๐ŸŽ Black Friday: 50% off Unlimited and 30% off Pro annual plans
Axe Compute Inc. (AGPU)
NASDAQ: AGPU ยท Real-Time Price ยท USD
6.89
-0.09 (-1.29%)
At close: Nov 18, 2026, 4:00 PM EST
7.20
+0.31 (4.50%)
Pre-market: Nov 19, 2026, 8:21 AM EST
โ† View all transcripts

Earnings Call: Q3 2026

Nov 17, 2026

Summary

Q3 revenue rose 125% QoQ, with positive operating and net income, though both included unrealized gains. More than $3B of signed build contracts is progressing toward deployment, while the qualified pipeline reached $14.1B.

Erin McMahon
CMO and Head of Investor Relations, Axe Compute

Good morning, and welcome to Axe Compute's third quarter 2026 earnings call. I am Erin McMahon, Chief Marketing Officer and Head of Investor Relations at Axe Compute. Joining me today are Christopher Miglino, Chair and Chief Executive Officer, Jeremy Yaukey-Witter, Chief Financial Officer, and Kyle Okamoto, President. Today's call is audio and slides. A replay and the presentation will be available on our investor relations site after the call. Before we begin, today's remarks include forward-looking statements as referenced on slide two. Important disclosures. Before we begin, a reminder that today's remarks include forward-looking statements. These include statements about signed contract value, customer prepayments, deployment timing, annualized run rate, margins, and financing. They involve risks and uncertainties that could cause actual results to differ materially.

Please see 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 are in this presentation and in our earnings release. With that, I will turn it over to Chris.

Christopher Miglino
Chair and CEO, Axe Compute

Good morning, everybody, and thank you, Erin, and thanks for joining the call this morning. I am Christopher Miglino, the CEO of Axe Compute. We told you last quarter we might do this call from our deployment in Sweden. Well, there is not enough snow there yet, so you get us from nice, warm Miami. The installation is going great there in Sweden, and we look forward to being there in the future. We had a great quarter, and we accomplished a lot. I will start with the numbers that matter the most this quarter. Revenue was $7.2 million, up 125% from $3.2 million in the second quarter. We did that while posting our first quarter of positive operating income of $6.8 million and our first quarter of positive net income of $11.5 million. We ended the quarter with $147.7 million in cash.

This is our first quarter posting results purely as a GPU-as-a-Service business, following the sale of our legacy medical business in September. That revenue growth was funded by our growing customer base, responsible spending and growing our current business, and divestment of previous obligations from the legacy business. We did not add corporate debt related to the incremental revenue growth. Jeremy will take you through the details later in the call. The headline is simple. Revenue more than doubled, the quarter was profitable, and we are just getting started. For anyone new to the story, here is what we do. Our core business is to build, own, and operate dedicated AI infrastructure. We co-engineer each cluster with the customer, we own the hardware, and our team runs it around the clock. We partner with data center owners to rent their power and space.

Every build that we do sits on a multi-year take-or-pay contract with an enterprise-grade customer, and the customer prepays. Our builds are in the U.S. and Sweden, and we source capacity worldwide. Our team talks to data center developers all over the world. We deploy the latest NVIDIA GPUs, working with partners including NVIDIA, Supermicro, and Penguin Solutions, and work with those data centers that we seek out globally. Many customer relationships start with our product called Axe Compute Access. It's a shared capacity offering, and they tend to grow from there into a dedicated build. You can think of this part of our business as a lead generation and client development product. We have clients that start out as small compute users that have grown much larger over time. I want to review how we finance growth.

This has been a major conversation in our industry recently and is different for every neocloud business. It explains the numbers Jeremy will show you. Every build is its own project. It starts with a signed contract, multi-year, and take-or-pay. The customer prepays at signing, and the prepayment funds a significant portion of the hardware. Where a project needs financing, that financing sits inside that project entity and is secured by the project's own contract and hardware. Axe Compute owns the cluster and operates it for life of the contract. Then we do it again for the next build. What that means in practice, each build is contracted and prepayment is deposited before we order the hardware. Each project is designed to stand on its own, and our growth is paced by signed contracts and customer cash.

