5 trillion during this year and $2 trillion by 2027. By 2030, data center capacity globally ought to approximately double, arriving at 200 GW. Token usage—a fair proxy for actual workloads—grew 25x over a span of 10 months. Demand is growing at a quicker pace than infrastructure can be built. That shortage is what we sell into, and we do so using contracted capacity, with a meaningful prepayment—usually 20%-40%—collected before construction begins. Demand in the market remains robust, and capacity is steadily coming online. The value we distinctly offer the market—choice, experience, worldwide reach, and a trusted partnership—resonates with our customers, and we intend to continue capturing that demand. Beyond the impressive growth we just discussed, there are three key business developments to report since our previous call. The team put in diligent effort to carry out all of them, and I'm quite proud of what was achieved. First, we divested Helomics, the AI-based cancer diagnostics operation that belonged to the public company at the time we started the AI infrastructure business. Axe Compute is now a pure-play neocloud, which translates into reduced operating costs for the company. Second, there's our Georgia cluster contract. Since we signed the Georgia agreement, its character has shifted considerably — and the shift has been favorable. To begin with, as of September 30th, the cluster is ours: 2,304 NVIDIA B300 GPUs energized and provisioned. Next, we retired the cluster’s asset-backed loan from cash we already had, and we didn’t issue any new equity. There’s a deferred purchase price that the cluster secures, and the customer lengthened the contract to five years, running through 2031. Because Axe now owns the cluster, all of the contract’s economics flow to Axe. Across the contract’s life, we anticipate about $352.6 million in revenue, and ownership should drive a substantial gain in gross margin and free cash flow. For Axe Compute, that marked a big milestone, and it’s the kind of build we intend to pursue moving forward. Everyone who helped get this transaction done has my thanks, since it was a considerable undertaking. Lastly, I'd like to highlight the deals we closed during Q3. Back in July, we announced build agreements totaling $1.3 billion spanning the U.S. and Europe as well, along with a dedicated Blackwell cluster valued at $1.5 billion over a five-year period. Since our Q2 earnings call, we've secured an additional $266 million in contracts and expansion. That lifts the total value of signed contracts over the past seven months to $3 billion. I'm revisiting those July numbers for a specific reason: Q3 revenue landed at $7.2 million, and none of the $3 billion-plus associated with our build project has been counted in that number yet. Revenue starts flowing the moment each build goes live, and Kyle is going to walk you through precisely where things stand on all of them. That said, I'll now pass things over to Jeremy, who can lay out the specifics underlying those figures.
Thanks, Chris. I’ll go over the third quarter results and the balance sheet. Unless noted, every figure is as reported in our Form 10-Q filed on November 16th. I’m really excited to call out the results for the third quarter of 2026, which are only achievable because of the excellent work our team is doing for customers. Revenue for the three months ended September 30th came in at $7.2 million, versus $3.2 million in the second quarter, a rise of 125%. All of the Q3 revenue came from Axe Compute Access. We recognized no revenue from Axe Compute Build during the third quarter. Operating income was $6.8 million, versus an operating loss of $16.4 million in the second quarter. Operating income takes in a $10.5 million gain on digital assets. We posted net income of $11.5 million, compared with a net loss of $17.2 million. That net income figure also reflects a $4.6 million unrealized gain on the equity securities and convertible note that came to us through the Helomics sale. For the quarter ending September 30th, 2026, adjusted EBITDA came in at a loss of $2.3 million. Adjusted EBITDA is a non-GAAP measure, and you can find a reconciliation to income/loss from continuing operations in the appendix to this presentation and in our earnings press release. We start with $11.7 million in income from continuing operations, then subtract $0.3 million of net interest income, add back $0.7 million in stock-based compensation, remove $9.9 million of unrealized gains on digital assets, and also remove the $4.5 million unrealized gain on the DataMEDS securities and convertible note. As of this quarter, we've changed our definition of Adjusted EBITDA to exclude unrealized gains and losses on the DataMEDS securities and convertible