Home / Transcripts / Predictive Oncology Inc. (AGPU) · November 17, 2025

Predictive Oncology Inc. (AGPU) Earnings Call Transcript

November 17, 2025

NASDAQ US Health Care Health Care Equipment and Supplies earnings 34 min

Earnings Call Speaker Segments

Raymond Vennare executive
#1

Hello, everyone, and thanks for joining Predictive Oncology's presentation of Q3 financial results and a business update, which covers an overview of our newest digital asset investment — in Aethir's ATH token — meant to establish the first strategic compute reserve anywhere in the world. Through this investment, a new business line opens up for Predictive Oncology, and we think we'll be able to put ATH to work for revenue-generating purposes; on September 29th, 2025, we introduced the digital asset treasury centered on Aethir's ATH token — tokens that drive Aethir's AI infrastructure as well as our capacity to gain directly from active digital asset management of that AI infrastructure. This matters, and it's important, since we've come to learn that a major obstacle standing in the way of speeding up AI-driven medical research and discovery is no longer merely good science — it's getting computing power that's both dependable and inexpensive. An expanding limitation on innovation in drug discovery or personalized medicine isn't only the science; it's the infrastructure needed to refine computational models and build the AI methodologies that unlock that science. Acknowledging this difficulty, we think we've placed ourselves strategically alongside Aethir to meet our own future infrastructure requirements while putting money into an enabling technology capable of unlocking the coming decade of discovery and innovation, in healthcare and well beyond, across many industries. Our approach is to convert Aethir tokens into graphic processing units, or GPUs, through the Aethir partnership and network, and convert GPUs into revenue. Today's deck will demonstrate precisely how we intend to produce cash flow while backing the AI compute ecosystem. To begin, here's the agenda, so everyone understands what to expect from today's presentation. Beyond what I've already covered, I'll give you a quick company update, then a financial report from our CFO, Josh Blacher. Next we'll get into our new digital asset treasury business and the vision for using the ATH token, presented by Tom McLaughlin, our Chief Investment Officer, and then by Kyle Okamoto, who sits on Predictive Oncology's Crypto Advisory Board. Kyle will take a deep dive into how we intend to leverage our strategic compute reserve to fulfill expertise needs and market demand and to bring in revenue. As I mentioned at the outset, we were thrilled to announce this past September that we've begun a digital asset treasury strategy centered on the Aethir ecosystem and its native ATH token. Though we're definitely not the first company to launch a digital asset treasury strategy, Predictive Oncology's investment in ATH extends far past merely holding digital currencies in our treasury hoping to boost short-term returns, even though that is certainly one very significant part of the initiative. What truly sets our digital asset strategy apart is that, as a participant in the Aethir ecosystem, we're able to play a direct role in broadening access to high-performance AI infrastructure and sharply reducing costs relative to the big cloud providers. We see this as a possible fresh source of revenue, and we think it will turn into meaningful value over the long haul for both the company and those who hold our shares. When it comes to growth and demand for AI compute power, you really cannot say enough about it, and many reports lately have shown that this demand is quickly moving ahead of what infrastructure can supply; in fact, an estimate from one report by Citigroup puts worldwide spending on AI infrastructure at $2.8 trillion by 2029; we think our approach to digital assets will let us gain from that expected expansion. Since our company has depended on AI to power our drug and biomarker discovery operations, we know well the kind of bottleneck that arises when AI infrastructure availability falls behind worldwide demand, or when the expense becomes too high to bear. In our view, this holds back breakthrough innovation not only in healthcare but in every sector. With this investment in ATH, Predictive Oncology has set up the first strategic compute reserve anywhere in the world, with the goal of buying GPU compute on the Aethir network to help make AI access more democratic on a broad scale, and we believe we will meet the fast-growing need for sophisticated computing capacity. This marks a major moment in our company's history, and I feel enthusiastic about what it offers. Now I would like to hand the call to Josh Blacher, our CFO, so he can go over our financial results for the third quarter. Josh.

