AI computing power remains at the intersection of policy and industrial demand.

The application version submitted to the HKEX shows that Shanghai JiLiu Technology Co., Ltd. has filed for a listing. Founded in 2023, the company specializes in AI computing power cluster products and operational services. Its core offerings include the Galaxy computing power cluster, Galaxy Service operational services, and the self-developed Mercury computing power network system and Venus computing power operating system.

JiLiu Technology's label is clear: as of 2025, it is China's largest independent provider of AI computing power clusters by revenue. By the end of 2025, the company held approximately a 10% market share in China's 10,000-card cluster service market, having served four 10,000-card-scale AI computing power clusters with an SLA-guaranteed online operational computing power scale exceeding 34,000 PFLOPS.

The State Council's "Artificial Intelligence Plus" action plan emphasizes the coordinated development of intelligent computing power, while the Ministry of Industry and Information Technology's inclusive computing power policy proposes exploring "computing power banks" and "computing power supermarkets," as well as flexible payment models based on card-time, core-time, and Token-based billing. However, while policies drive demand, they also imply coordination, sharing, cost reduction, standardization, and inclusiveness.

For computing power service providers, an expanded market space also means continuous comparisons in unit computing power pricing, scheduling efficiency, and service delivery capabilities. The ability to transition from project delivery to sustained operations represents the first critical threshold for JiLiu Technology's valuation.

Being the "largest independent" provider has value, but the qualifier cannot be omitted. According to the prospectus, JiLiu Technology ranked first among China's independent AI computing power cluster providers by 2025 revenue. However, in the broader Chinese AI computing power cluster provider market, the company held a roughly 1.1% market share, ranking ninth. This position indicates its entry into the computing power infrastructure arena but falls short of full market dominance.

AI computing power infrastructure is not a singular technological competition. Telecom operators control networks, data centers, and government-enterprise clients; cloud vendors possess platform gateways; model companies connect application scenarios; and hardware manufacturers control critical supply. JiLiu Technology's "independence" facilitates serving diverse clients but also requires competing for projects, pricing, and repeat business among industry giants.

The application version reveals that from February 20, 2023, to December 31, 2023, in 2024, and in 2025, JiLiu Technology's revenues were RMB 31.802 million, RMB 324.672 million, and RMB 520.261 million, respectively, with gross margins of 7.6%, 20.1%, and 21.8%. In 2025, the company recorded a net loss of RMB 355.555 million, primarily due to fair value changes in financial liabilities associated with preferred shares. Under non-IFRS metrics, adjusted net profit was RMB 31.123 million.

Revenue growth does not equate to a closed cash flow loop. During the same periods, the company's net operating cash flow was -RMB 9.410 million, -RMB 21.164 million, and -RMB 11.11 million, respectively. While 2025 approached positive territory, the three reporting periods still showed no positive operating cash inflow. As of the end of 2025, the company's cash and cash equivalents stood at RMB 297.946 million, declining to approximately RMB 62.19 million by February 28, 2026.

Liquidity metrics must also be viewed holistically. The company's net current liabilities expanded from RMB 2.059 million at the end of 2023 to RMB 49.276 million at the end of 2024, further rising to RMB 449.266 million by the end of 2025. This reflects both IFRS reporting factors for financial liabilities associated with preferred shares and operational factors such as increased interest-bearing bank borrowings and trade payables. If the IPO is completed and triggers relevant conversions, the capital structure will improve; however, receivables, inventory, payables, and operating cash flow must still be validated by the business itself.

Series D financing provided a buffer. According to the prospectus, the company's Series D financing totaled RMB 1.16 billion, settled on April 21, 2026, with a post-money valuation of RMB 9.16 billion. Based on this valuation, the static price-to-sales (PS) ratio for 2025 revenue is approximately 17.6x. While not the IPO issuance valuation, this multiple serves as a reference for market observers assessing pricing flexibility.

