Sharetronic Data Technology: From Foxconn Hardware OEM to China's AI Compute Rocket Ship

I. Introduction & Episode Roadmap [00:00 - 07:00]

On July 13, 2026, in a Shenzhen conference room, a shareholder asked the chairman of Sharetronic Data Technology a straightforward question. Until roughly eighteen months earlier, the company was known mostly for manufacturing hard-drive enclosures and doorbell cameras. Could the chairman provide concrete figures on the contracts underpinning the tens of billions of yuan in servers Sharetronic had been purchasing?

耿康铭 Geng Kangming declined. The company, he explained, had applied to the exchange for an exemption from disclosing detailed contract terms, precisely because it sought less public attention, not more. What mattered, he said, was delivery and supply reliability—not announcement volume.¹

That exchange captures the central tension surrounding the company’s recent market transformation. 协创数据 Sharetronic Data Technology Co., Ltd. (300857.SZ), listed on the 创业板 ChiNext board of the 深圳证券交易所 Shenzhen Stock Exchange, spent fifteen years as a low-margin contract manufacturer. In 2025, it reported revenue of ¥12.24 billion, up 65.13% year over year, and net profit of ¥1.164 billion, up 68.32%.² For the first half of 2026, it guided for net profit between ¥1.5 billion and ¥1.9 billion—representing year-over-year growth of 247% to 340%.³ In effect, six months of profit was set to surpass its earnings over the previous three years combined.

The primary engine of this growth was a business line that barely existed two years earlier. Revenue from "intelligent computing products and services"—GPU server clusters built, owned, and leased to clients training and running AI models—reached ¥2.761 billion in 2025. That marked a 1,727% increase over 2024 and accounted for 22.57% of group revenue.² Combined with its adjacent server-and-peripheral remanufacturing line, which grew 169% to ¥2.575 billion, nearly 40% of group revenue came from operations added after the launch of ChatGPT.

A note on identity, given the crowded Chinese small-cap sector: this is 协创数据, ticker 300857.SZ, not 深圳市智动力精密技术 Zhidongli Precision (300686.SZ), a separate Shenzhen precision-components manufacturer. Sharetronic's background spans storage devices, IoT cameras, and now high-performance compute.

Management's thesis is straightforward. Geng contends that Sharetronic applied the operational discipline of Taiwanese contract manufacturing—reflecting his six years at 富士康 Foxconn—to an AI hardware sector constrained by tight supply, high deployment costs, and strict time-to-market demands. The company describes this architecture as an integrated "compute-connect-store" stack: GPU servers and rental capacity, optical interconnect modules, and enterprise storage under one roof.

This operational shift coincided with a dramatic repricing in the equity market. At the start of 2023, Sharetronic was a micro-cap with a valuation of roughly ¥4 billion. By July 2026, it had joined the ranks of ¥100 billion-plus companies, closing at ¥220 per share on July 20 for a total market capitalization of ¥107.66 billion.¹⁸ A nearly 27-fold valuation increase in three and a half years reflects a fundamental shift in how the market evaluates the business model. That elevated market standing—and the credit capacity accompanying it—became an active input to its strategy rather than merely a passive outcome.

Skeptics focus not on end-user demand, but on balance-sheet leverage. Sharetronic's asset-liability ratio climbed from 52.39% at the end of 2023 to 85.89% by the end of March 2026.⁴ Interest-bearing debt reached approximately ¥25.9 billion by that date.¹ Furthermore, since the start of 2025, the company has announced cumulative server procurement plans exceeding ¥32 billion—compared to its 2024 group revenue of about ¥7.4 billion.⁵ This represents one of the most leveraged balance-sheet expansions among listed Chinese hardware firms in the current cycle.

Here is how this story unfolds: First, the Foxconn origin and the OEM decade that taught Geng cost discipline while confining the business to single-digit margins. Second, the 2020 ChiNext listing and subsequent efforts to move beyond hardware commoditization through storage and software. Third, the 2023–2025 pivot into AI compute alongside the remanufacturing business that remains widely misunderstood. Fourth, an examination of revenue sources and segment economics. Fifth, strategic acquisitions and the proposed ¥8 billion equity raise. Sixth, an evaluation of whether these developments establish a durable competitive advantage. Seventh, an assessment of a management team that has executed aggressively while taking on significant leverage. And finally, the key metrics and catalysts for investors to monitor.

---

↑ Back to Top

II. The Founder's DNA: Foxconn School of Hardware & The OEM Era (2005–2015) [07:00 - 17:00]

A section chief goes south

Born 耿思华 Geng Sihua in 1974 in Bozhou—an agricultural prefecture in northern Anhui better known for traditional medicine markets and 古井贡酒 Gujing Gongjiu baijiu than for semiconductors—Geng graduated from university in 1995 and joined the migration of ambitious workers to southern China.⁶

In 1990s Shenzhen, 富士康精密组件（深圳）有限公司 Foxconn Precision Components (Shenzhen) Co., Ltd. offered one of the world's most rigorous operational environments for hardware manufacturing. Geng joined Foxconn in March 1995 and remained until January 2001, rotating across engineering, quality assurance, procurement, sales, and manufacturing before advancing to 课长, or section chief.⁷

That rotation established the operational foundation for Sharetronic’s eventual business model. Foxconn's success stemmed not from primary invention, but from industrializing external technology at scale and cost levels that made brand owners dependent on its infrastructure. In this environment, procurement experience underscored how fractions of a cent saved on a component compound across millions of units; sales and quality management demonstrated how minor field-failure rates destroy high-value client contracts; and factory floor leadership framed manufacturing yield as compounding cash flow. Geng operated under a framework where product design was secondary to process efficiency.

The Foxconn model also incorporated a major financing component. Beyond assembly, contract manufacturers frequently fund working capital by purchasing components, holding inventory, building to client forecasts, and accepting delayed payment terms. The contract manufacturer's balance sheet functions, in practice, as a credit facility for the brand owner. Six years in this environment developed a familiarity with holding inventory and managing client operational risk through leverage. Two decades later, Geng applied that same balance-sheet strategy to GPU server clusters.

After leaving Foxconn in 2001, Geng operated a workshop producing computer chassis and cooling components.⁶ In November 2005, he formally established Sharetronic. The company was structured as a foreign-invested enterprise with registered capital of HK$50 million, wholly owned by a Hong Kong holding entity that was 61% controlled by Geng and 39% owned by an independent third party, POWER CHANNEL LIMITED.⁷

Building the white-label engine

In 2005, Shenzhen's electronics manufacturing ecosystem was expanding rapidly following China's accession to the World Trade Organization. Component supply chains in districts such as Bao'an and Longgang condensed development cycles significantly. While sourcing components and assembling prototypes became straightforward, securing repeat orders from tier-one global brands required audited facilities, established quality management systems, and institutional procurement relationships.

Leveraging his Foxconn background, Geng targeted procurement divisions at major tech firms, securing volume Original Equipment Manufacturer (OEM) contracts with Lenovo and IBM that integrated Sharetronic into international storage supply chains.⁶ The company's early product portfolio encompassed personal computer peripherals, video storage units, digital set-top boxes, and external hard-drive enclosures. Sharetronic later expanded into smart cameras and Internet of Things (IoT) hardware for 360 Group, China Mobile, and Xiaomi ecosystem partner 创米科技 Chuangmi.⁸

As an Original Design Manufacturer (ODM), Sharetronic developed reference designs, sourced semiconductors, completed tooling and regulatory certifications, manufactured finished units, and packaged them under third-party brand names. Brand owners retained retail marketing, distribution, and consumer margins, while Sharetronic earned conversion fees representing single-digit percentages of wholesale prices. In return, Sharetronic gained volume scale and operational insight into component supply chains.