note as well, since those, much like the digital asset marks, are non-cash fair value adjustments that sit outside our compute operations. For the nine months ended September 30, 2026, the company posted an Adjusted EBITDA loss of $8.4 million. Now moving to the balance sheet as of September 30. By the end of September 30, total assets stood at $686.1 million — roughly a twelvefold increase over the $52.9 million recorded at year-end — with customer prepayments of $569.3 million providing most of that funding. We book those prepayments as contract liabilities, and they turn into revenue as service is delivered. Compute infrastructure for our build clusters represented our biggest asset at $366.7 million, and that figure includes the Georgia cluster we acquired on September 30. On top of that, we held $92.6 million in prepaid compute plus $38.1 million in lease right-of-use assets covering third-party capacity. Cash and cash equivalents came to $147.7 million. Contract liabilities from customer prepayments accounted for most of our liabilities, across both the short-term and long-term categories. In Q3, we recorded a liability for the deferred purchase price payable to Duos as a result of the acquisition. That obligation has a face value of $42.9 million and was carried at $36.6 million as of September 30, net of discount; it comes due by September 30, 2027, or sooner if refinancing occurs, and the cluster secures it. Stockholders’ equity stood at $46.3 million. For further information, please see our Form 10-Q filed with the SEC. Over to you, Kyle.
Thanks, Jeremy. Hi, everyone. Kyle Okamoto here, President of Axe Compute, and this is a call I've been eager for. In the next few minutes, I'll lay out precisely what we're constructing, the partners we're constructing it alongside, the customers who came on board after our previous call, the demand stacking up behind those customers, and the team making all of it happen. So let's dive in. We'll begin with what we're actually building in physical terms. As of now, Axe Compute has more than 18,000 NVIDIA B300 GPUs in delivery spread across four dedicated AI factories located in the U.S. and Europe. I want to take a moment to explain what we mean by AI factory, because for us that phrase is much more than a marketing tagline. NVIDIA defines an AI factory as infrastructure that carries out the entire AI lifecycle—covering data, then training, then high-volume inference—with intelligence as the actual product. That's exactly what we give our customers: not a pile of GPUs, not a pile of infrastructure, and not hardware. It is a complete factory. That means GPU compute, CPU compute to handle orchestration and data preparation, high-speed networking, a high-speed parallel file system for storage, purpose-built security, and the software stack sitting on top to orchestrate everything and ultimately produce tokens. Every bit of that comes as a service, and Axe Compute operates all of it around the clock, 24 hours a day. Let me take you through the four. Georgia comes first, as Chris noted, with 2,304 B300 GPUs. The cluster is energized, it's provisioned, Axe now owns 100% of it, and it sits in customer acceptance at present. Acceptance is the final gate in our process, and it's likely the one I care about most. We prove out the cluster jointly with the customer, right? It's a co-engineering effort, running their own production workflows against their performance criteria with their configurations before it moves into a steady state service. Now that's the right way to kick off a five-year partnership. Next up is Sweden. Our first AI factory on European soil, with more than 2,000 GPUs. It has entered commissioning — the stage where each node is racked, cabled, burned in, and tested before any customer workload touches it. Europe is important to us. Demand across the region is growing quickly. In its August results, NVIDIA pointed to a record 35 new AI supercomputers under development throughout Europe, and we're already on the ground there. The biggest of our factories is the U.S. large cluster: over 9,000 GPUs and 20 MW. That customer's payment is already in hand, and the project is now in the manufacturing and readiness phase—meaning our manufacturing partners are constructing the hardware while the site gets prepared for power, cooling, and network. As for the U.S. expansion cluster, it involves more than 4,600 GPUs for an existing customer who returned to order double the GPUs of their initial cluster and prepaid as well. That one is likewise in manufacturing and readiness, and some equipment is already on site and being stood up. When a returning customer doubles down, that's the strongest endorsement an infrastructure company could receive, so we feel very honored. Now, on