Home / Transcripts / Axe Compute Inc. (AGPU) · April 1, 2026

Axe Compute Inc. (AGPU) Earnings Call Transcript & Summary

April 1, 2026

NASDAQ US Health Care Health Care Equipment and Supplies earnings 29 min

What were the key takeaways from Axe Compute Inc.'s April 1, 2026 earnings call?

In the fiscal year 2025 earnings call for Axe Compute Inc. (AGPU:US), management reported a significant strategic shift with the launch of their GPU capacity platform and a rebranding effort. Revenue for FY 2025 was $125,000, a modest increase from $85,000 in FY 2024, primarily from the legacy Drug Discovery Services segment. However, management highlighted a strong start to FY 2026 with $12 million in new contracts signed in March, translating to an expected monthly contract value of approximately $850,000, signaling robust growth potential and a shift towards recurring revenue streams.

What topics did Axe Compute Inc. cover?

What were Axe Compute Inc.'s April 1, 2026 results?

Axe Compute is at a pivotal moment with strong contract signings and a strategic shift towards AI infrastructure. While the reported losses raise concerns, the underlying operational performance and market positioning suggest significant growth potential. Investors should watch for execution on contract deployments and further developments in the AI infrastructure market as key catalysts.

Earnings Call Speaker Segments

Josh Blacher executive
#1

Hello, all, and thanks for being with us this morning for the Axe Compute Fiscal Year 2025 Investor Presentation. Shares of our common stock are listed on NASDAQ, where they trade under the symbol AGPU. Speaking on today's call will be Chris Miglino, who serves as our Chief Executive Officer; our President, Kyle Okamoto; and myself, Josh Blacher, your Chief Financial Officer. Prior to getting started, please turn your attention to Slide 2, where you'll find key legal disclosures. I'd ask that you take a moment to go through them; and with that, I'll pass things over to Chris.

