First Research Report from Yili Hua's New AI Fund: The Wave of AI Computing Assetization, Axe Compute May Become the Most Undervalued GPU Computing Entry in the US Stock Market

Author: Axiom Investment, led by Yili Hua, founder of Liquid Capital (formerly LD Capital), and his team, who established or are leading the AI investment fund.

By the end of 2025 and into the first half of 2026, as other sectors (some overvalued growth, cyclical, purely narrative themes) fade, while AI Capex continues to be revised upwards, and semiconductor and data center-related stocks significantly outperform, a true market consensus will form --- "AI is no longer just one of the themes, but has become the absolute main line of the global capital market." With research into undervalued AI companies in the US stock market, Axe Compute has become a key focus for us this year. Recently, on July 22, the announcement of an additional $1.3 billion AI computing contract greatly strengthened our confidence in investing in Axe. If the contracts can be executed smoothly and the data can truly reflect in the financial reports, we believe "Axe Compute, a company currently valued at less than $100 million, may become the most undervalued GPU computing entry in the US stock market."

I. A Glamorous Transformation from Pharmaceuticals to AI Computing

Before its name change in December 2025, Axe Compute was formerly Predictive Oncology Inc. (NASDAQ: POAI), a typical micro-cap biopharmaceutical company in the US stock market. As a typical "small-cap biotech stock," POAI's performance in the pharmaceutical stage was mediocre: revenue remained at the level of hundreds of thousands of dollars for a long time, with continuous losses, and its market capitalization hovered in the tens of millions of dollars range, attracting very little attention from the capital market.

In September 2025, the company suddenly launched the Strategic Compute Reserve, clearly indicating that it would focus on the Aethir native utility token (ATH) as the core, continuing the Crypto Treasury strategy narrative, suggesting that the company would shift towards AI narratives and computing business.

In October 2025, the company completed two simultaneous PIPE financings, totaling $343.5 million in cash, financed through a mix of $50.8 million in cash and $292.7 million in nominal value of ATH. Through this financing, the company's balance sheet turned from negative equity to a positive $47.7 million profit, acquiring 6.348 billion ATH, formally binding the company with the Aethir network, allowing for a capital operation model that combines AI computing narratives and treasury company overlays, thus entering the observation horizon.

On December 11-12, 2025, the company underwent a brand overhaul, changing its name from Predictive Oncology Inc. to Axe Compute Inc., and its ticker from POAI to AGPU, continuing to trade on Nasdaq.

By the end of the first quarter of 2026, Axe Compute officially began operating as a new cloud service provider, with the associated party of the crypto project Aethir potentially becoming the largest shareholder, conveying a comprehensive transformation signal to the financial market:

On February 9, Charles L. Nuzum was appointed chairman, Christopher Miglino (who previously participated in the ATH transaction structure design) officially became CEO, and in March, the board was restructured with Kyle Okamoto (former Aethir CTO/GM) appointed as President.

On April 1, the company completed enterprise-level commercial access to the Aethir distributed GPU network (over 400,000 GPU containers, 200+ locations, 93 countries), signing the first batch of contracts worth approximately $12 million. The contracts were mainly for the Immediate Access Program, contributing an expected revenue of about $835,000/month, with payment methods being prepayment + monthly prepayment, which has begun to contribute a small amount of compute revenue (approximately $7,000 confirmed in Q1).

On April 22, 2026, the company disclosed a $260 million exclusive B300 cluster contract (the first order of the Build Program), with core terms including a 36-month take-or-pay agreement, delivering 2,304 NVIDIA B300 GPUs + AI high-speed storage (US Tier-3 data center, 4.8 MW dedicated power). The structured deposit + prepayment + monthly prepayment will result in approximately $21 million in quarterly revenue after going live in Q3 2026.

On May 27, 2026, it was confirmed that the company received a $43 million first payment for the B300 contract, marking the first true contract cash milestone and confirming that the Build model has been initiated as planned, with hardware procurement and deployment underway.

On June 16, 2026, the company finalized a $25.9 million long-term deployment contract for Blackwell / Grace Blackwell (12 months + 24 months, renewable), of which $12.9 million has been prepaid.