to timing. A quarter ago, we told you that our signed book ought to be completely deployed throughout Q4 2026 and on into 2027, and that goal remains the same as of today. This slide lays out the progression plainly. Customer acceptance is in progress in Georgia, commissioning is taking place in Sweden, and manufacturing and readiness are the phases for our two U.S. clusters. Revenue comes in as each cluster finishes customer acceptance and goes into service. During that window, the run rate climbs cluster by cluster. Take a look at the pattern. Prepayments from the clients back both of our newest factories, and we build against signed multi-year contracts. These clusters are performing exactly as designed. The main point is this: we have four AI factories with more than 18,000 GPUs in total, and every single one is essentially a full AI factory provided as a service and operated by Axe Compute. How can a company of our size run four AI factories at the same time? This slide is basically the layer-cake explanation. At the middle, Axe holds what counts most — the design, the hardware, the customer relationship, and daily operations. Surrounding that center, we've assembled a broad and flexible group of partners and suppliers across every level of the stack. Flexible is the term I'd like you to keep in mind. This isn't just about raw capacity. The idea is to scale capacity up or down, letting us execute several builds concurrently without maintaining a large fixed-cost structure in advance of demand. When it comes to GPUs, systems, and fabrics, we rely on the newest and most advanced AI infrastructure from vendors including NVIDIA, AMD, Supermicro, Dell, Celestica, and a number of others. Having several system sources ensures we are never dependent on a single supplier's allocation or lead time, while still guaranteeing uniform design and architecture. On the storage side, WEKA and DDN keep those GPUs supplied with data at full speed. On the deployment and support side, Penguin Solutions and Computacenter provide integration teams whenever we need them, which lets us rack, cable, and burn in at multiple locations simultaneously; beyond that, there's sites and power — a crucial element in any AI factory. CBRE said in September that data center vacancy in North America reached an all-time low, and roughly 80% of capacity currently being built is already pre-leased; that's the reason we partner with a range of providers and use several deployment models, from co-location to owning sites, so we never have just one route to space and power; regarding financing, as Chris mentioned, we collaborate with top investment firms to place project-level capital behind every build. There are several choices available at each stage, so to speak. When a given partner's queue reaches capacity, we move on to the following one. That's the way we achieve parallelism. Each build is owned and run by Axe, while our partners scale their capacity up or down as demand shifts, which in the end lets us meet what customers need. Now, on to new business. During the three months since we last spoke, we brought in $266 million in entirely new contract value. All of the dollars shown on this slide are new. We've likewise recorded new logos, expansions, and renewals—and that's separate from, and in addition to, the total contract value expansion at the Georgia cluster that Chris mentioned, where the term was extended from three years to five years. We've signed eight new deals since the Q2 earnings call — four with brand-new customers, and four that are expansions or renewals. In our line of work, a customer coming back is the clearest signal you can get. So what did they purchase? Over 2,100 NVIDIA Grace Blackwell 300 GPUs — that works out to 30 full NVL72 racks of Blackwell Ultra, which is NVIDIA's flagship rack-scale offering and, according to NVIDIA, the platform behind its data center growth in the prior quarter. Contract lengths run from a year up to four years. One figure on the slide matters a lot: $64.5 million was prepaid at signing, out of $266 million in total contract value. That money isn't pledged — it's in hand. Customer cash arrives ahead of capacity. And who were the signers? Firms built around AI, outfits that train and run models, platforms offering inference in the cloud, plus big businesses deploying AI to launch fresh offerings for their own customers — all of them are expanding quickly, and Axe is what they pick. On the timing of billing: certain of those agreements started producing revenue during Q3, while the remainder kicks in across Q4 and the start of Q1. Over the coming two quarters, we anticipate it reaching full steady-state ramp. The finest bit: the accounts we serve now become the builds of tomorrow.