Christopher Miglino executive
#2

How's it going? It's been a bit more than 30 days since I came on board as Axe's new CEO, and I'm thrilled about it. The progress made during those 30 days has been pretty extraordinary. I'll touch on that shortly. Back in September of 2025, we rolled out our strategic compute reserve, which is our GPU capacity platform. Come December, we began a strategic push into AI infrastructure and took on the new name ax compute on NASDAQ. This February, I took on the CEO position, and as of this month, effective March 26, we brought 2 outstanding industry leaders onto our board: Theodor Zoo and Thurston Dirks. Theodor Zoo holds over 85 U.S. patents and is a trailblazer in semiconductors; he started IoTelligent Technology and served before as President CEO of CELESTIAL Semiconductor. Among Europe's top technology leaders is Thurston Dirks, who brings nearly 20 years of board service along with prior chief executive positions at Telefonica Dutch land, Lufthansa, Eurowings, ePlus Groups and Dushlandglass pacer. Combined, Zu and Dirk's bring more than 40 years of joint expertise spanning semiconductor breakthroughs, car information at large-scale telecom operations, cross-border M&A and digitizing enterprises. I'd also like to announce right now that Kyle Arcimoto will come aboard as our President, and welcoming Kyle to the team fills us with real enthusiasm. We have arrived at the point where the company shifts into executing, and that will reveal how our aim of being a leader in the compute market becomes real. I'd also like to say a few words about the legacy piece of our business, Drug Discovery Services. We are looking at strategic options, and as that process advances, we'll provide further updates; let me now expand on our vision. Our mission is simple: we bring choice to the market. GPU compute has become one of the most valuable commodities worldwide, yet obtaining it has been restricted, sluggish, costly, and held by just a handful of hyperscalers, with queues running from 36 to 52 weeks. That is what we are changing. With AX compute, there are three things enterprises get that the established players simply can't deliver. The first is choice: whatever GPU you want, wherever you need it, in whatever configuration fits your particular workload — you aren't stuck with one vendor's hardware lineup. The second is speed. Depending on what's available, our development timeline can be as quick as 24 to 48 hours — days, rather than weeks or months — which lets you innovate, test, and shift direction on the fly. The third is reach: our network spans more than 200 locations worldwide, so we can serve enterprises in the places they and their customers actually are, while honoring data sovereignty rules throughout. What we're building here is a platform, and the opportunity in front of us is simply massive. I'd like to point everyone toward something important that's happened. Over roughly the past month, we've closed deals totaling about $12 million in total contract value, though that's contingent on deployment and performance conditions. As we head into Q2, once deployments go live, we anticipate monthly contract value of roughly $850,000, which works out to $7.7 million of contracted revenue for 2026 that's already been signed. On top of that, we currently serve more than 20 enterprise customers, with upward of 30 deployments actively operating. To put it plainly: our company has secured $12 million in signed contracts. Additional agreements are continuing to close, and we've made a very strong start. I also want to speak to how solid that revenue actually is, because not all income carries the same characteristics. Every agreement we sign is structured with a portion of the money paid at the start, and clients make a monthly commitment to hold capacity for themselves. That structure reduces the risk tied to receivables, and many of these arrangements even carry prepayment components ahead of completed delivery. The way these deals are designed, they produce recurring revenue from enterprise customers across every one of our native companies as well as our established enterprise verticals, and they cover a diverse set of hardware — H 100s, H200, B-20 RX590 and RX590 clusters among them. The breadth of our GPU lineup reflects the breadth of enterprise AI applications we support. As you know, we rolled out the compute reserve, the strategic compute reserve. Through that reserve, we get tokens usable for buying compute continually. What this slide shows you is what that compute reserve is worth. At the time of this presentation, the company's current market cap stands at roughly $8.6 million. The strategic compute reserve carries a value of $43 million. That figure relies on March 27, 2026 pricing, when this presentation was being prepared. The AI infrastructure token ATH is what our compute reserve consists of. Be aware that ATH token prices swing naturally, as do many cryptos. What it was worth as of March 27 captures a price at one moment, and that price is going to shift—we simply want to ensure you know substantial volatility exists there. Now, on the MNAV ratio — that's a multiple of market cap. It's currently 0.20, and if you look at it on a fully diluted basis, it comes to 0.79. What the public market is implying at these levels is a valuation below the reserve value, and that's before you assign anything at all to the operating business or to the contracted income we walked through on the prior slide. Keep in mind that NAV is a non-GAAP metric and shouldn't be viewed as a replacement for GAAP financial measures; for additional information on any discounted valuations, take a look at AX Compute's 10-K. On the compute reserve side, at present we're holding 5.9 billion tokens. That breaks down as $2.86 billion worth of unlocked liquid tokens, plus an additional 3.12 billion tokens that remain locked, and those unlock on a monthly schedule that runs all the way to December of 2028. As for what discount should apply, we're not telling investors what number to use. The figures are simply out there, and it's up to the market to decide for itself. In our view, the $8.6 million market cap doesn't reflect the operating business or the early commercial traction shown on the earlier slide, and we believe that delivering over time might shrink that gap. One thing I'd like investors to keep in mind: AX has a way to convert those unlocked tokens into cash — it can purchase compute and then sell it back. So let me walk you through how that business model actually works. The commercial traction you just sized is genuine. The slide lays out why it holds up structurally. There are 4 pieces to our model, and they operate as a flywheel together rather than as separate features. Reserve GP capacity: customers sign up for reserved clusters, priced per GPU per hour, paying monthly in cash upfront, and typically prepaying on long-term contracts. That removes pricing swings for the customer and receivable exposure for us. We collect before computers are delivered. On the disputed provider network point: there are over 200 sites globally plus a large pool of providers, which means buyers pick where they sit and still retain control over their data. Our AI infrastructure isn't constrained by the places we've already constructed — we show up wherever your operations are. Turning to OpEx arrangements: AX is able to fund and build purpose-built infrastructure, B-300 clusters among them. Through that route, clients get the capacity they need while keeping CapEx off their balance sheet. At a moment when every CFO is scrutinizing capital commitments carefully, that amounts to a genuine commercial advantage: strategic compute reserve. As we put the capital from these stakes to work acquiring additional ATH — our strategic compute reserve — we could receive a token-based incentive of up to 20%, depending on network mechanics. The model is built to grow the reserve as time goes on. The reserve adds capacity, and added capacity unlocks the next deal. So customers pay in cash. We take that cash and purchase ATH tokens, then use those to buy compute. We may get additional ATH tokens, currently aimed at as much as 20%, and subject to network mechanics, those can be used to purchase more compute later. Let me make sure everyone understands what's actually happening within our operations. The funds shown on that prior slide, held in the strategic compute reserve, are something we can turn into cash. Here's how: we sell that capacity, take the resulting tokens and purchase compute with them, then sell that compute onward to our clients — and that cycle brings cash into the business. At this point, I'll pass things along to our CFO, Josh Blacher, who will take you through the 2025 financials. Josh?