In November 2014, Taiwanese connector and component manufacturer 正崴精密 Cheng Uei Precision acquired POWER CHANNEL LIMITED, securing an indirect stake in Sharetronic. Cheng Uei is chaired by 郭台强 Gou Tai-Chiang, younger brother of Foxconn founder 郭台铭 Terry Gou.⁷ Through this transaction, a Gou family enterprise became Sharetronic's second-largest shareholder, and Gou Tai-Chiang was later named honorary chairman of Sharetronic.

The margin trap, quantified

During its first decade as an OEM, Sharetronic generated steady cash flows alongside high customer concentration. In 2019, 2020, and 2021, the company's top five customers generated 76.39%, 87.37%, and 88.93% of total revenue, respectively. Lenovo alone represented over 50% of annual sales in one of those years and approximately one-third in another.⁸ Gross margins remained in the mid-teens; between 2021 and 2025, blended gross margin rose from 11.56% to 19.24%.⁹

This structural concentration exposed the company to standard contract manufacturing constraints. With nearly 89% of revenue tied to five clients, pricing was negotiated annually by customers with precise visibility into underlying bills of materials. Capital investments in factory automation primarily lowered unit costs for clients rather than expanding manufacturer margins. Historically, contract manufacturers escape this pattern by moving upstream into proprietary components or diversifying into proprietary service models.

Geng ultimately pursued both avenues. Achieving that pivot, however, required access to public equity markets—a milestone reached with Sharetronic's initial public offering in 2020.

---

↑ Back to Top

III. Going Public & The Storage/AIoT Pivot (2015–2021) [17:00 - 28:00]

By the mid-2010s, Sharetronic's product mix had begun shifting from basic enclosures to hardware integrated with silicon and embedded software. Solid-state drives (SSDs) became the company's largest single revenue line, while Wi-Fi smart cameras grew into its second-largest category.

This product distribution was reflected in the company's listing prospectus. When Sharetronic listed on the ChiNext board on July 27, 2020, at ¥9.30 per share, SSDs represented 33.6% of total revenue and smart cameras accounted for 39.7%, with mechanical hard drives, dashcams, smart mirrors, and routers comprising the remainder.¹⁰ The initial public offering issued approximately 51.64 million shares, raising net proceeds of about ¥434 million.¹⁰ While modest compared to subsequent capital raises, this marked the company's first external equity financing, funding an Internet of Things (IoT) smart-terminal production base, an R&D center, and working capital.

Beyond the immediate capital, the public listing fundamentally altered Sharetronic's financial flexibility. Going public provided three key instruments unavailable to a private contract manufacturer: equity currency for acquisitions, expanded bank credit channels, and a public market valuation that could support debt capacity. Management deployed all three over the subsequent six years.

The expansion of debt capacity proved particularly consequential. Commercial banks in China evaluate lending based on tangible collateral, public disclosure standards, and sponsor standing. While a private Shenzhen contract manufacturer generating ¥5 billion in sales with thin equity typically secures basic working capital lines, a ChiNext-listed firm—backed by institutional shareholders, a Taiwanese industrial group as a major investor, and audited financial statements—can access structured project financing. In operational terms, much of Sharetronic's subsequent expansion relied on translating public listing status into institutional borrowing capacity.

In 2022, Sharetronic executed a follow-on private placement on ChiNext sponsored by 天风证券 TF Securities. Combined with its IPO, total equity capital raised during its first two years as a public company reached approximately ¥1.2 billion.⁸ Though modest relative to later transactions, these proceeds funded automated manufacturing facilities across Shenzhen, Dongguan, and Anhui that established the baseline infrastructure for its eventual hardware deployment.

What Sharetronic actually does in storage

Understanding Sharetronic's core storage segment clarifies its operational role within the broader hardware supply chain. A solid-state drive primarily consists of NAND flash memory chips paired with a controller integrated circuit to manage data routing. Sharetronic neither manufactures NAND flash nor designs controllers. Instead, it purchases raw flash wafers and dies, conducts testing and grading, integrates them with controllers, tunes firmware, and completes high-volume assembly with low field-failure rates for white-label brand clients. Operating margins in this segment are narrow, working capital demands are substantial, and profitability remains sensitive to global NAND price cycles. Nevertheless, the business achieved significant operational scale: by the time Sharetronic prepared its Hong Kong listing prospectus, industry researcher Frost & Sullivan ranked it as China's second-largest smart storage device manufacturer by SSD revenue and its second-largest consumer smart camera maker by shipment volume, holding a 6.6% market share in smart cameras.¹¹

A 6.6% market share in consumer smart cameras provided procurement scale for component sourcing, but lacked the brand equity required to command premium pricing. Seeking to move beyond contract assembly margins, management announced an acquisition strategy in March 2021 to expand into software and cloud services.

The acquisition target was 西安思华信息技术有限公司 Xi'an Sihua Information Technology, a video-cloud and indoor-positioning software enterprise founded in September 2016. Sharetronic proposed acquiring 100% of Sihua for a valuation of ¥360 million—comprising roughly ¥191 million in newly issued shares and ¥169 million in cash—funded in part by a concurrent private placement of up to 35 investors, to which Geng personally committed up to ¥50 million.¹²

Management's acquisition rationale acknowledged the structural limits of contract manufacturing. As a hardware manufacturer, Sharetronic noted that it operated primarily as a front-end data collection provider, lacking proprietary software for high-margin data analytics and cloud storage. In its disclosures, the company stated that it was frequently forced to concede margins to third-party software vendors because it lacked an in-house platform.¹² While Sharetronic manufactured physical hardware, external software developers captured the higher-margin recurring revenues.

Financial disclosures for the target, however, highlighted significant valuation risk. Sihua generated ¥38.7 million in revenue and a net loss of ¥1.84 million in 2019, followed by ¥47.0 million in revenue and ¥2.80 million in net profit in 2020—yielding combined two-year revenue of ¥85.7 million and total net profit of just ¥960,000.¹² Valuing a company generating under ¥1 million in two-year net income at ¥360 million represented an aggressive premium for unproven software capabilities rather than demonstrated earnings power.

The ultimate resolution of the Sihua transaction differs from common market assumptions. While Sharetronic cited collaboration with Sihua to develop end-to-end video capture and cloud content storage capabilities under its proprietary "FCloud" platform, Sihua was never fully integrated into Sharetronic's consolidated financial statements. In its 2026 related-party transaction disclosures, Sharetronic designated Xi'an Sihua Information Technology and its subsidiaries as a related party, budgeting ¥300 million in transactions primarily for compute services.¹³ Consequently, the original deal structure did not yield a fully owned, consolidated software subsidiary.

The resulting FCloud platform provided an operational video management layer—enabling video recording, storage, retrieval, and retail-oriented AI vision features—that Sharetronic could bundle with its camera hardware. This integrated software layer enhanced hardware customer retention and enabled partial margin capture, though it did not scale into a standalone subscription software business capable of altering overall group economics. Sharetronic subsequently reassigned the FCloud brand to its AI compute training and inference platform, pivoting the software initiative from standalone application sales toward compute service enablement.

Evaluating Sharetronic's software strategy highlights a clear strategic diagnosis alongside constrained operational execution. Management correctly identified the margin limitations of pure hardware assembly, but the effort yielded a partner-driven platform rather than a high-margin, consolidated software P&L. Cloud and software revenues failed to develop into a material contributor to group earnings.

With software unable to resolve the company's structural margin compression, Sharetronic's ultimate business transformation originated from a different vector. Rather than shifting away from hardware, the company expanded further into physical infrastructure—supplying specialized hardware under a rental and compute-service framework.

↑ Back to Top

IV. The Inflection Point: Pivoting to AI Computing Infrastructure & Remanufacturing (2022–2024) [28:00 - 41:00]

In February 2026, during an interview with 21st Century Business Herald (21世纪经济报道), Geng Kangming addressed the ¥21.2 billion in server purchases his company had announced across five separate filings in 2025—a sum roughly three times its prior-year revenue.