We've talked about the large contracts we've signed within eight months of business, and these market statistics illustrate the market demand that we're seeing. AI infrastructure spending is forecasted to reach $1.5 trillion this year and $2 trillion in 2027. Global data center capacity is expected to roughly double to 200 GW by 2030. And token usage, which is a reasonable proxy for real workloads, grew 25x in 10 months. Usage is growing faster than infrastructure can be built. That is a shortage that we sell into, and we sell into it with capacity that is contracted, and a substantial prepayment of typically 20%-40% is received before we build. Market demand is strong, and capacity is coming online. Our unique value to the market of choice, experience, global reach, and trusted partnership resonates with our customers, and we intend to continue to capture that demand.

Three major business updates since our last call, aside from the amazing growth that we just shared. The team worked diligently to execute on all of these, so I'm very proud of these accomplishments. First, we sold our Helomics business, which is the AI cancer diagnosis business that was part of the public company when we started the AI infrastructure business. Axe Compute is now a pure play neocloud. This translates into less operating expenses for the company. Second, our Georgia cluster contract. The nature of the Georgia contract has changed significantly since we signed it, and it has changed for the better. First, as of September 30th, we own the cluster, 2,304 NVIDIA B300 GPUs energized and provisioned.

Second, we paid off the cluster's asset-backed loan with cash on hand, with no new equity issued. We owe a deferred purchase price secured by the cluster, and the customer extended the contract to five years through 2031. Owning the cluster means the full contract economics accrue to Axe. We expect approximately $352.6 million of revenue over the life of the contract, with a significant improvement in gross margin and free cash flow as a result of this ownership. This was a major milestone for Axe Compute, and it's the type of build we want to do going forward. I want to thank everybody that was involved in completing this transaction, as it was a significant undertaking. Finally, I want to call out the contracts signed in Q3.

We previously announced build contracts we signed in July that spanned $1.3 billion across the U.S. and Europe, as well as a $1.5 billion five-year dedicated Blackwell cluster. Since our Q2 call, we've added an additional $266 million in contracts and expansion. That takes signed contracts value over the past seven months to $3 billion. I'm repeating the July numbers for a reason. Q3 revenue was $7.2 million, and none of the more than $3 billion from our build project is in that number yet. The revenue starts as each build goes live, and Kyle's going to show you where we are with every one of those. With that said, I'll turn the call over to Jeremy, and he can walk you through the details of the numbers.

Jeremy Yaukey-Witter
CFO, Axe Compute

Thanks, Chris. I'll cover third quarter results and the balance sheet. All figures are as reported in our Form 10-Q filed on November 16th, unless noted. I'm very excited to highlight our results for the third quarter of 2026, which are only possible thanks to the great work our team is doing to serve customers. Revenue for the three months ended September 30th was $7.2 million, compared with $3.2 million in the second quarter, an increase of 125%. Revenue in Q3 was entirely contributed by Axe Compute Access. No revenue was recognized from Axe Compute Build in the third quarter. Operating income was $6.8 million, compared with an operating loss of $16.4 million in the second quarter. Operating income includes a $10.5 million gain on digital assets. Net income was $11.5 million, compared with a net loss of $17.2 million.

Net income also includes a $4.6 million unrealized gain on the equity securities and convertible note we received in the sale of Helomics. Adjusted EBITDA for the quarter ended September 30th, 2026, was a loss of $2.3 million. Adjusted EBITDA is a non-GAAP measure, and a reconciliation to income/loss from continuing operations is included in the appendix to this presentation, as well as our earnings press release. We start from income from continuing operations of $11.7 million, remove $0.3 million of net interest income, add back $0.7 million of stock-based compensation, remove $9.9 million of unrealized gains on digital assets, and remove $4.5 million unrealized gain on the DataMEDS securities and convertible note.