A couple of those customers are now collaborating with us to define the scope of bigger dedicated clusters. As Chris mentioned, things start with us and grow into dedicated AI factories; this slide is the one that gets me most excited. In just one quarter, our qualified pipeline grew by more than 100%, going from $5.9 billion to $14.1 billion. The count rose to 118 opportunities from 98. What customers want amounts to about 100,000 GPUs. They aren't looking to buy a few hours of compute. Their goal is to reserve dedicated AI factories for multi-year periods. Consider the composition of their requests: NVIDIA Blackwell accounts for 69% of pipeline value, while Vera Rubin makes up 16%. During the prior quarter, I mentioned that Vera Rubin would begin ramping, and that ramp is now underway. NVIDIA's two most recent generations account for 85%. An additional 14% runs on AMD hardware. In July, AMD introduced Helios, its inaugural rack-scale system featuring 72 MI455X GPUs, and shipments begin in this year's second half, with Microsoft, Meta, OpenAI, and Oracle among those adopting it early. Customers seek options, and we provide them. Broadening the view, the market is shifting in our favor. Market research shows that spending on AI-optimized infrastructure as a service—which is exactly what we offer—is expected to nearly double in the coming year, reaching about $42 billion, and then rise to approximately $66 billion by 2027, with inferencing overtaking training at some point during this year. Neocloud revenue, according to other research firms, is anticipated to grow by more than $25 billion, on a path toward $400 billion by 2031. The pipeline, as always, isn't the same as signed contracts, and not every opportunity will convert; we qualify these opportunities on the basis of known technical requirements, known stakeholders, known timing, and known commercial parameters. What matters most here is that our markets and the direction of our business are unmistakable. Customers want dedicated AI factories running on the newest GPUs under multi-year terms. That is exactly what Axe builds. During the prior quarter, I said that contracts come first and hiring follows them. Let me show you what that means in practice. Every function required by an AI factory is being built, and we're doing it with our own people. Data center strategy comes first. As Chris noted, a dedicated team is locating and locking down space and power for our upcoming builds, under terms aligned with what each customer wishes to contract. Given that data center vacancy sits at record lows, this team gives us a genuine competitive edge. Next up is operations: these are experienced folks in deployment and program management who take a signed contract all the way through design, sourcing, commissioning, and customer acceptance, and they do it across several sites at once. Then engineering: people who specialize in systems, network, HPC, storage, and hardware and know these clusters down to the cable level. Automation means tooling and telemetry that spot issues early and fix many of them on their own, which lets our support scale up quicker than our headcount does. And 24/7 support — a follow-the-sun NOC keeping an eye on every cluster at all hours, escalation from tier one through tier three, and a single accountable point of contact for each customer. So why does all of this matter so much? Because when you're operating at AI factory scale, hardware breaks every single day. During the training of Llama 3 by Meta on more than 16,000 GPUs, 419 unexpected interruptions were logged across 54 days — about one every three hours — and over half of those were traced to GPUs and their memory. The operators who come out ahead are the ones quickest at detecting, swapping, and recovering. That is precisely what this team was built for. And talent like this is hard to come by. In Uptime Institute's 2026 global survey, over half of data center operators reported difficulty finding qualified candidates. We are finding them, and they are picking us. Each customer receives 24/7 monitoring and a fully direct relationship with Axe, from first boot all the way through renewal. As my head of ops always likes to tell me, delivery is the product. Axe Compute handles operations end to end. Just to sum up: we have over 18,000 GPUs spread across four AI factories that are in the process of being delivered, a flexible network of partners and suppliers enabling us to build simultaneously, $266 million in fresh contracts — with $64.5 million of that already collected — a qualified pipeline that grew to more than double, reaching $14.1 billion, and the people in place to make it all happen. My excitement about the direction of Axe Compute has never been greater. Chris, over to you.