Josh Blacher executive
#3

Thanks, Chris. Ahead of walking through the financials, I want to address something significant up front. Our income statement for fiscal year 2025 contains two sizable noncash items, and both weigh heavily on the net BOSS number we reported; neither one is indicative of how the business performed at its core. Let me walk through each of them, and then I'll get into the underlying results. To start, we recorded $152.5 million of losses on digital assets at fair value. Starting January 1, 2025, we began applying ASC 350 to 60. Under this guidance, digital assets get reported at fair value on every reporting date, and the resulting mark-to-market adjustment runs through the income statement. The ATH tokens we bought, which had a cost basis of $102.7 million through December 31, 2025, saw their market value fall to $24.4 million as of December 31, 2025. That works out to an unrealized decline of $78.3 million on those holdings, and it's noncash. Those tokens are still sitting on our balance sheet. Next, we booked $52.7 million in losses on derivative instruments. That figure ties to the Crypto pipe transaction we executed on September 29, 2025, since the value exchange was denominated in ATH token instead of a fixed dollar amount, U.S. Under GAAP, we had to account for the contract as a derivative liability carried at fair value through earnings, and an assessment was carried out to determine how the instrument's fair value shifted from the September 29, 2025 execution date to the October 7, 2025 settlement date. Once the deal closed on October 7, we removed the derivative liability from the books, recognized the ACH tokens as assets, and recorded the prefunded warrants as equity. That $52.7 million figure represents the change in fair value occurring between September 29 and the October 7 settlement, and once again, it is completely noncash. Please remember that combined, these 2 items account for $205.2 million of the $233.1 million net loss we reported. Now for the full year 2025 income statement highlights. For the fiscal year that closed on December 31, 2025, continued operations brought in $125,000, up from $85,000 during fiscal year 2024. All of that revenue came from our legacy Drug Discovery Services business segment. No revenue was generated by our compute services segment in 2025, since the go-to-market pipeline was still being built out while the company went through its strategic transition. Starting in fiscal year 2026, we anticipate compute services will become a meaningful contributor. For 2025, total operating costs and expenses came in at $28.6 million, versus $10.4 million during fiscal year 2024. On a per-share basis, the net loss worked out to 13.37%, calculated using a weighted average diluted share count of roughly 17.4 million shares. That weighted average share count reflects the sizable share and prefunded warrant issuances tied to our PIPE transaction in the fourth quarter of 2025. Now for balance sheet highlights. Moving to the balance sheet: we closed fiscal year 2025 with $10.8 million in cash and cash equivalents, up from $0.6 million as of December 31, 2024, and the driver there is the proceeds from our PIPE financing transaction in October 2025. Two asset categories have shown up on our balance sheet that weren't there twelve months ago. First, digital assets came to $24.4 million, representing the $2.8 billion of ATH tokens in our possession. Second, we have a digital asset receivable of $15.5 million — that's our contractual right to receive additional HTH tokens later on, subject to vesting tied to time, which people also call locked ATH tokens. We expect those tokens to be released on a schedule we can foresee, running through December 2028. On top of that, total assets were $52.9 million as of December 31, 2025, compared with $5.0 million a year earlier. When the year closed, shareholder equity stood at $47.7 million. That compares to a deficit of $0.2 million on December 31, 2024. Returning equity to positive territory and regaining compliance with Nasdaq's stockholders' equity requirement in December 2025 were significant achievements for us. That concludes the financial portion of my remarks. Now I'd like to turn it over to Kyle Okamoto, our President, who will discuss our out strategy and the near-term pipeline. Kyle.