Beginning this quarter, we have revised our definition of Adjusted EBITDA to also exclude unrealized gains and losses on the DataMEDS securities and convertible note, because, like the digital asset marks, those are non-cash fair value adjustments outside our compute operations. For the nine months ended September 30, 2026, the company had an Adjusted EBITDA loss of $8.4 million. Moving to the balance sheet as of September 30. Total assets were $686.1 million as of September 30, up nearly 12x from $52.9 million at year-end, funded largely by $569.3 million of customer prepayments. Those prepayments are recorded as contract liabilities and convert to revenue as we deliver service. Our largest asset was $366.7 million of compute infrastructure for our build clusters, including the Georgia cluster acquired on September 30. We also carried $92.6 million of prepaid compute and $38.1 million of lease right-of-use assets for third-party capacity.

Cash and cash equivalents were $147.7 million. Contract liabilities tied to customer prepayments comprised the majority of our liabilities, both short-term and long-term. During Q3, we also added a liability reflecting the deferred purchase price owed to Duos following the acquisition. $42.9 million face value, carried at $36.6 million, net of discount as of September 30, due by September 30, 2027, or earlier upon refinancing, and secured by the cluster. Stockholders' equity was $46.3 million. For additional details, please refer to our Form 10-Q on file with the SEC. I'll turn it over to you, Kyle.

Kyle Okamoto
President, Axe Compute

Thanks, Jeremy. Hello, everyone. I'm Kyle Okamoto, President of Axe Compute, and I've been looking forward to this one. Over the next few minutes, I'm going to show you exactly what we're building, who we're building it with, the customers who signed up since our last call, the demand lining up behind those customers, and the team that's delivering all of it. Let's get into it. Let's start with what we're physically building. Right now, Axe Compute has over 18,000 NVIDIA B300 GPUs in delivery across four dedicated AI factories in the U.S. and Europe. I want to take a second to elaborate on the words AI factory, because for us, it's not just a slogan. NVIDIA describes an AI factory as infrastructure that runs an entire AI lifecycle, from data to training to high-volume inference, where the product is intelligence itself.

That's exactly what we hand our customers, not a pile of GPUs or a pile of infrastructure or hardware. It's a complete factory. GPU compute, CPU compute for orchestration and data preparation, high-speed networking, high-speed parallel file system storage, dedicated security, and the software stack on top to orchestrate and ultimately generate tokens. All of that is delivered as a service, and all of it is run by Axe Compute 24 hours a day. Let me walk you through the four. Georgia's first, as Chris mentioned, 2,304 B300 GPUs. Cluster's energized, it's provisioned, it's now 100% owned by Axe, and it's in customer acceptance right now. Acceptance is the last gate in our process, and it's probably the one I care most about. We prove the cluster together with the customer, right?

This is a co-engineering effort, running their own production workflows against their performance criteria with their configurations before it moves into a steady state service. That is really how you start a five-year relationship the right way. Sweden is next. Over 2,000 GPUs, our first AI factory in Europe. It is now in commission. Commissioning is where every node gets racked, cabled, burned in, and tested before it ever sees a customer workload. Europe matters to us. Demand there is building fast. NVIDIA highlighted a record 35 new AI supercomputers in development across Europe in its August results, and we are already there. Our largest factory is the U.S. large cluster, over 9,000 GPUs and 20 MW.

The customer is already prepaid, and it is in manufacturing and readiness state, which means the hardware is being built by our manufacturing partners, and the site is being readied for power, cooling, and network.

The U.S. expansion cluster, over 4,600 GPUs for an existing customer who came back for twice as many GPUs as their first cluster and prepaid as well. That is also in manufacturing and readiness, with some equipment already on site being stood up. A returning customer doubling down is the best endorsement any infrastructure company can get, so we are very honored. Now, timing. Last quarter, we told you we expected our signed book to be fully deployed across Q4 2026 and into 2027, and that remains our target three months later. You can see the progression right on this slide. Georgia is in customer acceptance, Sweden is in commissioning, and our two U.S. clusters are in manufacturing and readiness. Revenue follows each of those clusters as it completes customer acceptance and goes into service. The run rate steps up cluster by cluster through that window. Notice the pattern.