JiLiu Technology positions itself as a "full-stack AI infrastructure" provider, but its cost structure reveals a heavy hardware delivery focus at this stage. In 2025, the company's cost of sales was RMB 407.022 million, with hardware procurement costs accounting for RMB 350.628 million, or 86.1% of the total.

R&D investment presents another contrast. From 2023 to 2025, the company's R&D expenditures were RMB 2.949 million, RMB 28.360 million, and RMB 37.276 million, representing 9.3%, 8.7%, and 7.2% of revenue, respectively. This does not negate its technical accumulations in Mercury, Venus, and other systems. However, for a company emphasizing full-stack AI, 10,000-card cluster scheduling, and optoelectronic interconnectivity, the declining R&D expenditure ratio suggests that technological barriers must still be validated through product gross margins, service revenue proportions, client repeat purchases, and sustained R&D investment.

More critical is the gross margin structure. In 2025, AI computing power cluster products generated RMB 436.624 million in revenue, accounting for 83.9% of total revenue, with a gross margin of 16.8%. AI computing power cluster operational services contributed RMB 83.637 million, or 16.1% of revenue, with a gross margin of 47.7%. While the service business boasts higher margins, product delivery remains the primary revenue driver. Capital markets may assign greater valuation weight to high-margin, recurring service revenue, but for a product-delivery-dominated revenue structure, valuation benchmarks require more cautious consideration.

Client concentration is a critical metric. In 2024 and 2025, the company's top five clients accounted for 98.9% and 56.6% of total revenue, respectively, with the largest client contributing 59.0% and 16.6%. While client concentration has decreased, the top five clients still generated over half of the revenue in 2025.

Supplier concentration is also declining but remains high. From 2023 to 2025, procurement from the top five suppliers accounted for 90.7%, 75.5%, and 54.2% of total purchases, respectively. The application version also discloses that some top clients are also suppliers, and vice versa, noting that this phenomenon is common in the AI computing power cluster industry.

Industry explanations hold merit, but due diligence remains essential. Overlapping clients and suppliers do not inherently indicate issues but increase the complexity of verifying revenue quality and transaction independence. Key questions include whether sales and procurement are independently priced, whether they are mutually conditional, whether payments are genuine, and whether hardware flows, capital flows, and acceptance milestones form closed loops—all of which influence market perceptions of revenue quality.

JiLiu Technology's HKEX listing window stems from three variables: policy expansion, AI infrastructure upgrades, and capital markets' revaluation of computing power as a service. It also faces three constraints: product delivery remains the primary revenue driver, operating cash flow has yet to turn positive, and client/supplier structures still require diversification over time.

While a listing can improve the capital structure, it cannot replace validation of the business model. JiLiu Technology must now answer not whether "AI computing power is a good track (track)" but whether it can translate 10,000-card cluster capabilities into sustainable revenue, convert hardware costs into service gross margins, and transform financing valuations into operating cash flow. The data has provided its first answer; the secondary market will continue grading.

Information Sources and References

[1] Company Financials and IPO Filings

· JiLiu Technology Prospectus (Application Version): Detailed disclosures on the company's core business composition, hardware-delivery-dominated revenue structure (83.9% from AI computing power cluster products in 2025), core full-stack products (Galaxy computing power cluster, Mercury computing power network system, Venus computing power operating system), historical financial and governance matters (book losses due to fair value changes in financial liabilities associated with preferred shares), and plans for using proceeds from the HKEX listing to fund Galaxy R&D iterations, expand operational service scale, and deepen vertical integration in key industries.

· JiLiu Technology Audit Reports and Core Financial Data: Revenue trajectories from 2023 to 2025 (CAGR of 304.5%), adjusted net profit (RMB 31.123 million in 2025), and still-negative operating cash flow; gross margin breakdowns by core business (16.8% for computing power cluster products vs. 47.7% for operational services); cost of sales composition (86.1% hardware procurement costs); and declining R&D expenditure ratios (to 7.2%) for structural comparison.