His response was direct: "这笔投入其实非常保守" ("This investment is actually very conservative"), adding that "客户的需求远远大于我们的投入" ("Customer demand far exceeds our spending").¹⁴

He then offered a framework that has defined market discussions ever since, comparing the contemporary AI compute sector to the Chinese real estate market of the 1990s and suggesting the industry had two decades of growth before reaching an expansion peak comparable to property in 2015.¹⁴

That analogy is revealing. China's real estate expansion between 1998 and 2015 was among the largest wealth-creation cycles in modern economic history, driven less by product differentiation than by early land acquisition and aggressive financial leverage. Geng's comparison serves as an explicit acknowledgment of Sharetronic's strategy: treating compute deployment as a land grab where scale, execution speed, and balance-sheet leverage take precedence over proprietary differentiation. Yet for observers aware of the real estate sector's post-2015 trajectory, invoking property development highlights the structural risks of debt-fueled asset expansion.

How did a contract manufacturer of smart cameras and solid-state drives execute this pivot? Two distinct operational threads converged.

Thread one: a shortage nobody could arbitrage away

The first thread was structural demand. Beginning in late 2022, Chinese technology conglomerates, AI startups, cloud-gaming platforms, and rendering studios sought high-performance GPU capacity simultaneously. Under tightening export controls, advanced accelerators became scarce and expensive. Enterprises requiring compute capacity for multi-month model training runs preferred renting infrastructure over purchasing server clusters that entailed five-year depreciation schedules. This shift created an opening for asset owners willing to assume equipment depreciation risk in exchange for recurring rental yields.

In practice, AI compute leasing operates as an equipment rental business with rapidly depreciating hardware assets. Sharetronic acquires server racks containing high-performance accelerators, installs them in third-party data centers equipped with power, cooling, and networking infrastructure, and executes lease contracts with enterprise clients for exclusive or shared compute access. Sharetronic retains hardware ownership, ongoing maintenance obligations, and residual asset risk.

The segment's underlying economics depend on four primary variables: unit acquisition cost, annual rental yields, asset utilization rates, and economic useful life before power costs exceed operating value. A server rack purchased for 100 units generating 35 units in annual rent at 90% utilization over four years represents a high-return asset. The same hardware yielding 20 units at 60% utilization funded at 6% interest destroys capital. Public disclosures currently provide limited granular utilization data, leaving analysts to evaluate where Sharetronic falls between these two economic outcomes.

A second-order dynamic unique to China's domestic market reinforced this opportunity. While capital availability typically resolves hardware shortages in unconstrained markets, access to accelerators in China remained governed by procurement allocations. This structure shifted competitive advantage from capital scale alone toward component procurement relationships—an area aligned with Sharetronic's contract manufacturing origins. Furthermore, operating in an environment of constrained supply required managing heterogeneous server fleets, integrating imported accelerators alongside domestic neural processing units (NPUs) into unified clusters. This integration relies on systems engineering and supply chain coordination rather than native semiconductor design.

Thread two: the business almost nobody else wanted

The second thread was the development of a specialized server remanufacturing operation. Through a controlled subsidiary, Semsotai North Inc., Sharetronic collects and dismantles retired enterprise servers in North America. Purchased equipment and components are transferred to Sharetronic's Shenzhen Yantian branch and bonded-zone branch, which handle R&D, production, sale, and related technical and repair services.¹⁵

This operation exploits a structural inefficiency in global data center lifecycles. Enterprise operators frequently decommission servers after three years to optimize power-efficiency ratios, despite hardware components remaining functional. By acquiring decommissioned hardware at low cost, testing and grading components, and reassembling warranted server configurations, Sharetronic arbitrages the valuation gap between hyperscaler retirement cycles and price-sensitive domestic enterprise demand.

This model draws directly on Foxconn-style operational management. Remanufacturing at scale requires rigorous component testing, yield optimization, and field-failure control—disciplines Geng developed during his early career. Additionally, established server manufacturers generally avoid secondary remanufacturing to prevent cannibalizing primary hardware sales, leaving the market open to independent operators.

Conviction becomes commitment

Sharetronic's procurement filings through 2025 reflected this strategic shift. Over the course of the year, the company issued five purchase announcements—¥3 billion, ¥4 billion, ¥1.2 billion, ¥4 billion, and ¥9 billion—totaling ¥21.2 billion, of which roughly ¥12.2 billion had been formalized across five filings by the autumn.⁵¹⁶ In February 2026, Sharetronic announced a sixth procurement plan of ¥11 billion. Because the transaction exceeded 50% of recently audited net assets and total assets, it required a shareholder vote.⁵ Cumulative announced server procurement passed ¥32 billion, representing more than four times the company's total 2024 revenue of ¥7.4 billion.⁶

The financial impact materialized rapidly. Net profit grew 119.46% in 2023 and 140.80% in 2024, from a small base.⁵ By 2025, segment contributions fundamentally altered group financials. Group revenue reached ¥12.236 billion; intelligent compute products and services contributed ¥2.761 billion, server remanufacturing added ¥2.575 billion (up 169.35%), and data storage devices grew 28.31% to ¥4.493 billion.² In the fourth quarter of 2025, revenue rose 93.77% year-over-year to ¥3.905 billion, while net profit increased 246.15% to ¥466 million, achieving a 22.62% gross margin.⁹

Conversely, legacy Internet of Things (IoT) smart terminals fell 30.42% to ¥1.572 billion, and its gross margin collapsed from 41.33% to 27.65%.⁹ While management attributed this decline to a strategic focus on higher-margin models, the simultaneous reduction in volume and margin reflects competitive pressure in white-label consumer hardware.

The primary financial indicator in the 2025 financial statements appears on the cash flow statement. Operating cash flow reached ¥1.100 billion, up 1,011%, while investing cash flow stood at negative ¥13.685 billion, offset by positive financing cash flow of ¥13.591 billion.² These metrics demonstrate a leverage-driven capital deployment model: Sharetronic borrowed approximately thirteen yuan in external financing for every yuan of operating cash generated, converting that capital into server infrastructure. The long-term return on this capital depends on sustained rental yields and hardware asset longevity.

Evaluating these return dynamics requires analyzing segment unit economics and balance-sheet sustainability.

---

↑ Back to Top

V. Segment Breakdown & Financial Economics: Where the Dollars Actually Flow [41:00 - 53:00]

A popular narrative on Chinese retail trading forums presents Sharetronic as a pure-play AI compute provider. Disclosed segment metrics, however, tell a more nuanced story. As of the 2025 full-year financial accounts, compute services were the fastest-growing and most high-profile division, but they had not yet become the largest contributor to group revenue.

Examining the actual segment weights makes this clear. Data storage devices remained the primary revenue driver at ¥4.493 billion, accounting for 36.72% of total group revenue.⁹ Intelligent computing products and services ranked second, generating ¥2.761 billion, or 22.57%.² Server and peripheral remanufacturing brought in ¥2.575 billion, while legacy Internet of Things (IoT) smart terminals contributed ¥1.572 billion.⁹ Cloud and software revenue was not broken out as a standalone material segment—providing a definitive answer on how the company's 2021 software ambition ultimately evolved.

Analyzing the individual segment economics highlights the distinct operational drivers behind each line.

Intelligent compute. This division represents the capital-intensive equipment rental model: Sharetronic purchases GPU servers, installs them into clusters, and leases compute capacity to enterprise clients running model training, inference workloads, cloud video, and cloud gaming. In the first half of 2025—the initial period in which compute was disclosed separately—the segment generated ¥1.221 billion in revenue at a gross margin of roughly 20.6%.¹⁶ While a 20% gross margin is respectable for a hardware-heavy leasing operation, it sits well below the hyper-profitable levels implied by retail AI narratives. Crucially, this gross margin is calculated after server depreciation but before financing costs. For a business funded largely through debt, interest expenses are substantial. Over the first nine months of 2025, Sharetronic’s finance costs surged 2,277% year-over-year to ¥178 million, with interest expense jumping 454% to ¥196 million.¹⁷ Ultimately, the profitability of this business hinges on the spread between rental yields and borrowing costs—a margin far narrower than headline top-line growth rates suggest.