Our two newest factories are both supported by prepayments from the clients, and we build against signed multi-year contracts. These clusters are working exactly as designed. The takeaway, four AI factories, over 18,000 GPUs, and each one is really a complete AI factory delivered as a service and run by Axe Compute. How does a company our size deliver four AI factories at the same time? This slide is really kind of that layer cake answer. At the center, Axe owns what matters most, the design, the hardware, the customer relationship, and day-to-day operations. Around that core, we have built a deep and elastic bench of partners and suppliers at every layer of the stack. Elastic is the word I want you to remember. This is not only about sheer capacity.

It is about flexing up and down so we can run multiple builds in parallel without carrying a big fixed cost base ahead of demand. For GPUs, systems, and fabrics, we build on the latest and greatest AI infrastructure with suppliers like NVIDIA, AMD, Supermicro, Dell, Celestica, and several others. Multiple system sources means we are never single-threaded on one supplier's allocation or lead time, all while ensuring consistent design and architecture. For storage, WEKA and DDN keep those GPUs fed with data at full speed. For deployment and support, Penguin Solutions and Computacenter give us integration crews on demand, so we can rack cable and burn in at more than one site at a time. Then there are sites and power, a critical component in any AI factory.

CBRE reported in September that North American data center vacancy hit a record low, with about 80% of capacity under construction already pre-leased. That is why we work with a variety of partners and deploy multiple types of models, from co-location to site ownership, so we always have more than one path to space and power. On financing, as Chris covered, we work with leading investment firms to put project-level capital behind each build. So multiple options at every layer. If one partner's queue is full, we go to the next. That is how we build in parallel. Axe owns and operates each build, and our partners flex with demand to increase scalability, which ultimately allows us to satisfy customers' demands. Let us talk about new business. In the three months since our last call, we signed $266 million of brand-new contract value. Every dollar on this slide is new.

New logos, expansions, and renewals. This is in addition to the Georgia cluster's total contract value expansion that Chris mentioned, where the term was extended from three years to five years. Eight new contracts since Q2 earnings call, four new logos, plus four expansions and renewals. In our business, a returning customer is the strongest signal there is. What did they buy? Over 2,100 NVIDIA Grace Blackwell 300 GPUs. That is 30 full NVL72 racks of Blackwell Ultra, NVIDIA's flagship rack-scale system, and the platform NVIDIA says drove its data center growth last quarter. Terms run from 12 months to 48 months. An important number on the slide, $64.5 million prepaid at signing against that $266 million in total contract value. Not promised, received. Customer cash before capacity. Who signed?

AI-native companies, training and serving models, cloud inference platforms, and enterprises using AI to create new services for their consumer clients. They are growing fast, and they are choosing Axe. When does it bill? Revenue from some of these contracts already started in Q3, and the rest activates through Q4 and early Q1. We expect to see it fully ramped to steady state over the next two quarters. The best part, today's customers are tomorrow's builds. Two of these customers are already scoping larger dedicated clusters with us. As Chris said, relationships start with us and grow into dedicated AI factories. This is the slide that gets me most excited. Our qualified pipeline more than doubled in a single quarter, from $5.9 billion to $14.1 billion. 118 opportunities, up from 98. Customers asking for roughly 100,000 GPUs. Our customers are not shopping for a few hours of compute.

They want to reserve dedicated AI factories for years. Look at what they are asking for. 69% of the pipeline value is NVIDIA Blackwell, and 16% is Vera Rubin. Last quarter, I told you that Vera Rubin was going to start to ramp up, and it has. 85% sits on NVIDIA's two newest generations. Another 14% is AMD-based. AMD launched Helios, its first rack-scale system with 72 MI455X GPUs in July, and it starts shipping in the second half of this year, with Microsoft, Meta, OpenAI, and Oracle among the early adopters. Our customers want choice, and we give it to them. So zoom out, and the market is moving our way.