[2] Industry Regulation and Listing Review Information

· Hong Kong Exchanges and Clearing Limited (HKEX) Disclosure Platform: Progress updates on JiLiu Technology's mainboard listing application and compliance disclosures regarding its request for waivers from strict adherence to Listing Rule 8.05(3) (related to shorter operating history) and Rule 8.05A (market capitalization/revenue tests).

· China Securities Regulatory Commission (CSRC) and HKEX Sponsorship and Review Rules: Regulatory guidelines for overseas securities offerings and listings by domestic enterprises (Pilot Measures for Overseas Listings), with a focus on review (review) priorities for issues that complicate revenue quality verification, such as overlapping top clients and suppliers (both exceeding 50% concentration).

[3] Industry Data and Macro Background

· AI Computing Power Infrastructure Industry Public Information: Macro policy context from the State Council's "Opinions on Deepening the Implementation of the 'Artificial Intelligence Plus' Action" and the MIIT's inclusive computing power policy, as well as technical trends driven by large model development, shifting computing power demand from single-node performance optimization (Scale-up) to cluster-scale horizontal expansion (Scale-out).

· Industry Data and Market Landscape Reports: Citing Frost & Sullivan statistics, projections for China's AI computing power cluster market size (RMB 389.1 billion by 2030); JiLiu Technology's market share in the 10,000-card cluster service market (~10%), cluster MFU improvement ratio (30%), average annualized SLA (99.97%), and its ranking as China's largest independent AI computing power cluster provider by market share.

Disclaimer

1. Regulatory and Compliance Disclaimer: All financial data breakdowns, hardware/service gross margin chain analyses, discussions on overlapping clients/suppliers, and valuation multiple (PS) calculations in this article are based on independent journalistic analysis of the company's publicly disclosed prospectus application version, audit reports, and related industry research. They aim to explore financial logic and business model evolution in capital markets. The conclusions herein do not constitute substantive judgments or guarantees by the Hong Kong Stock Exchange (HKEX), Hong Kong Securities and Futures Commission (SFC), or China Securities Regulatory Commission (CSRC) regarding the company's listing application. Whether the company meets listing conditions remains subject to official regulatory review results and the final prospectus.

2. Industrial Technology Disclaimer: This article is for informational and industrial research purposes only and does not constitute any form of technical procurement or commercial cooperation advice. Discussions on cutting-edge technologies such as large model computing architectures, GPU 10,000-card cluster scheduling, and optoelectronic interconnectivity (OCS/NPO), as well as market competition analyses, are intended to serve tech industry observation and commercial research. They should not be treated as absolute bases for enterprise intelligent computing center construction, equipment procurement, or technology selection; any commercial decisions should rely on actual technical verification, SLA agreements, and official performance test results.

3. Investment Disclaimer: This article is written based on publicly transparent data and adheres to an objective, neutral principle of "no concepts, only data." However, it does not guarantee the absolute real-time accuracy or completeness of underlying data. Quantitative features highlighted, such as "excessive hardware costs," "still-negative operating cash flow," "major client dependency and client-supplier overlap," and "valuation overhang risks," are derived from statutory disclosure documents and represent logical deductions at the time of writing. This article does not constitute buy/sell recommendations, invitations, or investment advisory opinions for JiLiu Technology or any related securities or proposed H-share offerings. Investors should fully understand the investment risks in Hong Kong's stock market and new economy enterprises and make independent, prudent investment decisions.

4. Conflict of Interest Disclaimer: Caijing Chaoxun (Finance Trends News), its authors, and the publishing platform have no conflicts of interest with the companies mentioned in this article or their commercial partners. Investors should base their decisions on independent judgment and assume corresponding risks. Investment involves risks; enter the market cautiously.

Note: This article is analyzed and integrated based on publicly available information. Please contact us promptly for corrections if errors are found.