As large language models transition into large-scale training and inference stages, the market's perception of AI infrastructure is undergoing a significant shift. Previously, capital markets tended to focus on GPUs, servers, and computing centers. However, as model parameters, training data, and inference requests continue to expand, how data is efficiently stored, managed, and delivered to GPUs with minimal latency is increasingly recognized as a critical bottleneck in AI infrastructure.

As China's largest independent distributed AI storage solutions provider, Beijing XSKY Technology Co., Ltd. ("XSKY") has officially filed its prospectus, seeking a mainboard listing through the HKEX Chapter 18C (specialist technology companies) channel. According to the prospectus, based on installation volume in 2025, the company is China's largest independent distributed AI storage solutions provider, holding a market share of approximately 10.6%. However, when the scope expands to the entire Chinese data storage market, the company's ranking by revenue falls to approximately 10th to 15th place, with a market share of only about 0.1%. In the broader distributed AI storage market—which includes both independent and non-independent vendors—its revenue share stands at approximately 2.7%.

As an independent storage vendor, XSKY faces a dual scrutiny from the capital markets. On one hand, amidst a landscape dominated by tech behemoths and closed-source ecosystems, can the survival space and premium pricing logic of independent storage sustain itself? On the other hand, the company just achieved profitability in 2025, but its operating cash flow remains persistently negative, and upstream hardware price volatility raises questions about the fundamental quality of its financial performance.

Betting on "Data Gravity"? Can Independent Storage Navigate the Grip of Giants

In the enterprise-grade IT storage market, a recognized classification line exists: "non-independent providers" offer bundled storage deeply integrated with their proprietary hardware, where software typically only works with their own servers. In contrast, "independent providers" like XSKY focus on software-defined storage (SDS), allowing their software to flexibly deploy across third-party hardware from various manufacturers and heterogeneous computing platforms.

From a monetization standpoint, XSKY's core business logic centers on delivering its proprietary distributed storage software stack, taking end-to-end responsibility for system design, hardware selection, procurement deployment, and after-sales maintenance. Its delivery models include "software-only solutions" (delivering just the software layer, with customers providing their own hardware) and "appliances" (integrating proprietary software into designated hardware as a single deliverable). Typically, customers cannot derive practical benefits from purchasing hardware or software separately—the two are deeply synergistic within the overall solution, and appliances are priced and contracted as a single deliverable. Additionally, post-deployment AI storage services—including warranty, maintenance, renewal, and capacity expansion—establish a high-retention recurring revenue stream.

With the explosion in AI model parameters, China's data volume is projected to surge from approximately 50 zettabytes (ZB) in 2025 to over 150 ZB by 2030, with unstructured data accounting for a significant portion. XSKY's distributed architecture supports linear expansion of PB-level or even EB-level storage pools, directly driving capacity "add-ons" from existing customers as data volumes swell. The prospectus reveals impressive net dollar retention (NDR) figures, rising from 46.5% in 2023 to 91.9% in 2024, reaching 132.5% in 2025, and hitting an all-time high of 160.0% in the first half of 2026.

This retention performance validates the "data gravity" business principle in enterprise storage: once core business data and AI workflows become deeply embedded in a platform, migration costs become prohibitively high, forcing customers to continually expand purchases as data capacity grows.

Regarding average transaction value, the prospectus shows that the average selling price of the company's AI data lake solutions rose significantly from RMB 192,700 in 2024 to RMB 255,300 in 2025, and further to RMB 497,600 in the first half of 2026. The average selling price of training-inference storage solutions also surged to RMB 494,900 during the first half of 2026. Behind this increase in transaction value is the acceleration of AI-related data infrastructure investments by customers, leading to greater purchase capacity per project and expanding project scale.

Meanwhile, the share of one-stop "appliance" deliveries has risen substantially. In the first half of 2026, appliances accounted for 55.3% of total revenue, up sharply from 42.7% in 2025. Since appliances include hardware value, their total contract amounts far exceed those of software-only projects, thereby lifting average transaction value at the sales level. Overall, XSKY's monetization loop represents a rolling business model that leverages "software instead of hardware, locks in existing customers through data gravity, and increases transaction value through one-stop appliances."