Even so, signs of operating leverage have begun to emerge. Group gross margin improved to 22.62% in the fourth quarter of 2025,⁹ and reached 22.53% in the first quarter of 2026.⁴ This stability indicates that as compute clusters reach higher utilization, scaling efficiencies are helping offset financing headwinds.

Data storage. As the company's cash-generative foundation, data storage grew 28% in 2025, fueled by rising AI-driven demand for enterprise solid-state drives (SSDs) alongside steady consumer volumes. Economically, this operates as a spread business based on NAND flash prices. When underlying memory chip costs drop, Sharetronic's margin expands; when memory prices spike, margins compress unless cost increases can be passed along to customers. Management has noted heightened volatility in memory markets, relying on long-term supply agreements and dynamic pricing terms to mitigate exposure.¹⁸ Yet an important governance detail remains: Sharetronic's 2026 related-party disclosures budgeted ¥3.05 billion for storage chip procurement from Sharetronic Data Compute (Hong Kong) Co., Ltd., a designated related entity.¹³ Investors must recognize that a major portion of the company's primary raw input routes through a related-party channel.

Remanufacturing. Growing 169% in 2025 and 119.49% in the first half of the year,¹⁴ remanufacturing represents perhaps Sharetronic's most operationally distinct initiative. Although management does not report a standalone gross margin for this unit, its financial mechanics differ fundamentally from compute leasing. Remanufacturing is a working-capital business centered on inventory turnover rather than long-term asset depreciation, where key operational risks stem from component sourcing availability and cross-border trade compliance rather than cluster utilization.

IoT terminals. Though in managed decline as a proportion of total revenue, legacy IoT terminals remain relevant for two reasons. They represent the manufacturing origin of Sharetronic's customer relationships and production facilities, and they are designated to receive a portion of the ¥1.814 billion in storage-expansion capital from the pending equity offering.

Does the stack actually circulate?

Management frames these four business lines as an integrated ecosystem, but evaluating whether the stack genuinely circulates requires closer scrutiny. In theory, assembling enterprise SSDs at scale allows Sharetronic to equip its own AI compute clusters at manufacturing cost rather than commercial retail prices. Simultaneously, harvesting components from decommissioned servers could feed usable drives and memory back into both its storage manufacturing and server build-outs. If this circular integration functions as intended, the four segments operate as a single supply chain with multiple monetization points, driving group gross margins higher as internal integration deepens.

While recent margin expansion aligns with this hypothesis, it does not offer definitive proof. Sharetronic does not disclose detailed intersegment transfer pricing, making it impossible to separate genuine vertical integration benefits from favorable component pricing cycles. Furthermore, rapid shifts in segment composition make single-period snapshots misleading. In the first half of 2025, storage devices accounted for 37.1% of revenue (¥1.835 billion), compute services represented 24.7%, remanufacturing stood at 16.9%, and IoT terminals held 17.7% of a ¥4.943 billion total.¹¹ By the fourth quarter, growth was driven almost entirely by compute and remanufacturing. Financial modeling based on any single reporting period risks misinterpreting a rapidly shifting revenue mix.

Two broader corporate metrics add important context to this operating picture. First, inventory surged 48.1% in a single quarter to reach ¥4.261 billion by the end of March 2026, comprising mostly storage components and compute products.⁴ While high inventory poses structural valuation risks during market downturns, holding key hardware inventory in a supply-constrained environment can serve as a critical operational advantage.

Second, research and development trends present a notable divergence. R&D spending grew 75.92% in 2025 to ¥423 million, supporting a portfolio of 437 domestic patents (including 50 invention patents) and eight overseas patents.² However, R&D intensity has steadily declined: R&D expenses dropped from 3.3% of revenue in 2022 and 3.4% in 2023 to 3.2% in 2024, falling further to 2.2% in the first half of 2025. Over the same period, R&D headcount shrank from 384 employees at year-end 2024 to 361 by mid-2025.¹⁷

For a company positioning itself as an advanced technology platform, spending a shrinking proportion of revenue on R&D while reducing technical staff conflicts with standard industry narratives. Instead, these metrics reinforce an alternative view: Sharetronic’s core strength lies not in proprietary semiconductor design, but in low-cost manufacturing, supply chain integration, and rapid capital deployment.

That operational profile is not an inherent flaw—disciplined assembly, procurement scale, and balance-sheet execution form the foundation of contract manufacturing. However, financial markets value capital-deployment platforms differently from proprietary technology developers due to higher cyclicality and lower barriers to entry. Investors must evaluate which business model they are underwriting—and why the company's next growth chapter relies heavily on acquiring technical capability externally rather than building it in-house.

↑ Back to Top

VI. Capital Deployment & M&A Strategy: Guangwei, Sihua, and the "Compute-Connect-Store" Stack [53:00 - 64:00]

Buying the wire

On April 11, 2026, Sharetronic announced it would inject ¥510 million into 光为科技（广州）有限公司 Guangwei Technology (Guangzhou) Co., Ltd. in exchange for 51% of the equity.¹⁹

The strategic rationale is driven by the physical demands of high-performance computing. Interconnecting hundreds or thousands of GPUs to train a single large AI model creates a data transmission bottleneck. Because copper cabling cannot carry high data volumes over necessary distances without excessive power consumption, data centers rely on optical transceiver modules to convert electrical signals into light and back. As cluster scale expands, module speeds have advanced from 400 gigabits per second to 800G and toward 1.6 Terabits per second (1.6T). Guangwei designs optical interconnect chipsets and modules operating across this 400G-to-1.6T range.¹⁹

Securing a 51% stake in an optical module manufacturer alongside server integration and storage assembly completes what Sharetronic terms its "compute-connect-store" stack. In theory, controlling the compute node, network fabric, and storage tier allows Sharetronic to manage three of the four largest cost components in a data center deployment—leaving third-party accelerators as the primary external procurement cost.

However, assessing the transaction requires examining financial reality rather than strategic positioning alone. Guangwei is not a mature, low-cost asset. As of December 31, 2025, it reported total assets of ¥191.34 million, total liabilities of ¥140.61 million, shareholders' equity of ¥50.73 million, revenue of ¥77.70 million, and a net loss of ¥20.89 million.¹⁹ Sharetronic's ¥510 million investment for a 51% stake implies a post-money valuation of ¥1.0 billion against a pre-money equity value of roughly ¥490 million.¹⁹ Relative to 2025 sales, that represents a post-money multiple of roughly 13 times revenue for an unprofitable enterprise generating under ¥80 million annually. Characterizing the transaction as a bargain relative to listed optical peers such as 中际旭创 InnoLight or 新易盛 Eoptolink is inconsistent with Guangwei's disclosed financial metrics.

That valuation does not necessarily negate the transaction's rationale. Early-stage optical module and chipset developers often command venture-scale multiples, and ¥510 million represents less than half of a single quarter's net profit for Sharetronic at mid-2026 run rates—effectively a modest option premium on a critical supply chain node. Still, it functions as a venture bet on an unproven, unprofitable entity rather than a value acquisition. The key test will be whether Guangwei's optical components are successfully integrated into Sharetronic's compute clusters, as vertical integration without internal adoption yields only a passive equity holding.

A broader pattern emerges when examining Sharetronic's three primary capability acquisitions. The Sihua transaction represented an initial attempt to add software capabilities, resulting in a commercial platform and partnership rather than a consolidated software division. The Semsotai North structure proved far more economically productive, establishing a proprietary server-sourcing channel that grew rapidly. Guangwei marks the third installment in this strategy.