Josh Blacher executive
#2

Thank you, Raymond. Before we get into the financials, I want to note that our balance sheet and P&L for this quarter both reflect certain complicated and very unusual consequences of the private placement we entered into on September 29th to support our new digital asset treasury strategy; each contains temporary entries required purely as a result of when the transaction closed. As everyone is aware, during September the company completed two private financing transactions in support of its treasury strategy — one involving cash, the other involving cryptocurrency, specifically Aethir tokens, or ATH. Both took legal effect on September 29th, yet they closed on October 7th, at which point the company received $50.8 million in cash along with $292.7 million of ATH tokens, issuing shares and pre-funded warrants in exchange. The cash transaction, or cash SPA, was fairly simple. On September 29th, the company committed to issuing a set number of shares and pre-funded warrants to investors in return for a set cash amount from those investors. The sole variable that moved between signing on September 29th and the October 7th closing was the company's stock price, which permits equity classification under U.S. GAAP, with no accounting event recorded as of September 29th or September 30th. The crypto transaction, or crypto SPA, entailed a value exchange of greater complexity. The company committed to issuing a set number of pre-funded warrants in return for a set number of ATH tokens, priced off an ATH market price established on September 26th. The market value of that set number of ATH tokens moved between the September 26th pricing and the September 30th close of the third quarter, and kept moving through the October 7th closing. Consequently, the economic value to the company and to investors shifted between September 29th and the September 30th close of the third quarter. Since that economic value shifted because of a variable not tied directly to the company's own equity, U.S. GAAP mandates treating the contract as a derivative liability at fair value through earnings. The liability was booked at fair value when the contract became effective on September 29th, then remeasured on September 30th, the quarter end, and once more on October 7th at settlement to capture the fair value of the contributed ATH and the pre-funded warrants as of that date. A Monte Carlo simulation modeled expected price movements in both ATH and the company's stock price across the brief period in order to compute gains and losses. Upon the October 7 closing, the derivative liability was derecognized, and the ATH tokens were booked as assets while the pre-funded warrants were booked as equity. The final effect on the P&L will be an entry to earnings capturing the ultimate shift in the relative market value of what investors contributed — ATH tokens — against what they received — pre-funded warrants to purchase the company's common stock. Because of everything just described, the numbers for September 30, 2025 don't reflect what the company's actual cash situation or financial results really look like. Let me highlight a handful of those items here. We finished the third quarter of 2025 holding $182,000 in cash and cash equivalents, down from $612,000 on December 31th, 2024. Once more, because of a derivative liability of $74 million tied to the securities purchase contract that launched our digital asset treasury strategy, stockholders' equity came in at $77 million, versus a stockholders' deficit of $203,000 on December 31th, 2024. For the third quarter of 2025, our net loss reached $77.7 million, compared with $3.1 million for the quarter that ended September 30th, 2024. To repeat, that big loss in the third quarter of 2025 came directly from the $74.4 million loss on remeasurement of the derivative liability, and that was a non-cash charge. Revenue of $3.6 million for the third quarter that ended September 30, 2025, stayed largely flat against revenue of $3.9 million in the same period of 2024. General and administrative expenses came to $2.6 million across the three months ended September 30, 2025, versus $1.5 million in the comparable period of 2024, and they consist mostly of professional and consulting fees. Operating expenses for operations and R&D were $529,000 in the third quarter of 2025, compared with $535,000 in the matching quarter of 2024. Operating expenses for sales and marketing were $133,000 in the third quarter of 2025, compared with $73,000 in the matching quarter of 2024. Importantly, looking at our statement of cash flows, cash used for operations came to $5.9 million over the nine months ended September 30, 2025, versus $8.0 million in the same period of 2024. That $2.1 million decrease during the period, or 26%, shows just how meaningful our cost reduction initiatives have been, and how the non-cash loss on remeasurement of the derivative liability distorts the P&L. The monthly burn implied by the statement of cash flows averages out to $656,000. That wraps up the financial overview. I'd like to hand the call over now to Tom McLaughlin, our new Chief Investment Officer, to walk through our investment strategy. Tom. Revenue of $3,600 for the third quarter that ended September 30, 2025, stayed largely flat against revenue of $3,900 in the same period of 2024. Revenue of $3,600 for the third quarter that ended September 30, 2025, stayed largely flat against revenue of $3,900 in the same period of 2024.