On July 22, 2026, the company announced an additional $1.3 billion AI infrastructure customer contract, based on five-year agreements with options for renewal, requiring a large upfront payment, and including terms for continuous GPU upgrades as new generations of GPUs are released. Expected revenue will begin to materialize by the end of Q4 2026, with the upfront payment to be made in Q3 2026, at which point the annual recurring revenue (ARR) will exceed $384 million. This $1.3 billion large order should mark the true starting point for the market to re-evaluate Axe.

II. Multiple AI Computing Solutions, Highly Flexible "Coreweave": Dissecting Axe's Business Model

Axe Compute Inc. is a technology company focused on providing high-performance computing infrastructure for artificial intelligence (AI) workloads. By acquiring large-scale GPU capacity from hardware manufacturers and infrastructure providers, it deploys this capacity to enterprise customers through long-term service agreements; the service scope includes hardware procurement, data center hosting, networking, storage, and financing. Axe also retains its oncology drug development solutions business, but this is currently not the company's main business.

1. Axe's business is divided into two product lines:

(1) Immediate Access Program

Targeted at customers who need rapid deployment and flexible scaling. Relying on the existing GPU inventory of the Aethir distributed network, deployment can be completed in as fast as 48 hours, covering over 200 global nodes. Suitable for inference, fine-tuning, and small to medium-scale training scenarios, with monthly payments based on reserved capacity.

(2) Build Program / AI Factory

Targeted at ultra-large scale, long-cycle dedicated computing needs. Axe is responsible for overall architecture design, data center site selection and power negotiations, hardware financing arrangements, and ultimately enterprise-level SLA (Service Level Agreement) operations, "design-deploy-own-operate."

The $260 million three-year large order landed in April 2026 is a landmark case for this model. The company plans to procure a dedicated cluster composed of 2,304 NVIDIA B300 graphics cards and AI-specific high-speed storage infrastructure from a Tier-3 data center facility in the US, with 4.8 megawatts of dedicated redundant power, specified deployment locations, and service standards, with deployment work scheduled to be completed in Q3 2026. A structured payment arrangement is in place, with the first $43 million already received. During the 36-month service period, the company will confirm approximately $21 million in revenue each quarter.

In June 2026, the company signed a $25.9 million long-term deployment contract for Blackwell and Grace Blackwell, covering inference infrastructure and simulation platform scenarios. $12.9 million has already been received as a prepayment.

In July 2026, the Build business line secured over $1.3 billion in long-term AI infrastructure contracts over five years across the US and Europe, significantly exceeding the annual signing target of $1 billion. Project upfront payments will be received in Q3 2026, with continuous revenue officially confirmed starting from the end of Q4 2026. Once all clusters are operational, the corresponding annual recurring revenue will exceed $384 million. The company's management stated that current market demand is strong, and related revenue will be included in the 2027 annual recurring revenue, continuously opening up medium to long-term growth space.

2. Reinterpreting Axe's Build computing business, the best comparison is Coreweave, one being a leader in centralized training, the other a new force in global mixed computing:

CoreWeave follows a heavy asset, centralized, and deep cultivation training scenario route, operating 49 large AI data centers across North America and Europe, with approximately 250,000 high-end GPUs. It builds single-room training clusters at tens of thousands of cards through InfiniBand high-speed interconnect networks and Kubernetes-native orchestration, achieving extreme performance for ultra-large-scale distributed training suitable for trillion-parameter training in leading AI labs like OpenAI, Meta, and Microsoft. It went public on Nasdaq in March 2025 and received an additional $2 billion strategic investment from NVIDIA in January 2026, becoming a benchmark enterprise for dedicated AI computing cloud (Neo-Cloud). However, due to all data centers being concentrated in North America and Europe, the 80-150 milliseconds of network latency caused by intercontinental transmission, combined with compliance requirements for data residency in various countries, has kept CoreWeave out of many regional markets in Asia-Pacific, the Middle East, and Latin America.

Axe Compute, on the other hand, follows a mixed model, distributed, and globally covered route. On one hand, it integrates global third-party data center resources through the Aethir distributed computing network, deploying over 200 computing nodes in 93 countries, providing access to a total of over 435,000 GPUs. On the other hand, it is vigorously expanding a new cloud business centered on assets worth over $1 billion. This enables it to penetrate the large-scale customized computing market, serving all types of GPU buyers and AI companies.