Viewed together, these transactions reflect a management strategy of small, early-stage, adjacent investments—avoiding transformational deals at prices that could jeopardize balance-sheet stability, while targeting segments where manufacturing capabilities provide operational insight. This disciplined M&A pattern contrasts with the value-destructive acquisitions seen among some Chinese mid-caps pursuing emerging tech trends. The counterpoint is that none of the three transactions has yielded an independent, consolidated profit center, suggesting the strategy currently reflects disciplined risk control more than proven value creation.

Ultimately, the company's largest capital allocation decision has not been M&A, but rather the choice to hold compute assets directly on its balance sheet rather than acting as a hardware broker or asset-light manager. Owning hardware maximizes potential returns during market upswings while concentrating asset-depreciation and debt-service risks—making asset ownership the central capital allocation call by which Sharetronic will be evaluated.

The financing machine

That shifts the focus directly to the financing structure supporting Sharetronic's compute buildout, where debt levels have expanded significantly. By the end of 2025, short-term borrowings reached ¥4.292 billion and long-term borrowings stood at ¥7.483 billion.²⁰ Earlier that year, total borrowings had risen to ¥12.811 billion by the end of August 2025—up over ¥10 billion from December 2024—with ¥6.292 billion due within 12 months.¹¹ The company's borrowing rate range also broadened, shifting from 2.4%–4.4% in 2024 to 2.2%–6.9% in the first half of 2025.¹⁶ The emergence of a 6.9% upper interest rate indicates that lenders began pricing in higher credit risk as leverage increased. For 2026, the board authorized credit line applications of up to ¥20 billion from banks and up to ¥80 billion in finance-lease capacity.²⁰ While these figures represent authorization ceilings rather than drawn debt, they underscore the scale of planned expansion.

To balance this leverage, Sharetronic disclosed a plan on June 26, 2026, to raise up to ¥8 billion through a private equity placement to no more than 35 investors. The proposed allocation targets ¥4.686 billion for intelligent computing center construction (out of a ¥4.930 billion total project cost), ¥1.814 billion for data storage expansion and upgrades (against a ¥2.054 billion total cost), and ¥1.5 billion for working capital and bank loan repayment.¹⁸²⁰ Shareholders approved the proposal on July 13, 2026.¹ Given that three-to-five-year depreciating assets are poorly suited to 6.9% bank debt, the equity issuance serves primarily as a partial deleveraging measure rather than purely expansionary capital.

Shortly thereafter, on August 4, 2026, Sharetronic disclosed a request for authorization to allocate up to ¥7 billion of internal funds to entrusted wealth management, including up to ¥6 billion in agreement deposits.²¹ For a company carrying an 85.89% asset-liability ratio, parking ¥7 billion in bank deposits reflects working-capital cash flow dynamics rather than surplus capital. Management attributes these cash balances to concentrated customer collections and upfront prepayments linked to large-scale compute cluster deployments. Substantial prepayments would indicate strong end-user demand and commercial leverage; however, because Sharetronic has not disclosed the precise breakdown between collections and advance deposits, this explanation remains a management assertion requiring verification in future reporting periods.

Whether this capital allocation strategy translates into a durable, risk-adjusted business model remains the central question facing investors.

↑ Back to Top

VII. Playbook & Strategic Frameworks: Helmer's 7 Powers & Porter's 5 Forces [64:00 - 74:00]

                       SHARETRONIC "COMPUTE-CONNECT-STORE" MATRIX
   ┌─────────────────────────────────────────────────────────────────────────────────┐
   │                                  FCloud SaaS                                    │
   ├─────────────────────────┬─────────────────────────────┬─────────────────────────┤
   │   COMPUTE (算)          │       CONNECT (联)          │       STORE (存)        │
   │   AI GPU Servers &      │   Guangwei 800G/1.6T        │   Enterprise SSDs &     │
   │   Remanufactured Nodes  │   Optical Modules (51%)     │   Storage Arrays        │
   └─────────────────────────┴─────────────────────────────┴─────────────────────────┤
   │                     Foxconn-Grade Low-Cost Supply Chain & Yield                 │
   └─────────────────────────────────────────────────────────────────────────────────┘

The matrix above reflects how Sharetronic seeks to be understood. The more useful analytical exercise, however, is asking which layers are truly load-bearing.

Seven Powers, honestly applied

Hamilton Helmer's 7 Powers framework asks whether a business possesses a structural condition that yields persistent, above-cost-of-capital returns that competitors cannot easily erode. Applied rigorously to Sharetronic, the balance sheet yields one plausible power, one genuine counter-positioning wedge, and a massive cyclical tailwind that should not be mistaken for a durable moat.

Process power — plausible, but unproven. Helmer sets a high bar for process power: an embedded operational capability built over years that rivals cannot replicate quickly even with full visibility into how it works. Toyota's production system is the canonical reference. Sharetronic claims its Foxconn-derived manufacturing discipline enables it to refurbish, test, integrate, and deploy server clusters faster and cheaper than conventional OEMs. The circumstantial evidence is visible—a server remanufacturing segment that grew 169% in 2025 without disclosed margin degradation, gross margins expanding alongside volume scale, and an executive team trained explicitly in production yield. The counter-evidence is equally stark: research and development intensity is falling while R&D headcount shrinks—rare traits for a true process-power enterprise. Verdict: plausible, under-disclosed, and a metric to monitor rather than assume.

Counter-positioning via remanufacturing — genuine, yet constrained. The remanufacturing channel represents a textbook counter-positioning play against incumbent server manufacturers. Established OEMs such as 浪潮信息 Inspur Information and 中科曙光 Sugon cannot aggressively market refurbished, three-year-old servers at a fraction of new-unit list prices without cannibalizing their core sales models. The limitation is supply-side dependency: the model relies on acquiring retired hyperscaler hardware overseas—a sourcing pipeline subject to regulatory and trade policy risks outside Sharetronic's control.

Scale economies — emerging, but unproven. Sharetronic purchases memory, motherboards, and optical components in expanding quantities, generating modest procurement leverage. However, the company's single most critical input—high-performance GPU accelerators—remains supply-constrained. In a tight market, scale secures supply allocation rather than price discounts. Once hardware bottlenecks ease, that allocation advantage evaporates.

Absent: brand, network effects, and switching costs. A rented GPU cluster is a functional commodity. While clients sign multi-year service agreements—with Geng noting that major contracts typically span five years¹⁴—a contractual commitment is a legal obligation rather than a structural moat. Compute capacity lacks data gravity, proprietary software ecosystems, or developer lock-in; when contract terms expire, lease rates reset to prevailing market spot prices.

Five Forces, and who is actually across the table

Porter's Five Forces framework clarifies the underlying competitive dynamics.

Supplier power: high. Sharetronic does not design or fabricate accelerators. It purchases semiconductors within a market shaped by U.S. export restrictions, where procurement remains governed by supply allocations. As Geng acknowledged at the July 2026 shareholder meeting, domestic GPU inventory had sold out by January 2026¹—a statement intended to highlight strong demand that simultaneously illustrates severe supply constraints facing buyers.

Buyer power: moderate, but rising. Current compute scarcity grants asset owners pricing leverage, driving recent margin expansion. However, Sharetronic's customer base is concentrated among major internet conglomerates and AI enterprises with formidable bargaining power. Given the company's historical dependency on top customers during its OEM era—where five clients generated up to 89% of revenue—buyer concentration remains a structural risk, even as Sharetronic has yet to disclose granular customer breakdowns for its compute division.

Threat of new entrants: moderate and accelerating. Capital represents the primary barrier to entry, and Chinese equity markets supplied substantial funding to compute operators throughout 2025 and 2026. Established IT distributors such as 神州数码 Digital China possess the balance-sheet scale and enterprise relationships required to enter compute leasing. Their primary operational gap is Sharetronic's specialized remanufacturing pipeline.

Threat of substitutes: low near-term. Direct substitutes for raw compute capacity do not exist for model training. The primary substitution threat stems from enterprise clients building proprietary compute centers rather than renting capacity as accelerator supply normalizes.