Tom McLaughlin executive
#3

Tom McLaughlin here — Chief Investment Officer — thanks, Josh, and good morning to everybody. I'll take you through how we approach investing and the way our assets are being put to work. First, let's look at what we hold and how we intend to execute; as of November 10th, 2025, the company's position was roughly 5.7 billion ATH tokens, worth about $145.9 million in U.S. dollars. That number comes from a per-token price of $0.0268, taken from the 4:00 P.M. Eastern Standard Time quote on the Coinbase Exchange dated November 10th, 2025. Every transfer and transaction involving tokens now requires multiple signatures, and oversight along with governance responsibilities are divided between Predictive Oncology and D&A Holdings, our primary asset manager. In our view, that structure delivers both protection and clear accountability for everyone involved. When it comes to our digital asset strategy, the investment goal is simple: put the assets sitting on our balance sheet to work so we generate the strongest risk-adjusted returns possible for Predictive Oncology. We intend to reach that goal by drawing on a mix of four distinct allocation models. Those four models are: first, passive ATH staking in AI and gaming pools — this approach delivers a baseline yield with very little active management needed. Second, active staking for Aethir network cloud hosts, where we expect to share in rewards from AI bookings by taking part directly in generating network revenue. Third, GPU leasing that makes use of our ATH — here the plan is to turn tokens into direct cash flows by booking GPUs and selling compute capacity. Fourth, rotating other tokens across ecosystems. There's a chance to book GPUs with ATH and earn other DePIN tokens — that stands for decentralized physical infrastructure network tokens — and then either keep those tokens or sell them for cash, which we'd use to buy more ATH. What we really want to emphasize is this: ATH is an active asset class in our view, not something you simply hold. This slide is meant to break down the company's assets overall. As I said on the prior slide, the company holds an estimated $5.7 billion in ATH tokens on its balance sheet as of November 10th. That works out to an MNAB metric of roughly $0.76. Looking at the chart on the right, you'll see that $3.74 billion of those ATH tokens are locked, with unlocking continuing through December 2028. Once those tokens become available, we can use them to generate additional returns. Since the transaction closed, our accumulation strategy has brought in another 927.9 million tokens as of November 10th, using $25 million in cash. With such a sizable ATH position, we think Predictive Oncology is well positioned, and we plan to keep accumulating more tokens off the open market. Now, a summary of our objectives. I want to be clear about our targets and why they matter. For fiscal year 2026, our aim is a high single-digit yield on our ATH tokens. We view this as a cautious yet reachable target, thanks to the various revenue channels we are bringing online, which I mentioned on the earlier slide. At the same time, we're pursuing a range of approaches within our digital asset plan: staking, leasing, and rotating tokens in order to lower risk and broaden where our revenue comes from. Our near-term focus is to establish a steady base of cash flow that keeps our operations running; this isn't only about constructing something for the future; we are bringing in money right now. Over the longer horizon, we're striving to build a robust, durable business model that we expect the public market to value favorably relative to peers, judged on revenue as well as profitability measures. When you examine our comparisons against peers, the AI infrastructure space rewards expanding revenue and disciplined margins, and we intend to deliver on both. A quick look at Aethir. Let me now walk through why Aethir sits at the heart of our approach at AI. The Aethir network serves as the backbone of our digital asset plan — an enterprise-grade, decentralized GPU platform that handles AI, enterprise, and gaming workloads. With more than 430,000 GPUs spread across over 200 data centers worldwide and roughly $150 million in annualized revenue, Aethir stands alone as the decentralized compute platform working at genuine enterprise scale. What we believe distinguishes Aethir is its tokenized participation model. The ATH token functions as the only way to interact with the network, serving both to bring GPU supply on board and to reserve compute demand. Each unit of value moving through Aethir gets settled in ATH; amid a worldwide compute shortage, where AI demand exceeds supply by over 10x, accumulating a strategic treasury in ATH offers a rare chance to hold an asset tied directly to the infrastructure layer of the AI economy. Here's why that matters: AI infrastructure spending among the big tech giants is expected to reach $2.8 trillion by 2029. This is no niche market; it is the bedrock of the coming decade of technological growth. The network model Aethir delivers represents a fresh approach to doing business. We're eager about the possibilities this system opens up. Kyle Okamoto will now go deeper on the usefulness of the strategic compute reserve and how we intend to turn it into enterprise revenue.

