Shengyi Electronics: The AI Compute PCB Turnaround and China's First A-Split-A Spin-Off

I. Introduction & Episode Roadmap

On the morning of February 25, 2021, a mid-sized circuit board manufacturer from Dongguan rang the opening bell on Shanghai's STAR Market. The company sold 166.36 million new shares at RMB 12.42 apiece, raising just over RMB 2.06 billion gross and lifting its registered capital to RMB 831.8 million.¹ By the standards of a Chinese initial public offering that year, the debut was quiet — the stock rose sharply on its first day, which on the STAR Market in early 2021 was standard fare.

What made the listing historically notable had nothing to do with day-one pricing. Shengyi Electronics (生益电子) was the first company in the history of China's A-share market to complete an "A-split-A" (A拆A) spin-off — a domestically listed parent carving out a domestically listed subsidiary. The China Securities Regulatory Commission had published pilot rules permitting the structure only in December 2019.² Shengyi Electronics filed its application in May 2020, cleared the listing committee in October, secured registration approval on January 5, 2021, and listed seven weeks later.² A capital-markets mechanism that had existed on paper for fourteen months had its first working example in a printed circuit board factory.

Five years later, on February 2, 2026, those same shares appeared in a very different context. The parent company, Shengyi Technology (广东生益科技股份有限公司), completed a RMB 2 billion private placement of exchangeable bonds secured by 25.6 million Shengyi Electronics shares — representing 3.08% of the subsidiary. The bonds carried a 0.01% coupon, were oversubscribed 8.64 times, and established an initial exchange price of RMB 97.44 per share.³ Put simply, a parent that floated its subsidiary at RMB 12.42 per share was, five years later, structuring a monetization at nearly eight times that baseline, while institutional investors queued up to lend funds at virtually zero interest for the option to convert.

Between those two events sits a volatile business story. The trajectory runs through a 5G telecommunications expansion that peaked just as IPO proceeds arrived; an industry downturn that turned a business earning a 28.6% return on equity in 2019 into one posting a net loss in 2023;⁴⁵ a capital-allocation bet in Jiangxi that management eventually unwound and redirected; and an artificial intelligence compute cycle that took revenue from RMB 3.27 billion in 2023 to RMB 9.49 billion in 2025, turning a RMB 25 million net loss into RMB 1.47 billion in profit.⁵⁶

Printed circuit boards rarely attract broad investor interest, which makes this case study particularly instructive. The business lacks a consumer brand, software-like gross margins, network effects, or a traditional patent moat. Instead, it offers a clean test of three fundamental corporate questions: whether vertical integration lowers operating costs, whether high-end technical complexity commands durable pricing power, and whether management teams learn from past capital-allocation mistakes. Shengyi Electronics has run all three experiments in public view, backed by audited financial results over a single decade.

Three main themes run through this analysis, each evaluated against operational data rather than taken at face value.

The first is the parent-subsidiary dynamic. Shengyi Electronics is majority-owned by China's dominant manufacturer of copper-clad laminates — the resin, glass, and copper substrate forming the core of every circuit board. Because laminate represents the single largest raw material cost in board manufacturing, vertical ownership intuitively suggests a permanent cost and development advantage. However, the company's financial performance during 2022 and 2023 provides strong evidence against that assumption.

The second theme contrasts technical capability with commercial leverage. Fabricating a 30-layer board with buried microvias and controlled impedance for 224-gigabit-per-second data transmission is technically demanding. Whether that complexity prevents corporate buyers from qualifying secondary vendors and squeezing margins remains a separate question — one that the 2022–2023 telecommunications downturn answered clearly, and that the current AI cycle has not yet re-tested.

The third theme involves capital allocation across cyclical swings. In this sector, constructing a single production line costs more than an entire year of trough net profits, takes eighteen months to complete, and thereafter either operates at 90% capacity or generates heavy depreciation losses. Executive judgment regarding construction timing often proves more consequential than product selection alone. The historical record reveals one clear misstep, a subsequent strategic correction, and an ongoing high-stakes expansion.

A simple operational framework governs this industry's performance. Two fundamental metrics determine whether a given year yields profit or loss: plant utilization rates and average revenue per square metre of board. Layer counts, material formulas, client rosters, and supply chain positioning ultimately serve as levers to drive those two core numbers.

Evaluating corporate developments against this framework reveals their real financial impact. Adding a factory expands physical capacity. Securing a customer qualification yields no financial result until order volumes materialise. Introducing higher-tier technology raises average selling prices, but only until competitors match the capability. Disclosures from Shengyi Electronics make both utilization and pricing observable over time.