Industry rivalry: intense and intensifying. Sharetronic competes against state-backed server champions with lower capital costs, industrial powerhouses like 工业富联 Foxconn Industrial Internet operating at vastly greater assembly scale, and a long tail of independent leasing firms offering undifferentiated compute capacity.

Evaluating this competitive landscape requires distinguishing among distinct market segments, as "AI compute in China" encompasses several discrete business models.

State-backed server OEMs like Inspur Information and Sugon design and sell hardware directly to state-owned enterprises, telecommunications carriers, and government agencies, benefiting from procurement preferences in public tenders that Sharetronic cannot match. While superior in R&D depth and institutional channels, these incumbents cannot easily replicate Sharetronic's secondary remanufacturing model.

Foxconn Industrial Internet—Geng's former employer at mega-scale—assembles AI servers for global hyperscalers in volumes far exceeding Sharetronic's throughput. Yet it operates as a contract manufacturer for third-party brand owners rather than an owner-operator retaining server assets on its balance sheet, representing a fundamentally different risk profile.

Distributors like Digital China leverage deep corporate client relationships and favorable corporate financing, though they lack in-house manufacturing and remanufacturing infrastructure.

Finally, a broad network of regional compute operators—frequently backed by municipal governments establishing local intelligent computing centers (智算中心)—competes directly on rental rates using subsidized land, power, and capital.

This municipal cohort represents a critical, understated competitive threat. Government-backed compute centers operate under policy mandates rather than market-driven hurdle rates. If localized buildouts flood regional markets with capacity, marginal lease rates will be set by operators unconstrained by capital return requirements. Sharetronic's ability to navigate this environment depends on maintaining input costs low enough to preserve margins—reaffirming why the longevity of its remanufacturing engine remains paramount.

The strategic synthesis is clear. Sharetronic's operational advantage centers on two specific capabilities: its secondary remanufacturing pipeline and its rapid deployment speed. Its primary growth driver, compute leasing, remains a capital-intensive commodity business whose elevated returns reflect temporary supply shortages rather than an enduring structural moat. The company has achieved strong profitability by deploying leveraged capital early into a supply-constrained asset class. That represents a classic timing advantage—highly lucrative during market expansions, yet inherently non-durable when supply balances normalize.

Consequently, the central investment question turns on management's ability to execute through changing market cycles.

↑ Back to Top

VIII. Management Credibility & The Skeptical Investor Stress Test [74:00 - 83:00]

The record so far

In July 2025, Forbes China named Geng Kangming to its Best CEOs list as the sole honoree from the computers and peripherals category. The selection methodology weighted market capitalization, share price performance, net profit growth, and return on equity across two consecutive years, during which Sharetronic's market valuation expanded from roughly ¥3.97 billion at the start of 2023 to approximately ¥28 billion by mid-July 2025.[^22]

Industry awards reflect market momentum rather than definitive proof of long-term value. However, evaluated on operational execution—specifically whether leadership delivers on its stated targets—Geng's record is solid. The pivot into compute was announced, funded, executed, and monetized within roughly two years. Management has consistently met or surpassed its earnings guidance: the 2025 pre-announcement net profit range of ¥1.05 billion to ¥1.25 billion⁵ materialized at ¥1.164 billion; first-quarter 2026 net profit reached ¥750 million, up 343.45%;⁴ and first-half 2026 guidance of ¥1.5 billion to ¥1.9 billion implies second-quarter net profit between ¥750 million and ¥1.15 billion.³ Meeting or exceeding guidance across four consecutive reporting periods demonstrates operational reliability.

Communication across channels has also remained consistent. The operational framework Geng outlined privately to analysts in February 2026—highlighting supply shortages, five-year contract durations, and debt financing via bank loans, finance leases, and shareholder borrowing¹⁴—mirrored his public statements to shareholders five months later. Management has not altered its stated growth drivers or attributed performance to shifting external conditions, largely because the company has yet to experience an operational miss. Crucially, management credibility is tested most rigorously during downturns, which Sharetronic has not yet faced. A stress test must therefore focus on structural balance-sheet exposure rather than historical crisis management.

Evaluating the business requires addressing the primary concerns of institutional investors.

Stress test one: the balance sheet. Balance-sheet leverage represents the central financial risk. Sharetronic's asset-liability ratio increased steadily over five consecutive quarters, rising from 55.66% at the end of 2024 to 75.94% by mid-2025, 78.33% at the end of the third quarter, 81.48% at year-end, and reaching 85.89% by March 31, 2026.⁴¹⁷ As of the first quarter of 2026, total assets reached ¥36.9 billion against total liabilities of ¥31.693 billion.⁴ Furthermore, the company's Hong Kong listing prospectus disclosed a net gearing ratio of 256.4% at mid-2025—an increase of more than 180 percentage points over six months.¹¹ Interest-bearing debt of approximately ¥25.9 billion now stands against a comparatively modest equity base.¹

Management defends this leverage profile on three grounds. First, it asserts that customer prepayments and efficient collections fund ongoing working capital, a claim supported by the recent ¥7 billion treasury authorization. Second, it notes that the pending ¥8 billion private equity placement will directly reduce debt levels. Third, operating cash flow performance has strengthened: following a negative ¥1.374 billion across the first nine months of 2025,¹⁷ full-year operating cash flow recovered to positive ¥1.100 billion, while first-quarter 2026 generated ¥272 million, up 146% year-over-year.²⁴

An objective evaluation indicates that while cash conversion metrics have improved and equity financing is the appropriate tool, an ¥8 billion capital injection moderates debt obligations rather than completely restructuring the balance sheet. Sharetronic acknowledged in its placement filings that expanding debt levels increase overall financial risk.²⁰ Operating at nearly 86% leverage within a capital-intensive sector subject to rapid hardware obsolescence means an unexpected slowdown in compute demand would strain debt covenants alongside earnings.

Stress test two: depreciation and obsolescence. Hardware assets depreciate rapidly, with Sharetronic's annual depreciation expense already surpassing ¥621 million.⁶ At the July 2026 shareholder meeting, Geng addressed this risk by citing cost timing, stating that server clusters acquired for slightly over ¥3 million in 2023 had reached market valuations exceeding ¥14 million—reflecting a fourfold price surge that highlighted the early-mover advantage.¹ He similarly rejected concerns over market overcapacity, describing compute demand as vast and asserting that server clusters across the industry remained fully utilized.¹

Neither claim can be independently verified from public disclosures. While hardware appreciation occurred during the acute accelerator shortages of 2023 and 2024, price escalation reflects market scarcity rather than an intrinsic asset characteristic. Relying on secondary market appreciation to dismiss depreciation risk emphasizes how strongly unit economics depend on sustained supply constraints. Similarly, because Sharetronic does not publish cluster utilization rates, statements regarding full capacity utilization remain unverified operational claims.

An analytical paradox also exists within this framing. If compute hardware appreciates substantially above acquisition costs, liquidating equipment offers immediate capital gains. Continuing to lease the fleet indicates management expects long-term rental yields to exceed net resale values. While that operational choice is logical, it sits in tension with using asset appreciation arguments to downplay structural depreciation risks. Investors should view historical hardware appreciation as a reflection of prior acquisition timing rather than a model for future asset residual value.

Stress test three: what cannot be verified

Stress test three: disclosure and governance. Certain disclosure practices warrant scrutiny from institutional investors.

Counterparties to major server procurement announcements have been withheld on grounds of commercial confidentiality, while procurement execution shifted from the parent entity to operating subsidiaries in subsequent filings.⁵ Additionally, Sharetronic secured regulatory exemptions from disclosing granular contract terms and declined to provide specific customer details during shareholder meetings.¹ Meanwhile, projected related-party transactions for 2026 rose to ¥7.8 billion. This total includes ¥3.05 billion in storage chip purchases from a Hong Kong related entity, ¥850 million in compute services provided to MetaBase (Beijing) Technology—an entity whose equity Sharetronic transferred to an independent third party in February 2026, removing it from consolidated accounts—alongside ongoing transactions with Sihua and Qishuo (Shenzhen) Technology.¹³ While these structures comply with reporting standards, they restrict external analysis of arm's-length demand and customer concentration.