Market research expects spending on AI-optimized infrastructure as a service, which is exactly what we deliver, to nearly double this year to about $42 billion, and to reach about $66 billion in 2027, with inferencing overtaking training this year. Other research firms see neocloud revenue growing more than $25 billion toward $400 billion by 2031. As always, the pipeline is not signed contracts, and not every opportunity will close. These opportunities are qualified as they have known technical requirements, known stakeholders, known timing, and known commercial parameters. The key point here is that our markets and our business trajectory is unmistakable. Customers want dedicated AI factories on the latest GPUs on multi-year terms. This is precisely what Axe builds. Last quarter, I told you hiring follows contracts. Here's what it looks like in practice.

We are building out every function an AI factory needs, and we're building it in-house. It starts with data center strategy. As Chris mentioned, we have a dedicated team finding and securing space and power for our next builds, on terms that match what each customer wants to contract. With data center vacancy at record lows, this team is a real competitive advantage. Operations, experienced deployment and program leadership that takes a signed contract through design, sourcing, commissioning, and customer acceptance across several sites at the same time. Engineering, systems, network, HPC, storage, and hardware specialists who know these clusters down to the cable. Automation is tooling and telemetry that catches problems early and resolve many of them automatically, so our support capabilities can scale faster than our headcount.

And 24/7 support, a follow-the-sun NOC, watching every cluster around the clock, tier one to tier three escalation, and one accountable point of contact for every customer. But why does this matter so much? Because at AI factory scale, hardware fails every day. When Meta trained Llama 3 on over 16,000 GPUs, it logged 419 unexpected interruptions in 54 days, roughly one every three hours, and more than half traced back to GPUs and their memory. The operators who win are the ones who detect, swap, and recover fastest. This is exactly what this team is built to do. And this talent is hard to find. In Uptime Institute's 2026 global survey, more than half the data center operators said they struggled to find qualified candidates. We are finding them, and they are choosing us.

Every customer gets 24/7 monitoring and a 100% direct relationship with Axe, from first boot through renewal. Delivery is the product, as my head of ops likes to always tell me. Axe Compute runs end-to-end operations. To recap, over 18,000 GPUs across four AI factories moving through delivery, an elastic partner and supplier bench that lets us build in parallel, $266 million of new contracts with $64.5 million already in the bank against those contracts, and a qualified pipeline that more than doubled to $14.1 billion, and the team to deliver all of it. I've never been more excited about where Axe Compute is heading. Chris, back to you.

Christopher Miglino
Chair and CEO, Axe Compute

Thanks, Kyle. Here is where the quarter leaves us. We signed more than $3 billion in Q3. Those contracts are now moving through delivery. Georgia is in acceptance, Sweden is in commissioning, and more clusters are in process. Each one includes prepayments, and builds will be project financed. Behind them is a $14.1 billion qualifying pipeline. Put that next to the financials, our first quarter with positive operating and net income, revenue up 125%, and $147.7 million of cash, and a business now purely focused on AI infrastructure. The fourth quarter is about bringing those builds online and turning signed contracts into revenue. While a pipeline is not signed contracts, we are very excited about the future of this business. Thank you to our customers, our partners, and the Axe Compute team. Back to the operator, where we can take some questions from our audience.

Erin McMahon
CMO and Head of Investor Relations, Axe Compute

Thank you, Chris. We will now begin the question and answer session. Here is how to participate. If you are joining via webcast, please select the react button and then raise your hand. Alternatively, you can type a question using the Q&A functionality. If you are joining from your phone, please dial star nine to raise your hand. Once you are called, press star six to unmute yourself. We ask that you limit yourself to one question and one follow-up. Our first question comes from Ben Sommers from BTIG.