This narrative traces the company's trajectory from a 1985 joint venture in Dongguan through the 5G expansion and landmark spin-off, through the 2023 financial loss and regulatory inquiry, into the AI compute surge, concluding with an analysis of potential risks to its current growth trajectory.

↑ Back to Top

II. The Parent Company Foundation: Shengyi Technology & The CCL Substrate Empire (1985–2018)

The corporate foundation rests on permit document 粤经贸委资【1985】836号, issued by the Guangdong Provincial Economic and Trade Commission on May 22, 1985, authorizing the creation of a joint venture named Dongguan Shengyi Electronics (东莞生益电子有限公司).¹ Its initial three shareholders comprised a county-level electronics bureau, the trade development arm of the Guangdong Provincial Foreign Trade Corporation, and Hong Kong-based Fumin Development, with a registered capital of USD 3.8 million. The setup embodied the classic mid-1980s Pearl River Delta structure: Hong Kong capital and market access paired with mainland land and labor under a local state sponsor.

At its core, a printed circuit board functions as the nervous system of an electronic assembly—a laminated sandwich of insulating substrate sheets with copper wiring etched onto each layer and interconnected by micro-drilled, conductive plated holes. While a basic consumer device may use a four-layer board, telecom base stations require twelve to twenty layers, and artificial intelligence server backplanes can exceed thirty layers. Each added layer exponentially increases manufacturing complexity, requiring precise alignment of wiring patterns across all layers to micron-level tolerances while ensuring drill points penetrate accurately throughout the stack.

Two operational realities stem from this technical architecture, both central to the commercial investment case.

First, a PCB facility operates less like an assembly plant and more like a high-precision chemical processing facility. Boards pass sequentially through chemical plating baths, etching lines, lamination presses, and laser drills. Operational performance manifests primarily in yield—the percentage of panels that complete production without electrical shorts, layer misalignments, or impedance variance outside strict specifications. For a basic four-layer board, yield loss is minor. For a thirty-layer board requiring multiple build-up cycles, a single percentage point change in yield can represent the entire profit margin.

Second, the industry demands an unforgiving form of capital intensity. Production machinery is capital-intensive, highly specialized, and difficult to reconfigure. A manufacturing line built for high-layer server boards cannot be pivoted to produce smartphone boards during a compute downcycle, yet it continues to depreciate on a fixed schedule. This structural friction—volatile end-market demand paired with fixed overhead costs—drives the sharp earnings swings typical of the sector.

The insulating layers within a board consist of copper-clad laminate (CCL), which forms the core of the company's corporate lineage. Parent company Shengyi Technology, listed on the Shanghai Stock Exchange in 1998 under ticker 600183, grew to become China's largest CCL manufacturer and a leading producer globally.⁷ In high-performance applications, CCL is far from a commoditized input: resin chemistry determines high-frequency signal attenuation, thermal expansion rates, and endurance under thermal stress. Consequently, developing next-generation boards requires close, early-stage collaboration with substrate formulators.

Although Shengyi Electronics does not disclose an itemized bill-of-materials breakdown, public filings provide clear structural context. Raw materials constitute the vast majority of production costs, with laminate and prepreg serving as the primary inputs alongside copper foil and specialty chemicals. The company explicitly notes that sharp increases in material costs that cannot be offset by internal efficiencies or downstream price adjustments directly impair profitability.⁵ Conversely, during 2023, as market prices for laminate, prepreg, copper foil, and chemicals declined simultaneously, unit production costs decreased by 7.56%—demonstrating how closely board manufacturing expenses mirror substrate input prices.⁹

By 2013, the original foreign shareholder had exited, leaving Shengyi Electronics as a wholly owned subsidiary of Shengyi Technology.¹ Corporate restructuring toward an eventual public listing began in 2016. The parent company transferred minority equity stakes to four Xinyu-registered investment partnerships serving as management and employee holding vehicles, then converted the business into a joint-stock enterprise. The conversion utilized audited net assets of RMB 1.11 billion, minus a RMB 150 million dividend, as the equity base at a conversion ratio of 1:0.6257, establishing 600,829,175 shares.¹

In March 2017, municipal state-owned entity Dongguan Guohong Investment (东莞市国弘投资有限公司) invested RMB 150 million at RMB 2.321 per share, acquiring a 9.712% equity stake.¹ This finalized the pre-IPO ownership structure among the listed parent, a municipal state investor, and employee management partnerships.