3. Financial Analysis:

Axe Compute's financial performance for the first quarter ending March 31, 2026

Press enter or click to view image in full size

As of March 31, 2026, the company holds $6.9 million in cash and cash equivalents, a $20.2 million ATH digital asset position (approximately 2.83 billion tokens), and $9.4 million in current digital asset receivables, totaling approximately $36.5 million in liquidity. Management believes this is sufficient to support the company's operations through fiscal year 2026 and beyond.

In the first quarter of 2026, revenue was $35,000, compared to $110,000 in the first quarter of 2025. The sales in the first quarter of 2026 primarily came from the traditional drug discovery services sector, with the computing services sector contributing only $7,000. According to the company, a $43 million initial payment for the B300 large order was received in May, and an additional $25.9 million long-term contract with Blackwell was added in June, both of which have yet to be reflected in the income statement.

Once the $260 million exclusive cluster officially launches in the third quarter, it is expected to generate approximately $21 million in computing revenue in a single quarter, equivalent to 600 times the total revenue of the first quarter. Assuming a $1.3 billion order launches in the fourth quarter, quarterly revenue could increase by an additional $65 million to $86 million, representing over 400% quarter-on-quarter growth. The company is on the brink of a transition from quarterly revenues in the hundreds of thousands to millions, yet the current market pricing does not fully reflect the certainty of this step-up in revenue.

In the first quarter of 2026, the net loss was $7.7 million. This net loss includes a non-cash mark-to-market loss of $4.3 million from the company's ATH digital asset holdings. As of March 31, 2026, accounts receivable stood at $659,000, up from $32,000 on December 31, 2025. This quarter saw significant increases in both accounts receivable and contract liabilities, reflecting the monthly prepayments required from Compute Services customers following project launches at the end of the first quarter.

Axe Compute CEO Christopher Miglino stated, "Our goal this year is to sign contracts worth $1 billion, and the contracts finalized in July have far exceeded that target... We believe that signing contracts worth $2 billion this year is not out of reach, which will help boost next year's annual recurring revenue (ARR)." Combining his public statements from the first half of the year, Axe Compute currently has potential business orders exceeding $4 billion and has signed contracts worth over $1 billion, aiming to sign a total of $3 billion in contracts this year.

4. Valuation Analysis:

Model 1: FY2026E Forward Price-to-Sales Ratio (Forward P/S)

Estimated annual revenue

The following are confirmed orders that have been officially released, from which FY2026 revenue can be estimated at approximately $125 million.

3 Wall Street analysts predict that AGPU's revenue for 2026 will average $163,935,524, with a minimum revenue forecast of $157,505,455 and a maximum of $168,752,872. By 2027, this figure is expected to reach $254,372,663, with a minimum revenue forecast of $244,405,017 and a maximum of $261,853,600. Currently, we conservatively estimate confirmed revenue at approximately $125 million.

CoreWeave's Forward P/S is approximately 3.88 times, while Axe Compute's actual confirmed revenue for 2026 is about $125 million, with a total share capital of 11.385 million shares, currently priced at $6.85.

Axe's market value = $125 million × 3.88 = $485 million

Corresponding stock price: $485 million ÷ 11.385 million shares ≈ $42.60/share

Relative increase from current price: 42.60 ÷ 6.85 ≈ 6.21 times

Model 2: P/ARR (Forward Scenario Calculation)

The P/ARR (Annual Recurring Revenue) ratio is a commonly used steady-state valuation metric in the computing infrastructure industry, suitable for business models centered around multi-year locked computing contracts, better reflecting the intrinsic value of a company's long-term stable cash flow. This time, we reference the industry leader CoreWeave's P/ARR valuation center of approximately 2.4x for July 2026, as a fair pricing benchmark for mature computing service providers.

As of now, the company's total Build business long-term order forward steady-state annual recurring revenue (ARR) can reach $384 million.

Axe's reasonable total market value = $384 million × 2.4 = $921.6 million

Corresponding target stock price = $921.6 million ÷ 11.385 million shares ≈ $80.94/share

Relative increase potential from current price: 80.94 ÷ 6.85 ≈ 11.8 times

Comprehensive calculations indicate that Axe's stock price has an upside potential of 6 to 11 times, and the current market value is severely undervalued. The above calculations do not account for valuation discounts based on the differences in business scale and maturity between the two companies, and the actual reasonable valuation center may have downward adjustment space.