Insider transactions provide further context. Sharetronic's controlling shareholder divested approximately 7.26 million shares (2.98%) for ¥284 million in 2024, followed by an additional 7.36 million shares (3.00%) for ¥696 million in February 2025 at a weighted average price of ¥94.37—totaling roughly ¥980 million over the two-year period.²² In April 2026, POWER CHANNEL LIMITED, the Cheng Uei-linked second-largest shareholder, initiated a non-public inquiry transfer of a 3% equity stake. Valued at approximately ¥3.29 billion based on the April 17 closing price of ¥316.90, the transaction was completed at ¥241.37 per share, subject to a six-month buyer lockup, and drew demand 2.65 times oversubscribed.²²

Proponents highlight that alongside the equity transfer, POWER CHANNEL committed up to ¥2 billion in financial support through shareholder and entrusted loans at interest rates capped at 4.5%, designated for compute infrastructure, research, and talent acquisition.²³ Providing capital at 4.5% to a firm borrowing at market rates up to 6.9% represents tangible financial support.

Nevertheless, the broader pattern remains clear. Over a two-year period in which management requested shareholder authorization for an ¥8 billion equity issuance to support expanded borrowing, the two primary shareholders monetized approximately ¥4.3 billion in equity. While portfolio diversification and structured inquiry transfers with lockup provisions represent orderly transaction methods, substantial equity sales by core insiders warrant careful consideration by institutional investors evaluating long-term operational conviction.

---

↑ Back to Top

IX. The Investment-Story Spine & Risk Radar: Why Sharetronic Wins (and Why It Might Not) [83:00 - 90:00]

Strip away market commentary, and Sharetronic's core investment thesis reduces to a single testable proposition: whether a manufacturer with an unusually low cost of acquiring and deploying compute assets can earn a spread over its cost of capital long enough to service its debt before those assets become obsolete.

Everything else is detail.

That proposition does not require artificial intelligence demand to expand indefinitely—only that utilization holds long enough to cover a defined amortization window. Nor does it depend on Sharetronic achieving technological breakthroughs, or on its optical and software initiatives generating outsized returns. The business relies on two core variables: maintaining high cluster utilization and ensuring rental income comfortably exceeds combined interest and depreciation expenses. Every bullish and bearish scenario ultimately reflects one of those two conditions.

Evaluating both sides clarifies where operational momentum meets structural balance-sheet constraints.

Why it wins from here.

The stack is real where it counts. Strategic advantage lies not in branding, but in component sourcing. Sharetronic's server remanufacturing channel secures an input cost advantage on compute and storage hardware that competitors purchasing new equipment cannot match. Having expanded that segment from inception to over ¥2.5 billion in revenue within two years, Sharetronic occupies a distinct niche: providing price-sensitive enterprise clients with 30% to 50% cheaper compute capacity on previous-generation hardware. That positioning remains defensible as long as decommissioned enterprise hardware continues to flow through its sourcing channels.

Execution velocity is demonstrated, not asserted. Sharetronic progressed from announcing its entry into AI compute to generating ¥2.761 billion in segment revenue within eighteen months. First-quarter revenue in 2026 jumped 193% compared to the prior-year period.⁴ Beyond balance-sheet leverage, the underlying operational infrastructure functions effectively. Management's Foxconn-trained background has enabled faster deployment than state-backed peers, allowing the company to convert hardware supply during tight market conditions directly into long-term contracted revenue.

The margin trend supports the operating-leverage thesis. Blended gross margin expanded during this rapid volume growth, rising from 11.56% in 2021 to 19.24% for full-year 2025, and reaching 22.62% in the fourth quarter.⁹ Expanding gross margins alongside scaling revenue indicate that top-line growth reflects genuine operational leverage rather than margin-sacrificing volume accumulation.

Why it may not.

The equity raise must convert to yield fast. Deploying ¥8 billion into hardware clusters that run at sub-optimal utilization would destroy capital. Sharetronic has not previously deployed capital of this magnitude in a single cycle, and its pending equity placement arrives as industry-wide compute capacity accelerates across China. As management acknowledged in its Hong Kong prospectus disclosures, an inability to secure financing on acceptable terms could force the company to delay, curtail, or terminate its expansion plans.¹⁶

A capex digestion pause poses a primary structural risk. Rather than an industry-wide slowdown, a multi-quarter pause in procurement by major Chinese technology platforms would depress utilization rates and force rental yields lower. On a balance sheet with an asset-liability ratio approaching 86%, fixed depreciation and escalating interest expenses would collide with softening revenue. In that scenario, operating leverage would operate in reverse.

Verification risk remains a practical constraint. Confidential counterparty designations, ¥7.8 billion in budgeted related-party transactions, disclosure exemptions on contract details, and the absence of published cluster utilization rates prevent external investors from independently verifying revenue quality. Given past industry instances where announced compute contracts failed to materialize as projected, this information gap increases the implied cost of capital.

Myth versus reality

Three widespread market assumptions warrant examination against regulatory disclosures.

Myth: Sharetronic is a pure-play AI compute provider. Reality: Based on 2025 financial disclosures, data storage devices remained the largest revenue segment, with intelligent compute contributing 22.57% of total sales. Including remanufacturing brings the AI-adjacent share to roughly 40%. While segment composition is shifting rapidly, current revenues reflect an enterprise storage and hardware manufacturer with an expanding compute division attached.

Myth: The Guangwei acquisition demonstrated strict valuation discipline. Reality: The transaction represented a venture-style investment in an unprofitable enterprise valued at approximately thirteen times trailing revenue. While small enough to avoid straining group liquidity, the deal reflects early-stage capability acquisition rather than value-focused asset pricing.

Myth: Software and cloud capabilities provide recurring, high-margin revenue. Reality: Cloud and software revenues have not been disclosed as a material standalone segment, and the primary software entity cited during the 2021 initiative operates as a related party rather than a consolidated subsidiary. Recurring revenues stem from physical compute leasing rather than software-as-a-service, carrying a fundamentally different margin profile and capital intensity.

Risk radar. Three primary risks require explicit evaluation.

Financial leverage and cost of capital. Impact: high. Likelihood: moderate. Risk centers on refinancing mechanics rather than immediate default, highlighted by ¥6.292 billion in short-term debt maturing within twelve months as of August 2025¹¹ and a widening interest rate range on corporate borrowings. Execution timing and pricing of the pending equity placement represent the key indicators to monitor.

Hardware depreciation and obsolescence. Impact: high. Likelihood: high. Technological obsolescence remains certain in direction, though variable in timing, as successive accelerator generations reduce the power-efficiency of older fleets. Sharetronic's remanufacturing capability provides a partial structural hedge by enabling component harvesting and redeployment, offering an operational advantage over pure leasing entities, though it does not eliminate asset write-down risks.

Cross-border compliance. Impact: moderate. Likelihood: moderate. The remanufacturing model relies on importing decommissioned enterprise servers from North America into Chinese bonded zones. This supply chain operates under evolving export controls, customs classifications, and environmental regulations, where operational disruptions would directly affect Sharetronic's most differentiated business line.

Two secondary operational factors also warrant monitoring.

The first is human capital development. Sharetronic's disclosed employee turnover reached 127.2% in 2022, 138.8% in 2023, and 124.1% in 2024, before moderating to 77.3% in the first half of 2025.¹⁷ While high turnover is common across Pearl River Delta assembly operations, total research and development headcount declined during the company's operational pivot. Transitioning from white-label consumer assembly to enterprise compute infrastructure requires specialized engineering talent in power management, thermal design, and cluster networking that filings do not yet show expanding at scale.