Kyle Okamoto executive
#4

Thanks, Tom. Hello, everyone. I'm Kyle Okamoto, a member of Predictive Oncology's Crypto Advisory Board and the CTO and AI General Manager of Aethir. Today I'll take you through the broader AI infrastructure market landscape and, more importantly, how Predictive Oncology will use its strategic compute reserve to create real-world utility, plus the next steps toward delivering shareholder value as a result. Now that Tom has given you a quick overview of the Aethir network, let's pull back a bit and look at the global AI infrastructure market as a whole. This data truly paints a powerful picture. As Tom noted, Citigroup forecasts that AI infrastructure spending will reach $2.8 trillion by 2029. That same projection has only grown larger with every passing quarter. According to Bain & Company, worldwide power capacity demand should reach 200 gigawatts by 2023. Buildouts are taking place right now, yes, but they remain around the single-gigawatt mark, which means things are still quite early. Through 2030, the AI infrastructure market is expanding at roughly a 40% CAGR; that growth isn't linear — it's exponential acceleration. A supply and demand imbalance of more than 1,000% clearly exists. Let me say that figure again. Compute demand outstrips supply by over 10 times, correct? Jensen from NVIDIA, Elon from Tesla and X, Sam from OpenAI, and even more crucially, the enterprise clients we serve daily all confirm the shortage of affordable, suitable AI infrastructure. By 2026, the top five hyperscalers by themselves will shell out close to $0.5 trillion in AI CapEx, as reported by both Citigroup and Reuters. Herein lies the inherent issue: such a precious resource sits under the control of a handful of companies. Aethir's mission, together with Predictive Oncology, is to alter that, right? Predictive Oncology can now assist in delivering on that mission. Our view is that this isn't merely a passing blip.