Financial Quality and Hidden Concerns Behind Hardcore Technology

For evaluating a high-growth specialist technology company under Chapter 18C, financial analysis typically centers on the balance between "revenue elasticity, gross margin structure, R&D efficiency" and "cash burn rate, capital structure, liquidity safety." Reviewing XSKY's prospectus reveals a financially compelling yet tense picture.

On the one hand, the company experienced a "turning point" in 2025, successfully achieving profitability with an annual profit of RMB 7.137 million, and maintaining a marginal profit of RMB 307,000 in the first half of 2026—primarily attributed to revenue surge driven by AI demand and significant operating leverage. As core underlying technologies matured, the company's R&D expense ratio dropped dramatically from 65.9% in 2023 to 28.9% in 2025, further diluting to 22.3% in the first half of 2026. This was mainly due to a unified and modularized R&D process across business lines, allowing substantial reuse of core components, greatly improving per-engineer output efficiency, and reducing the incremental costs of developing new solutions.

However, beneath this impressive turnaround lies deeper concern, including a plunge in gross margins, persistent cash outflows, and potential off-balance-sheet obligations. First, the deterioration of margin structure: during the track record period, the company's overall gross margin remained stable around 63% in 2024 and 2025, but sharply dropped to 51.8% in the first half of 2026. The root cause lies in the product mix shifting toward lower-margin appliances. Software-only solutions boast gross margins exceeding 99%, while appliances—which include significant externally purchased third-party hardware components (such as SSDs, HDDs, memory, CPUs, and network cards)—have extremely low margins, falling to just 18.9% in the first half of 2026.

Moreover, the global AI boom has spurred robust demand for high-performance storage media, causing structural supply-demand imbalance and sustained price increases. XSKY cannot fully pass on rising costs to downstream customers and has been forced to increase hardware procurement reserves to guard against supply disruption risks, squeezing its gross margin further.

Additionally, from 2023 to 2025 and the first half of 2026, the company's net cash flows used in operating activities were RMB -153 million, RMB -39.266 million, RMB -25.459 million, and RMB -93.48 million, respectively. Operating activities have never generated positive cash flow. As of June 30, 2026, the company's cash and cash equivalents had been depleted to just RMB 26.335 million, reflecting an extremely tight position. To sustain operations and inventory reserves, the company had to raise RMB 28.019 million in new bank borrowings during the first half of 2026.

In terms of competitive landscape, although XSKY holds a 10.6% market share in distributed AI storage installations as of 2025—earning titles of "China's largest independent distributed AI storage supplier" and "China's second-largest distributed AI storage supplier"—its revenue share in the overall data storage market is only 0.1%, and just 2.7% in the distributed AI storage market. Compared to non-independent giants with server manufacturing lines and extensive customer networks, the company lags in market dominance and hardware procurement bargaining power.

Additionally, the prospectus highlights high customer concentration risk: revenue from the top five customers accounted for 55.5% in the first half of 2026, with the largest customer contributing 22.1%, and most of these top five are distributors. Any fluctuation in distribution channel stability or procurement cycles of major end-projects would directly trigger significant revenue volatility.

For this Hong Kong listing, XSKY intends to use the net proceeds to strengthen R&D for next-generation AI-native storage platforms, adapt to vertical industries, expand ecosystem compatibility (through joint certification with leading GPU/CPU vendors), and build localized sales and demonstration centers in key economic hubs.

Overall, XSKY presents itself as a hardcore specialist technology company sprinting on the golden track of AI, with solid technical barriers and high customer stickiness—but simultaneously burdened by cash flow pressures, gross margin restructuring, and the grip of giants. Its post-listing performance will serve as a crucial litmus test for the real valuation anchor of China's software-defined storage in the AI era.