Kyle Okamoto executive
#4

Okay, thank you Josh. I appreciate it. I'm energized this morning. Good morning to everybody. I'm genuinely excited to be with you today to explain the way our company operates and the reasons the market conditions are so positive for what we're creating. Let's begin with the market. To put it simply, we're inside an AI supermarket — a supercycle for AI. During 2026 by itself, Gartner estimates that spending on AI will hit $2.5 trillion, along with the team. Spending on AI data centers should expand more than 31% per year through 2030. And during 2026 by itself, CapEx for global AI infrastructure will be above $1 trillion, compared with only $290 billion 2 years earlier. That's enormous expansion, plus a cumulative projected outlay of $6.7 trillion by 2030, taken from numerous credible research firms, and I believe just about anyone in this industry who spends day to day would consider those figures simply much too conservative. What counts most for aXcompute, though, is the structural change. By 2030, AI workloads rise from 25% to 70% of all activity in global data centers. That means the share of workloads inside data centers triples in only 4 years; it isn't merely more AI. It represents a basic shift and a recomposition of what compute infrastructure is genuinely meant to do. In my view, 4 factors are producing this paradigm shift. First, combined hyperscaler CapEx is running at $600 billion throughout 2026. Second, 7 gigawatts of fresh capacity will come online by 2030. Third, by 2027 inferencing is overtaking training as the main AI workload. That means distributed, latency-sensitive compute that can't simply operate in 3 U.S. regions must sit where people are, where applications or agents reside, and so on; and finally, enterprises are deliberately shifting toward bare metal performance as they become smarter and better able to use GPU-based compute, and away from hyperscaler abstraction. At AXCompute, our position sits squarely in the path of this shift. Let me explain how we truly serve customers. When you compare a compute offering directly with the public Neocloud segment, the difference is quite striking; and honestly, the investment case turns out to be very obvious. As Chris said, our distribution model reaches more than 200-plus locations across over 100 distinct configurations. That's choice. Most of our competitors run in just one country, or perhaps a handful of data centers across limited regions, typically with one configuration option. So AXCompute has centered itself on delivering choice within a highly centralized and consolidated market. We also aim for deployment time lines of this scale — 24 to 48 hours, or occasionally quicker, naturally, depending on availability; compare that with weeks and months, as Chris noted, for the remainder of the market, correct, the market as a whole. We mostly rely on an OpEx model, which permits highly predictable economic scaling, yet ultimately creates less drag. There's no need for us to spend 18 months constructing a data center; we partner with companies that are constructing data centers. For our competitors to keep pace with sales, they need to spend heavily on CapEx infrastructure. Across a very broad note, we are able to satisfy that demand right now. From day one, we were built to provide 100% bare metal performance, inside Tier 3 and Tier 4 data centers, supported by enterprise-grade SLAs that are truly meant to handle those high-performance workloads. Clearly, our rivals provide an assortment of virtualized environment offerings, but at the end of the day, this high-performance enterprise-grade workload is what we were constructed for from the ground up. Then consider valuation. As Chris and Josh both noted, our basic market cap is $8.6 million. Neocloud comparables stretch from hundreds of millions to more than $38 billion a couple of days ago. Revenue PDAs for our peers sit at 7 to 15x. On roughly 1x expected income from signed contracts, we trade — 1 versus 7% to 15% — relative to the $52.9 million audited asset base. As the slide puts it, the bottom line here is that aXcomputes and neocloud, with real income and real assets, are not yet being priced by the market; we think that shifts as we keep simply executing. So allow me to take you through the pipeline and some deal economics. How predictable are income streams is one of the questions investors ask us. To highlight the more detailed mechanics of our business, let me take you through an example of that. Take, for instance, a 10-node H-100 cluster on the 12-month commitment, which produces roughly $1.3 million in ARO. On a deal like that, gross margin comes in at around 8%, perhaps a bit higher; customers prepay 10% of the total contract value at sign in advance; that demonstrates a commitment to remain for the full 12 months and to put money on the table. The first month's rent is also paid in advance by them — meaning before we tie up compute resources, we have cash in hand. Additional ATH token incentives, currently targeted at that up to 20% level by Christmas, may then be generated by each settlement, which then compounds the strategic compute reserve, so that for the next month, each month our COGS for that compute we book becomes 20% cheaper, and over and over that compounds. Income, therefore, is priced on a reserve capacity basis at a per hour per GPU rate. Thus, with no hidden fees, clients receive a consistent bill each month. We've built acts around what is wrong with today's centralized cloud world. And you get a surprise bill at the end of the month, not knowing where your spending is, not knowing what you're paying for egress or hidden storage charges or API calls or service tickets or things like that — that is 1 of those key tenets; so no hidden fees. Typically, contracts last 6 to 12 months or longer and always was monthly advanced payment. The structure is truly designed to lower receivables exposure, but also to give customers predictability in their underlying economics and to align with their monetization model. On top of that, we provide add-on upsell offerings, such as shared storage, options like VAST or WECA, CPU nodes for when clients run bigger clusters for orchestration, or offerings such as managed Kubernetes that comes right out of the box at the software and API level. All of these can attach to a deployment, which in the end lifts margins and increases stickiness with customers. What that produces is truly a business in which agreements are built with upfront payment portions before deployment, agreements last a long time, agreements repeat, and each deal grows our P&L along with our token reserves at the same time; so that's the compounding structure. That's the flywheel Chris mentioned; it isn't linear. So now let me turn to how customers actually begin with X and how they can expand over time. Essentially, we serve clients via 2 primary routes, with a smooth and deliberate path linking them. Tier 1 is what we refer to as start fast; it's built for speed. It's built for experimentation, and in the end, it delivers enormous choice. Customers can be up and running in under a day, 24 to 48 hours typically, on any GPU type, any configuration, across all 200-plus locations, with flexible term lengths, full parametal performance from day 1; this is the lowest friction on ramp in the market; they can view all of our selections, what's available, where it is, what the network and the configs and all the small details that they need and care about most, and in most cases, they can easily try it before they buy. Tier 2 is what we refer to as X Forge, and it's meant for enterprises that want to scale very strategically using a larger cluster. And here, either our partners or AX itself can handle the CapEx build costs. So clients can remain on that OpEx model even at major scale; we're discussing thousands of high-end B-300 GPUs, for instance. We rely on NVIDIA reference architecture design; we can provide service level commitments if we deploy late, to stand behind our promises to deliver that cluster on time and at the performance levels that client expects. And we also let clients upgrade during the committed term as technology advances to the newest and best sheep use. The entire aim is to build a long-term partnership, not a vendor transaction with rigid lock-in, letting customers grow and excel and surpass over time. Clients can also shift from Tier 1 to Tier 2 as their needs grow, right, with no disruption, no penalty, no lock in. That's the entire purpose of this model. In short, we want to grow alongside our customers, not trap them. I hope that was helpful. And with that, I'll hand it back to Chris to close the sale.