Ben Sommers
Analyst, BTIG

Hey, good morning, and thank you for taking my questions. I just wanted to ask a bit about the GPU rental pricing market. Given the recent noise around the potential of, I guess, tighter regulation market for data center development in the U.S., just kind of curious how you think that can potentially reflect on GPU rental pricing for capacity that can be online over the next, let us call it, 6-12 months.

Christopher Miglino
Chair and CEO, Axe Compute

Thanks for joining us today. I appreciate the question. This is Chris Miglino. While the market is tightening across the United States because of the political environment, there is plenty of stuff that is coming online and has been working for the last couple of years. We were just going through it yesterday. We have over 75 active data center opportunities to look at. There is a lot of capacity that is going to be coming online here in the next year or so. But I think that also the U.S. regulations that are going to go in place internationally are going to push a lot more activity back into the United States. The United States is going to control the distribution of the chips outside of the United States, and so that is going to push a lot more activity back into data centers in the United States.

I think that will increase the pricing of the GPUs per hour. I think you're going to start to see an increase in the pricing on a per-hour basis for anybody that's looking to rent capacity. Even though there's going to be an abundance of space available and power available, I think that you're still going to see a lot of demand coming from global customers that are going to now need to come into the U.S. to rent that capacity.

Ben Sommers
Analyst, BTIG

Got it. Super helpful. Then just kind of, you made a comment on you're starting to see the shift more towards inference workloads. Just kind of how do you think about positioning your power portfolio for the longer term shift in demand towards inference workloads? Thank you.

Christopher Miglino
Chair and CEO, Axe Compute

Kyle, you want to take that one?

Kyle Okamoto
President, Axe Compute

Sure. I think one of the main premises that we've built Axe Compute on is choice, and the reason why we've done that is because clients have a strong need to be in specific places to support things like inferencing. Rather than simply telling a client, This is where we're building out our next cluster, and if you want AI infrastructure, you're going to get this spec, this configuration, and this location, we've come at it from, we want to be able to say yes to where a customer wants to be. And the main reason for that, obviously, is inferencing, right? You need to be as local as possible while still thinking global, and that's why we have an entire team dedicated to going around the world and finding the next places that we can put AI infrastructure for our clients.

I think that will only continue. If you look at the evolution of the internet, the evolution of cloud, the evolution of edge computing, the evolution of IoT, all of these different waves of technologies all started out very centralized and then ultimately got more and more distributed, and I think you will see the same thing for AI infrastructure. So, the ability to build a very fast and distributed set of AI infrastructure. Again, I talked about elasticity earlier, and we obviously talk a lot about choice. I think those are two very important tenets going forward for the AI community.

Ben Sommers
Analyst, BTIG

Super helpful. Thank you guys for taking my questions, and congrats on all the progress.

Christopher Miglino
Chair and CEO, Axe Compute

Thank you. Appreciate it.

Erin McMahon
CMO and Head of Investor Relations, Axe Compute

Thank you for your question, Ben. Our next question comes from Kevin Dede at JonesTrading.

Kevin Dede
Analyst, JonesTrading

Hey, Chris, Kyle.

Kyle Okamoto
President, Axe Compute

Hey, Kevin.

Kevin Dede
Analyst, JonesTrading

Thanks for having me on the call.

Kyle Okamoto
President, Axe Compute

Thanks for joining us.

Kevin Dede
Analyst, JonesTrading

Congrats. Yeah, no, congrats. It is fun to see all the acceleration given all the things that we have worked through together in the past, Chris.

Christopher Miglino
Chair and CEO, Axe Compute

Yep. Thank you.

Kevin Dede
Analyst, JonesTrading

Can you give me maybe a little more color on your customers? I think some of the investor pushback I get is that there is not a lot of transparency, and I know that you are confident in their financial standing, but it is not necessarily clear for those of us from the outside.

Christopher Miglino
Chair and CEO, Axe Compute

Yeah, I think the obviously, this whole sector, and if you look at all the deals that get done here, how many times it says undisclosed customer, because in our industry, it is very common for the customer to restrict the knowledge of who they are. They do not want people to know where their data center is, what they are doing there, especially in situations like ours where we are primarily the only customer in a particular data center at any given time. So, it is a very big security issue for the customer. So, we get locked up with NDAs to not be able to disclose any of that information.