From a horizontal comparison with peers, AGPU's current market pricing shows a significant mismatch with its business scale and growth potential. As of now, the company's market value is only about $8 million, while based on the long-term contracts that have been established, its guided ARR has reached $384 million, corresponding to a P/ARR of only 0.2 times. In contrast, peers such as Nebius, CoreWeave, IREN, and WhiteFiber have P/S on ARR of 6.9 times, 2.4 times, 4.0 times, and 10.4 times, respectively. Even considering that AGPU is still in the early stages of commercialization and the revenue recognition rhythm has not been fully released, its valuation level is still far below the industry average. With the B300 exclusive cluster and subsequent contracts exceeding $1 billion contributing revenue in the second half of 2026, the company's ARR is expected to be rapidly realized next year, and the current extremely low valuation multiples provide significant safety margins and flexibility for investors.

AI x Crypto Capital Model Design: "Compute + Treasury" Dual-Drive Model

In addition to the expected impressive Compute business data, AGPU also has a highly imaginative flywheel model, namely the ATH Treasury strategy. Unlike companies that simply hoard BTC and ETH, ATH is embedded within a company that generates positive cash flow from related businesses. Compute orders directly drive ATH demand and settlement, while Treasury appreciation feeds back into Compute expansion, creating a self-reinforcing positive flywheel.

1. What are ATH and Aethir?

The Aethir network is a decentralized physical infrastructure network developed by the Panama Foundation DCI Foundation (referred to as "DCI"). The Aethir network aggregates enterprise-level GPUs contributed by independent data centers, enterprises, and other hardware owners into a globally distributed network. This network aims to provide instant GPU computing resources for AI training and inference, cloud gaming, and other virtualized computing workloads, often at lower prices than centralized cloud providers. Within the network, three roles work together to ensure the availability, applicability, and quality of computing resources: "containers," which are the actual containers executing computations; "verifiers," which test and monitor containers to validate their integrity and performance; and "indexers," which match computing resource users with suitable containers. Buyers of computing resources utilize the network's computing resources.

ATH, as an agent unit of GPU computing power, serves as the transaction medium and incentive unit for participants in the Aethir network. To become a provider of computing resources, network participants need to acquire ATH and stake it as collateral to contribute GPU resources and qualify for processing computation requests. Once computing resources are delivered and verified, ATH flows from the users of computing resources to the providers as payment and rewards. Resource providers earn ATH through "capacity proof" rewards (to maintain availability and readiness) and "delivery proof" rewards (for completing workloads), as well as service fees paid by computing resource users. Service providers can re-stake, hold, lend, or sell the ATH they receive. The treasury of Aethir is responsible for managing protocol fees and allocating ATH for protocol development, while the blockchain settlement layer records transactions and facilitates the transfer of ATH.

2. Capital Design of Axe and ATH

The capital structure of ATH Treasury is not a simple "buy and hold" strategy but is designed through two layers, deeply binding Axe Compute's business entity with the Aethir (ATH) ecosystem, forming a closed loop of business-capital-token. The core advantage of this design is that each Compute order landing will translate into incremental demand and value capture for ATH in the future, rather than relying solely on external market liquidity and sentiment like traditional treasury companies.

(1) Axe's Access business is embedded in the Aethir network

The Access model (instant access) is the core of AGPU's light asset expansion, fully relying on the Aethir distributed GPU network (400,000+ GPU containers, 200+ locations, covering 93 countries). After clients place orders through AGPU's Access platform, the computing tasks are executed directly in the Aethir network. Each invocation of an order consumes or stakes ATH, creating real demand.

Each Access order = directly drives ATH demand + generates positive cash flow (prepaid income). This design makes AGPU's Access business a "natural demand engine for ATH." Linking business growth to ATH prices --- the more orders there are, the more ATH is consumed/staked, and the stronger the price support.

(2) Axe's treasury strategy: holding ATH and forming strategic reserves

AGPU's Treasury strategy is an upgraded version of BTC/ETH treasury. Companies like MicroStrategy passively hold BTC as "digital gold," relying on external Bitcoin halving and market cycles, lacking intrinsic cash flow support. AGPU's ATH is "embedded" within a Compute business that generates positive cash flow. ATH is not only a reserve; the Aethir network uses ATH for staking and settlement, and Axe's Access business operates on it, naturally forming a closed loop.