Christopher Miglino executive
#5

Kyle, thanks for coming aboard — we couldn't be happier to have you with us, and we can't wait for you to put your know-how to work on the AX compute family. A few points I'd like to make before we wrap up: this market is enormous. Its expansion is quicker than anything I have witnessed in my whole life — it takes me back to the early days of the Internet, with one difference: the demand doesn't let up. And in my view, we're only at the leading edge of firms entering this arena. Plenty of organizations have already adopted AI, sure, yet I believe there remains a whole world of businesses that are only now beginning to move into it. Our position is strong when it comes to serving those folks as they come online across the globe, and doing so fast. Over the past 30 days, we have signed contracts totaling $12 million. For a 30-day stretch, that is serious money, and we are only getting underway. That works out to roughly $835,000 in monthly contract value, spread across more than 20 separate customers; and I can tell you there is far more sitting in the pipeline that we can see. So what we're witnessing makes us bullish. Here's the concept: we can keep deploying the tokens we hold, take them into the market, and convert them into compute, which in turn converts into cash — I think investors truly need to grasp this, since it gives us the means to avoid going to market to raise extra capital. Beyond the cash we already hold, we also have the strategic compute reserve, which can likewise be converted into cash by turning it into compute; so as you get to know our business better, I think you'll come to appreciate the genuine worth of that compute reserve and how, by participating in purchasing more of the ATH tokens over time, we can obtain additional tokens. So anyone who'd like to arrange a one-on-one call with us — we'd be glad to have that discussion; just reach out at ir@axcompute.com. And I want to express my gratitude to the investors who have traveled this road with us; should you have any questions, just tell us. Hopefully today's presentation offered you a bit more insight into where we stand and where we're headed. Thanks for being with us today.

Operator operator
#6

Goodbye.

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.