S o, that is something that is very important to our customers. But I think the best way to look at it is, who can write a check for $300 million, $500 million, $120 million? Those are the type of customers that are doing this.

I can tell you they're S&P A+ rated. It's difficult for us to say exactly who the customers are because of the nature of our agreements, but that's an industry-wide thing, not specific only to us.

Kevin Dede
Analyst, JonesTrading

Yeah, I understand. I've gotten the same feedback from many other companies. I completely understand that. Maybe understand there also is a mix between enterprise and AI-forward companies. Is there any color you'd maybe be able to share on how that mix leans?

Christopher Miglino
Chair and CEO, Axe Compute

Yeah. Most of our customers are enterprise. Most of the stuff that comes into our Axe Compute Access program, so the stuff that is a little smaller in nature. When I say smaller, it's because of the nature of a billion-dollar transaction. Some of those deals are more AI-native transactions.

Kevin Dede
Analyst, JonesTrading

Just a quick one before I hop, Chris. Sorry.

Christopher Miglino
Chair and CEO, Axe Compute

Sure.

Kevin Dede
Analyst, JonesTrading

Qualified-

Christopher Miglino
Chair and CEO, Axe Compute

Ask as many as you want. Ask as many as you want. No limit.

Kevin Dede
Analyst, JonesTrading

Okay. Well, thanks. You know how I can be. I am curious about the qualified business that you mentioned expanding from $5.9 to 14.1 billion. How is qualified technically defined?

Christopher Miglino
Chair and CEO, Axe Compute

Kyle, you want to take that?

Kyle Okamoto
President, Axe Compute

Yeah, sure. I would be happy to. A qualified opportunity is we know their technical requirements. We know their stakeholders, who is influencing the decision, who is making the decision. We know the timing, right? When do they want to sign? When do they want to be up and running? We understand the commercial parameters, right? Target pricing, prepayments, term lengths, et cetera. We have had several conversations, and the great part about this industry is our customers are incredibly intelligent, right? They know exactly what they want. Because they are running their businesses on top of this infrastructure, and it has become a strategic imperative for them. That means that they also have a large amount of attention on these opportunities. Scaling out infrastructure for AI is now a board-level conversation for a lot of these companies. That means that they have very detailed requirements.

They know exactly what they want, when they want it, where they want it. Once we understand that information, we consider that opportunity qualified.

Kevin Dede
Analyst, JonesTrading

Thanks, Kyle. I will go for one more bonus since Chris gave me the option.

I have your attention. You have 18,000 plus GPUs destined for customers, but what is difficult to see is the power. I know you extended your data center relationships beyond Duos. Congrats on that. I am just wondering if you can quantify how much power you think you will be consuming by the end of this year and by the end of 2027.

Kyle Okamoto
President, Axe Compute

Yeah, sure. Georgia is 5 MW for that 2,300. Sweden is also 5 MW for that 2K. The larger U.S. cluster is 20 MW, and the U.S. expansion is 10 MW for their initial deployments. We have also secured a large amount of double-digit megawatts at other facilities, as well as the ability to scale those clusters beyond their initial deployments. One thing that we found is very important in 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 have fully realized that AI infrastructure is a strategic imperative to their business, and they do not want to be inhibited on growth. Those are initial deployments with room to run.

Kevin Dede
Analyst, JonesTrading

Perfect. Thank you very much, gentlemen. I really appreciate the response and being allowed to join the call. I will cede the floor. Thank you again.

Kyle Okamoto
President, Axe Compute

Thanks for the question.

Thank you very much, Kevin. I appreciate it.

Erin McMahon
CMO and Head of Investor Relations, Axe Compute

All right. Thank you, Kevin. This concludes today's conference call. Thank you for joining, and for those participating, a copy of this presentation and recording will be available via our website, and you may now disconnect.