Kyle Okamoto executive
#5

Analysts who know the space well, research houses, major banks, lenders, and the investor community keep revising their projections for GPU compute demand, and our view is that this pattern will persist in the coming years. Regarding what Raymond said earlier about structural limits, when enterprise GPUs carry lead times averaging 40 to 50 weeks, businesses can't afford to sit around waiting to purchase hardware on their own or waiting for legacy infrastructure vendors to scale up; they require solutions immediately. I'd like to read something Sam Altman, OpenAI's CEO, said not long ago, emphasizing that creating a strategic national reserve of computing power is highly sensible; Predictive Oncology has done precisely that: constructing a strategic compute reserve. The approach is fairly straightforward, but it packs a punch. Let's dig into the details. It genuinely begins with setting up the strategic compute reserve, which has been completed and made public, and our attention is now on what comes next. Predictive Oncology has developed the industry-standard core component needed to open up fresh methods of deploying compute, and we have an excellent team assembled that has already begun putting that reserve to use. The second step involves stimulating supply; that effort is currently underway. Predictive Oncology intends to monetize and expand its ATH holdings through staking GPU clusters on Aethir's network, extending loans to cloud hosts that supply GPUs to the network, boosting supply, and ultimately rewarding the community with superior options and a greater number of options, while further meeting that unquenchable demand and earning accretive value in return. The third step is meeting that demand, which is likewise moving forward. Predictive Oncology expects to turn ATH into fiat revenue by renting GPUs on the Aethir network and then delivering that AI infrastructure straight to enterprises. We'll dig into this particular point shortly. Finally, the fourth step involves expanding the reserve. Predictive Oncology expects to purchase additional tokens on the open market using cash proceeds and to receive from the Aethir Foundation an extra 20% of the quantity of ATH tokens that Predictive Oncology buys on the open market. Additionally, generating ATH rewards through staking and lending physically enlarges the total ATH war chest. This produces a virtuous cycle of meeting enterprise demand and growing not just the digital assets held on the balance sheet, but cash holdings too. In essence, we're turning capital into productive compute at market velocity, which we think will assist in resolving that 40 to 52 week supply chain challenge while expanding the strategic compute reserve and producing positive cash flow. Alright, no presentation would be complete without the notorious eye chart. Here's that chart offering a detailed look at the transaction workflow that generates the primary utility—not the sole utility, but the primary utility—of the native ATH token, which is meeting enterprise demand for premium AI infrastructure. Let's go through this and how it operates in practice. It's a straightforward four-step operation. The first thing Predictive Oncology does is provide ATH to cloud hosts so they can stake GPUs, after which those hosts collect extra ATH rewards through the Aethir network. In the second step, Predictive Oncology takes ATH and uses it to lease GPUs that get deployed for enterprise customers. Then comes step three, where Predictive Oncology pulls in fiat cash revenue at a positive margin, producing cash flow by supplying those GPUs to those enterprise clients. Step four has Predictive Oncology spending part of that fiat revenue buying additional ATH on the open market while also getting a 20% grant from the Aethir Foundation, with whatever cash remains going toward operations, further investments, and profits — this is what makes the cycle virtuous. ATH unlocks compute access, compute throws off cash, cash buys more ATH, and additional ATH brings more compute capacity online. This capital allocation engine was built to keep optimizing continuously through staking, rental, and sales cycles, and at the same time it ties financial returns to market expansion goals. It also opens the door for any cloud host or GPU provider to plug into the Aethir network, since Predictive Oncology can assist them through staking, lending, and currency appropriation, which further boosts the supply side of the equation, giving those enterprise clients more choices, faster solutions, and better prices — which leads us to the near-term pipeline and how we're turning this demand into money right now. Through its partnership with Aethir, Predictive Oncology can tap a near-term pipeline carrying enterprise-scale revenue potential. Let me take you through several small samples of the major opportunities we're working on at present. These six active prospects cover GenAI platforms, inferencing leaders, multimodal AI companies, and so forth. None of this is speculative. These are well-funded, high-growth companies, mostly located in the U.S., with backing from top-tier VCs and strategic investors such as NVIDIA, AMD, Meta, Sequoia, and so on. You'll see contract sizes starting around $3 million for a short-term training run for a real-time generative video firm on NVIDIA H200 GPUs; that company has already used Aethir successfully for training runs like this before. The range extends to a $120 million long-term engagement on the newest, latest, and greatest NVIDIA GPUs, the Blackwell B300 series, which is moving into production globally this quarter. Past these six examples, there are 100-plus additional prospects in the pipeline, from startups looking to experiment with GPUs for the very first time as they stand on the brink of their next groundbreaking innovation, to universities or research firms that badly need GPUs at the right price to finish their studies and produce that next big idea. Although the newest, you know, AI unicorns in the enterprise space might grab the spotlight on this slide and this earnings call, and the latest and greatest Hopper and Blackwell GPUs are clearly what's highlighted here, the go-to-market teams are serving a broad customer base across many GPU configurations. Because AI is very much horizontal, it has touched or is touching every industry, every market, every geography, every vertical. So, the goal for the team is a client base that's diversified, which aligns strongly with the vision and mission of democratizing access to such a valuable resource. The main points, I'd say, are that top-tier, industry-leading AI infrastructure is highly sought after when priced correctly. A number of significant enterprise opportunities are being worked on alongside Aethir, and there are many smaller deals making progress too. Typically, contract terms are long-term, correct? This is dedicated, reserved bare-metal infrastructure. We aim for 6 months or more, and payment is always upfront, which reduces risk. The market we're now fully engaged in holds billions in unmet demand, as Raymond mentioned. We're truly set up to grab a meaningful portion. Let's consider what that could mean from an investor's viewpoint. AI infrastructure peers are trading at 19 to 67x, you know, P to S ratios even with losses. Predictive Oncology intends to generate profitable revenue with zero infrastructure, no hardware infrastructure needs, which creates considerable scarcity value. To give this market context, let's examine the neocloud public company peer group and point out the main differences, since what we're doing is uncommon and distinctive here. If you look at the AI infrastructure sector, investors heavily prioritize revenues and growth, mostly overlooking big losses. Firms such as CoreWeave, White Fiber, Applied Digital, and Nebius are trading at PS multiples from 18, 19 to 67x, even with enormous net losses of hundreds of millions of dollars each quarter.