Thanks, Kyle. So this is where we stand at the end of the quarter. During Q3, we signed contracts worth over $3 billion. Delivery is now underway on those agreements. Acceptance work is happening in Georgia, commissioning is taking place in Sweden, and additional clusters are moving through the process. Prepayments are part of each of these, and project financing will fund the builds. Supporting all of that is a qualifying pipeline of $14.1 billion. When you set that alongside the financials — our first quarter delivering positive operating and net income, revenue climbing 125%, and $147.7 million in cash — and a company that is now focused purely on AI infrastructure, you get the picture. The fourth quarter is all about getting those builds online and converting signed contracts into revenue. A pipeline is not the same as signed contracts, of course, but we are very excited about what lies ahead for this business. Our thanks go to our customers, our partners, and the Axe Compute team. I'll hand things back to the operator so we can take some questions from our audience.
Thank you, Chris. At this point we'll move into the question and answer portion. Here's how you can take part. Should you be with us on the webcast, click the react button and raise your hand. You can also submit a question through the Q&A feature. For those on the phone, dial star nine to raise your hand. After your name is called, press star six so your line is unmuted. Please keep to one question plus one follow-up. Ben Sommers from BTIG has our first question.
Hi, good morning, and thank you for taking my questions. I'd like to ask a bit about how GPU rentals are priced in the market. Given the recent chatter about the possibility of, I suppose, stricter rules governing data center builds in the U.S., I'm curious how you figure that might affect the pricing of GPU rentals for capacity that could come online within the next, say, 6 to 12 months.
Thank you for being with us today. I appreciate the question; this is Chris Miglino. The political climate has caused the U.S. market to tighten, but a great deal of capacity is arriving and has already been functioning for the past two years or so. We reviewed all of it just yesterday. More than 75 data center opportunities are currently active for us to examine. A substantial amount of capacity will arrive within roughly the next twelve months. That said, I believe U.S. regulations that will take effect internationally will also drive considerably more activity back into the United States. The US will be the one deciding how chips get distributed beyond its borders, and as a result, a great deal more work is going to be pulled back into American data centers. In my view, that drives up what people pay for GPUs by the hour. I believe you'll begin seeing higher per-hour rates for anyone seeking to rent capacity. Even with plenty of space and power on hand, I think demand will remain strong from customers worldwide who will now have to enter the U.S. in order to rent that capacity.
Understood. That's really useful. Then just sort of — you mentioned that you're beginning to see the shift leaning more toward inference workloads. How are you thinking about positioning your power portfolio for the longer-term move in demand toward inference workloads? Thank you.
Kyle, would you like to handle that one?
Sure. I believe choice is among the core principles on which we built Axe Compute, and the reason we took that approach is that clients really need to operate in particular locations to support things such as inferencing. Instead of just informing a client, this is where we're constructing our next cluster, and if you want AI infrastructure, you'll receive this spec, this configuration, and this location, we've approached it from the standpoint of wanting to say yes to wherever a customer wants to be; and the primary reason for that, clearly, is inferencing, correct? You have to be as local as you can while still thinking global, and that's why we have a whole team devoted to traveling around the world and identifying the next locations where we can place AI infrastructure for our clients. I think that will just keep going. If you consider how the internet evolved, how cloud evolved, how edge computing evolved, how IoT evolved, all of these various waves of technology began in a very centralized way and then eventually became more and more distributed, and I think you'll witness the same pattern for AI infrastructure. So, the capacity to build a very fast and distributed set of AI infrastructure. Again, I mentioned elasticity earlier, and we clearly talk a lot about choice. I think those are two very important tenets moving forward for the AI community.
That's really helpful. Thanks, guys, for taking my questions, and congratulations on all the progress.
Thanks. I appreciate that.
Thanks for that, Ben. We’ll take our next question from Kevin Dede, who is with JonesTrading.
Hey, Chris, Kyle.
Hey, Kevin.
Appreciate the invitation to be part of this call.
We're glad you could be with us.
Congratulations. Yeah, no, congratulations. It's great to witness all that acceleration after everything we've been through together previously, Chris.
Yes. Thanks.
Could you shed a bit more light on who your customers are? Part of the pushback I hear from investors is that transparency is limited, and while I realize you have confidence in their financial position, that picture isn't always clear to those of us on the outside.