The second involves accounting policy judgments. The valuation assigned to useful asset life and residual server value represents a central accounting estimate. When questioned regarding the impact of heavy depreciation on net profit, management responded that its depreciation policies comply with relevant accounting standards and that stable product cash flows cover short-term debt service.⁴ Because extending useful life assumptions or assuming elevated residual values flatters reported net income, any future adjustments to depreciation schedules in regulatory filings will require close examination.

Finally, the pending Hong Kong listing represents a notable structural milestone. Sharetronic filed its H-share listing application with the Hong Kong Stock Exchange on October 22, 2025, listing TF International, BOCOM International, and Guangyin International as joint sponsors.¹⁶ A dual A+H listing structure would expand international capital access while subjecting related-party transactions and revenue recognition policies to secondary regulatory oversight.

↑ Back to Top

X. Epilogue & Key KPIs to Watch [90:00 - 95:00]

Watching a stock surge 172% in half a year creates a temptation to search for a grand technological explanation. Sharetronic's actual story is both more prosaic and more instructive.

A former Foxconn section chief spent two decades learning a fundamental rule of hardware manufacturing: the winner is whoever can produce the same product for less and deliver it faster. Geng applied that discipline to hard-drive enclosures, solid-state drives, and smart cameras, discovering each time that contract assembly caps financial returns. Contract manufacturing rewards operational excellence with survival rather than outsized profits. When the AI infrastructure boom arrived, he identified a dynamic most peers overlooked—that compute expansion was primarily an asset-acquisition race rather than a technological breakthrough. His edge lay not in primary invention, but in sourcing, refurbishing, integrating, and deploying hardware at a cost structure competitors could not replicate. He then leveraged the balance sheet aggressively to execute that thesis.

Whether that strategy reflects foresight or recklessness will be determined by the utilization rates and rental pricing of a massive server fleet over the next 36 months. Both interpretations remain defensible on disclosed evidence, and any definitive conclusion today relies more on conviction than verifiable data.

What makes evaluating Sharetronic challenging is that bull and bear theses rely on the same baseline facts. Both sides acknowledge triple-digit revenue growth, extreme balance-sheet leverage, and rapid asset depreciation. They diverge on two critical variables: the long-term durability of the rental spread—an unobservable metric—and cluster utilization, an observable metric the company does not publish. This yields a focused analytical question. Investors do not need to forecast the broader AI macro cycle to evaluate the business; they simply need to monitor a concise set of disclosed metrics to determine which direction the rental spread is moving before it manifests in financial statements.

Three core indicators will determine the outcome:

One: the gross margin of the intelligent compute segment. Every other element of the investment thesis flows from this metric. The segment generated a 20.6% gross margin during its first separately disclosed half-year. If this margin holds or expands as server deployment scales, it confirms that rental rates remain firm, capacity utilization is high, and operating leverage is taking hold. If gross margin compresses while revenue grows, it indicates Sharetronic is securing market share through aggressive pricing, signaling that debt service will become a headwind before top-line deceleration becomes visible. This metric will provide the earliest indication of margin health.

Two: the asset-liability ratio evaluated alongside operating cash flow conversion. Neither metric provides clear insight in isolation. A declining leverage ratio driven strictly by equity issuance is financial engineering; rising operating cash flow paired with expanding debt reflects a continuous capital treadmill. The critical combination is a declining leverage ratio alongside operating cash flow funding an increasing share of capital expenditures—the operational marker of a business transitioning from aggressive asset acquisition to sustainable cash generation.

Three: internal deployment of Guangwei's optical modules. This offers the clearest test of whether the "compute-connect-store" framework functions as a genuine cost-reduction engine or merely a strategic marketing narrative. Guangwei was unprofitable, generating under ¥80 million in revenue prior to its acquisition. If 800G and 1.6T optical modules scale into Sharetronic's server builds and push Guangwei toward profitability, the vertical integration thesis gains empirical support. If Guangwei remains an isolated business unit, the integration narrative reduces primarily to compute leasing and storage assembly.

Secondary noise—executive accolades, daily stock fluctuations, property-market analogies, and headline server procurement figures—is downstream of these three operational variables.

The contract manufacturer once constrained by tier-one procurement terms now manages tens of billions of yuan in AI compute infrastructure. The coming three years will determine whether Sharetronic successfully broke free from the OEM margin trap—or built a larger, far more leveraged version of it.

References

1. 直击协创数据股东会！董事长耿康铭回应：算力无闲置，300多万购入设备市价已超1400万 — 网易/中金在线, 2026-07-13 ↩↩↩↩↩↩↩↩

2. 协创数据2025年营收122.36亿元 智能算力服务成营收增长核心引擎 — 新华网, 2026-03-17 ↩↩↩↩↩↩↩

3. 协创数据2026年半年度业绩预告及净利润指引 — 证券时报, 2026-07-15 ↩↩

4. 智算成核心引擎 协创数据Q1净利润同比增长超3倍 资产负债率升至近86% — 新浪财经, 2026-04-27 ↩↩↩↩↩↩↩↩↩

5. 再购110亿元服务器！协创数据去年逾200亿投入算力 但净利增速不及前两年 — 新浪财经, 2026-02-12 ↩↩↩↩↩↩

6. 安徽大佬，狂砸320亿"豪赌"算力 — 钛媒体, 2026 ↩↩↩↩↩

7. 从富士康课长到1600亿市值掌舵人，起底协创数据耿康铭的算力"赌局" — 金融界, 2026-06-30 ↩↩↩

8. 协创数据(300857.SZ)公司公告与信息披露 — 巨潮资讯网 CNINFO ↩↩↩

9. 协创数据(300857.SZ)行情与财务概览 — 东方财富网 ↩↩↩↩↩↩↩

10. 上市了！协创数据：首次公开发行股票将于7月27日在创业板上市 — 每日经济新闻, 2020-07-23 ↩↩

11. 【看新股】协创数据港股IPO：智慧存储设备龙头布局A+H 上半年偿债压力剧增 — 中国金融信息网, 2025-10-28 ↩↩↩↩↩

12. 协创数据拟收购思华信息转型软件服务商 标的两年收入8574万净利96万 — 腾讯新闻, 2021-03-18 ↩↩↩

13. 算力牛股欲"补血"扩产！协创数据拟定增80亿，年内关联交易预计金额激增至78亿 — 时代周报 / Investing.com, 2026-06 ↩↩↩

14. 21专访｜狂砸200亿买服务器 协创数据董事长直言这非常保守 — 21世纪经济报道, 2026-02-06 ↩↩↩↩↩

15. 协创数据技术股份有限公司财务信息与公告汇总 — 新浪财经 ↩

16. 122亿元"梭哈"算力，债务激增的协创数据加速冲刺H股 — 证券时报, 2025-10 ↩↩↩↩↩

17. 协创数据港股IPO：有息负债激增近90亿元 财务费用暴增2277% 研发人员及费用率皆逆势下滑 — 新浪财经, 2025-10-30 ↩↩↩↩↩

18. 80亿元定增刚过会，千亿算力巨头半年揽超15亿元净利 — 腾讯新闻, 2026-07-20 ↩↩↩

19. 协创数据：关于增资取得光为科技(广州)有限公司51%股权的公告 — 新浪财经/公司公告, 2026-04-11 ↩↩↩↩

20. 直言"这非常保守"后 协创数据再抛80亿定增 — 21世纪经济报道, 2026-06-29 ↩↩↩↩

21. 协创数据：拟使用不超70亿元自有资金进行委托理财 — 证券时报, 2026-08-04 ↩

22. 协创数据第二股东拟套现约33亿 近2年大股东套现9.8亿 — 新浪财经, 2026-04-20 ↩↩

23. 股东减持与资金反哺同步推进 协创数据算力布局获20亿元股东资金支持 — 每日经济新闻, 2026-04-21 ↩