Kyle Okamoto executive
#6

It's obvious their share values have gone through the roof, since people putting money in understand the AI story isn't going anywhere and that we're still at the beginning. Now, here's the part that sets things apart. With every AI infrastructure client contract it might carry out on the Aethir network, Predictive Oncology intends to be profitable. Whereas rival neoclouds are bleeding money on building out infrastructure, Predictive Oncology is able to supply big GPU clusters without any need to spend on physical infrastructure. Such is the strength of that strategic compute reserve. You get immediate AI infrastructure access on an OpEx model, drawing on the native ATH token to open up that capacity. In the same way, this group of peers carries hefty operating expenses and overhead, yet Predictive Oncology receives a ready-made, turnkey GPU-as-a-service offering, with no requirement for enormous software development budgets; because of this, Predictive Oncology stays extremely competitive on price within the sector and in the end makes GPUs within reach for developers pursuing that next world-changing breakthrough, and for researchers seeking to solve the toughest problems life presents. I'm displaying the figures from a few weeks back, October 15, for the neocloud peer group. The truly distinctive point here is that there is no public neocloud out there producing positive net income and free cash flow. That places Predictive Oncology above everyone else. Even though the market pays for growth and revenue rather than profitability, Predictive Oncology's ultimate goal is to hand over both rising revenue and profitability; we're pretty thrilled about what will arrive shortly. Predictive Oncology is broadening past medicine to turn into a major participant in AI infrastructure. As we anticipate what comes next in the near term, past wrapping up deals in the pipeline's later stages, which I covered a moment ago, Predictive Oncology is genuinely centering its mission on compute infrastructure that fuels the new AI economy; our belief is that the Predictive Oncology strategic compute reserve, constructed on Aethir's ATH token, will let Predictive Oncology employ ATH to deliver the newest and finest AI infrastructure to enterprises across the world. That generates genuine utility for the digital assets held in that reserve, producing revenue and profit for the company so it can expand the reserve further and, in the end, make clients delighted. This enormous war chest of compute tokens should, in expectation, let Predictive Oncology bring on board NVIDIA's newest GPUs, the Blackwell B300s, to win over marquee enterprise logos hungry for that cutting-edge technology to drive innovation ahead. Finally, the team expects a near-term brand and stock ticker refresh to mirror the strategic focus, sustaining Predictive Oncology's leadership in AI and machine learning-based drug discovery and therapy techniques, while adding what we hope becomes a high-growth business spanning AI infrastructure and digital asset management, with further exciting digital asset utility use cases still to come. This covers this section of the prezo, and we'll return to Raymond for closing remarks. Thank you.

Raymond Vennare executive
#7

I appreciate everyone for listening in, and that brings today's presentation to a close. Our outlook on what lies ahead stays highly positive, and we're incredibly excited about the collaboration we have with Aethir and D&A.

For developers and AI pipelines

Programmatic access to Predictive Oncology 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.