Yeah, I mean, obviously this entire sector—just look at all the deals completed here and how often the customer is listed as undisclosed—because in our industry it's extremely common for the customer to limit knowledge of their identity; they don't want people knowing where their data center sits or what they're doing there, particularly in cases like ours where we're mostly the sole customer at a given data center at any one time. So for the customer that's a huge security concern, which is why we're bound by NDAs that prevent us from disclosing any of that information. So that matters a great deal to our customers. But I think the best lens is this: who can write a check for $300 million, $500 million, $120 million? Those are the kinds of customers doing this. I can tell you they're S&P A+ rated. It's hard for us to name exactly who the customers are given the nature of our agreements, but that's something across the whole industry, not something specific only to us.
Okay, I understand what you're saying. A lot of other firms have shared the very same feedback with me. I've got that completely clear in my mind. I'd imagine it's a mix of enterprise clients and AI-forward organizations as well. Could you maybe shed some light on which direction that blend usually goes?
Sure. Enterprise accounts make up the majority of our customers. Most of what comes into our Axe Compute Access program, meaning the ones that are somewhat smaller in nature. By smaller, I mean that's because of what a billion-dollar transaction looks like. Several of those deals lean more toward AI-native transactions.
One quick thing before I go, Chris. Sorry about that.
Of course.
Qualified-
Go ahead and ask as many as you'd like. Ask as many as you'd like. There's no cap.
All right, well, thank you. You know the way I am. I'm interested in that qualified business you said grew from $5.9 billion to $14.1 billion. Technically speaking, what does qualified mean?
Kyle, would you like to handle that?
Absolutely, glad to do that. What counts as a qualified opportunity is that we understand their technical needs. We know who their stakeholders are, who has influence over the decision, and who actually makes it. We know the timing as well, correct? When are they looking to sign? When do they want to be operational? We grasp the commercial parameters too, correct? Target pricing, prepayments, term lengths, and so on. We have had multiple conversations, and the wonderful thing about this industry is that our customers are extremely intelligent, correct? They know precisely what they want. That is because their businesses run on top of this infrastructure, and for them it has turned into a strategic imperative. As a result, they also devote a great deal of attention to these opportunities. Scaling out infrastructure for AI has become a board-level discussion at many of these companies. Consequently, their requirements are very detailed. They know precisely what they want, when they want it, and where they want it. Once that information is clear to us, we regard that opportunity as qualified.
Thanks, Kyle. Since Chris gave me the option, I'll take one more bonus. Now that I have your attention: you've got 18,000 plus GPUs headed to customers, but the hard part to see is the power. I know you've expanded your data center relationships past Duos. Congrats on that. I'm just wondering whether you can quantify how much power you figure you'll be consuming by the end of this year and by the end of 2027.
Yeah, sure. For that 2,300, Georgia is 5 MW. For that 2K, Sweden is also 5 MW. The larger U.S. cluster comes in at 20 MW, and the U.S. expansion is 10 MW for their initial deployments. At other facilities we've also locked up a large amount of double-digit megawatts, plus the ability to scale those clusters beyond their initial deployments. One thing we've found matters a great deal when selecting a site is the ability to grow. All of our customers are thinking well ahead into the future, and as I mentioned before, they've fully realized that AI infrastructure is a strategic imperative for their business, and they don't want to be inhibited on growth; those are initial deployments with room to run.
Great. Thanks so much, gentlemen. I truly appreciate the reply and the chance to be on the call. I'll hand things over now. Thanks once more.
Appreciate the question. Many thanks, Kevin. That means a lot.
Okay, thank you, Kevin. That brings us to the end of today's conference call. We appreciate everyone joining us, and if you took part, a copy of this presentation along with a recording will be posted on our website, and you're free to disconnect now.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Axe Compute Inc. transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.
Get an API key View API docs →For developers and AI pipelines
Programmatic access to Axe Compute Inc. earnings transcripts and 132,000+ others is available through the
EarningsAPI REST API and the hosted MCP server.
Quarterly plans from $162 - full transcripts, speaker segments, full-text search,
and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.