3. How AGPU and ATH's positive flywheel operates

(1) AGPU and ATH's Business Flywheel: "Order-Demand-Value Addition" Cycle Driven by Access Model

Order fulfillment, increase in access orders --- Increased demand for ATH --- Endogenous value addition of ATH --- Expansion of Axe's balance sheet --- Value addition of AGPU --- Acquisition of more AI computing power orders

(2) AGPU and ATH's Capital Flywheel: "Performance-Funding-Stock Increase" Cycle Driven by Treasury Value Addition

Order fulfillment --- Improvement in company performance, increase in available funds --- Purchase of ATH, increase in ATH holdings --- External value addition of ATH --- Expansion of Axe's balance sheet --- Value addition of AGPU --- Acquisition of more AI computing power orders

The business flywheel provides endogenous demand and cash flow (Access orders directly feed ATH demand). The capital flywheel provides leveraged value addition and asset expansion (Treasury stock increases amplify ATH price effects), forming two deeply nested driving models --- performance and ATH prices jointly drive AGPU price increases and performance expansion. This model may become a new paradigm for the "AI x Crypto" capital model.

IV. Potential Risk Variables for Axe

The story of Axe Compute (AGPU) is highly imaginative, and the current market pricing does not reflect the optimistic expectations for future contract fulfillment and ATH reserve value. However, like any highly elastic growth asset, the story leads financial realization, and valuation is more based on future GPU contract delivery and ATH price performance rather than confirmed revenue and profit. Historical revenue remains at an extremely low base, and the real conversion of large orders and verification through financial reports will take time. The following are the main risk variables that investors need to assess carefully.

1. Contract Execution and Delivery Risks

The Build Program is key for AGPU's transition from a light asset Access model to a semi-heavy asset customized cluster. The $260 million B300 dedicated cluster (expected to generate about $21 million in quarterly revenue after Q3 2026) and subsequent $1.3 billion global customer contracts have been signed. However, there are execution risks related to hardware procurement, data center coordination, power deployment, and enterprise-level SLA implementation. If the launch is delayed or customer acceptance is postponed, revenue recognition will be delayed, potentially affecting cash flow and market confidence.

2. Revenue Conversion and Financial Report Verification Risks

In Q1 2026, revenue was only $35,000 (with minimal contribution from Compute services), while nearly $1.6 billion in signed orders have not yet been significantly converted into revenue. Wall Street analysts have an average revenue expectation of about $164 million for 2026, but these forecasts include assumptions for conversion in the second half of the year. If order fulfillment is slower than expected, actual revenue may be significantly below consensus. Non-cash ATH valued at market value will continue to fluctuate, and increases in accounts receivable and contract liabilities also reflect potential bad debt risks under the prepayment model.

3. Macro and Market Valuation Risks

If AI Capex is adjusted downward due to economic slowdown or technological iteration, order demand may be affected; tightening GPU supply, energy costs, and data center compliance requirements may all increase execution costs. Forward P/S and P/ARR calculations are based on assumptions that financial data can be realized, and the actual reasonable valuation center may be discounted due to differences in scale and maturity, with current high elasticity also meaning amplified volatility.

Overall, the story of Axe Compute is ahead of financial realization, and the rhythm of revenue recognition and the next quarterly or semi-annual financial report will be key verification windows. The above risk variables are not exhaustive, and investors should conduct their own due diligence, fully understand the associated risks, and make independent decisions based on their risk tolerance.

In summary: AGPU has completed a remarkable transformation from traditional biotechnology to AI GPU Compute entry in less than a year, with a business model that encompasses a hybrid AI computing power solution of the light asset "Access model" + large-scale cluster construction and leasing "Build model," achieving an astonishing $1.6 billion in order contracts. Coupled with the "Compute + Treasury" dual-driven model, combined with business model, asset reserves, and valuation not priced comprehensively, AGPU has a growth potential of 6-11 times relative to the current stock price, making it a highly elastic asset worth paying close attention to in the wave of AI computing power assetization.

Disclaimer: This content is provided for general branding and informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online events, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets or to use any services. Crypto assets are highly volatile and may result in loss. WEEX services and online events may not be available in all regions and are subject to applicable laws, regulations, and eligibility requirements. You are responsible for ensuring that your use of WEEX services complies with local laws and for carefully assessing the risks before participating in any crypto-related activities.