This ownership structure created a widely held assumption across capital markets regarding structural synergy:

• The Assumption: As the direct fabrication arm of China's premier laminate producer, Shengyi Electronics secures raw materials at cost, gains exclusive early access to advanced substrate formulas, and maintains a permanent gross margin advantage over standalone competitors.
• The Disclosed Reality: Regulatory filings and operational results show a far more constrained commercial arrangement.

During the 3G and 4G infrastructure expansions, the company established itself primarily as a telecommunications board specialist. Its major client base included equipment vendors—Huawei (华为技术有限公司), ZTE (中兴通讯), Ericsson, Nokia, and Samsung Electronics (삼성전자)—alongside enterprise computing clients such as IBM and Inspur (浪潮).⁸ This focus enabled the company to build technical expertise in heavy-copper construction, complex backplanes, and low-loss high-layer board designs.

However, that specialization resulted in significant customer concentration. By 2019, the company's top five customers generated 74.08% of total revenue, with Huawei alone contributing RMB 1.381 billion—up from RMB 644 million in 2018.⁸ That single client relationship accounted for approximately 45% of total company sales.

Financial disclosures challenge the thesis that parent ownership guarantees superior profit margins, revealing two primary structural limits.

First, supply volume from the parent is far lower than commonly assumed. In 2025, during a year of record sales, Shengyi Electronics purchased RMB 764.18 million of materials from Shengyi Technology, representing 14.51% of its total procurement.⁵ By comparison, its largest supplier that year—an unaffiliated third party—accounted for 30.00% of procurement.⁵ While the parent remains a key partner, it is not the dominant supplier.

Second, pricing terms are strictly regulated. Domestic listing rules require related-party transactions to take place on arm's-length commercial terms, subject to audit committee review, independent board approvals, and annual public disclosure—commitments formally affirmed in governance documents by both the parent and its directors.⁵ Consequently, the parent cannot legally subsidize the subsidiary during industry downturns, preventing the relationship from acting as a structural margin floor.

The primary operational benefit of parent alignment lies in product development velocity. When cloud hyperscalers update signal-integrity specifications, having an in-house substrate developer accelerates the cycle between raw material qualification and prototype fabrication. This offers a clear advantage in turnaround speed rather than unit production cost—an edge that generates value when end markets reward technical innovation.

↑ Back to Top

III. The 5G Supercycle Boom & The Landmark STAR Market Spin-Off (2019–2021)

In 2019, operational execution aligned with market demand. China's telecommunications carriers accelerated 5G base station deployment at an unprecedented scale, creating immediate demand for the high-frequency, high-layer circuit boards Shengyi Electronics had manufactured for decades. That year, revenue reached RMB 3.096 billion and net profit rose to RMB 441.2 million, driving a weighted-average return on equity of 28.62%.⁴ For a capital-intensive contract manufacturer, a return on equity approaching 30% indicates severe supply scarcity and temporary pricing power well above production costs.

The demand spike stemmed from structural changes in 5G radio architecture. Unlike 4G base stations that relied on simplified antennas linked to separate radio units, 5G active antenna units (AAUs) integrated dozens of transmit-and-receive channels directly behind the antenna arrays. Each channel required dedicated high-frequency signal paths within a single tower-mounted unit. Fabrication demanded higher layer counts, ultra-low-loss dielectric materials engineered for millimetre-wave signals, and heavy copper layers to manage high power and thermal dissipation. A single 5G AAU consumed several times the circuit board value of its 4G predecessor, multiplying order volumes as state carriers deployed units nationwide.

While this demand surge played directly to the company's technical capabilities, it also exposed the business to systemic customer concentration. Growth relied not on a diversified customer base, but on three state-owned telecom carriers executing national infrastructure targets through a concentrated group of equipment vendors.

Recognizing the market momentum, Shengyi Technology announced plans in February 2020 to spin off its PCB subsidiary onto Shanghai's STAR Market.⁸ Regulatory authorities had published pilot rules for domestic spin-offs only fourteen months earlier, establishing strict eligibility criteria including three years of parent listing history, three consecutive profitable years, and caps on subsidiary earnings contributions.² Shengyi Electronics submitted its application on May 28, 2020, cleared the listing committee on October 16, and became the first of fourteen queued applicants to receive registration approval, on January 5, 2021.²⁸

However, the listing proceeds fell significantly short of initial projections. The original regulatory filings targeted approximately RMB 4 billion in capital.⁸ The actual offering raised RMB 2.066 billion gross, yielding net proceeds of roughly RMB 1.975 billion — half the original target.¹⁵ Management allocated these funds across four areas: the Dongcheng Phase IV high-speed, high-density board expansion in Dongguan; a Phase II multilayer board plant at Ji'an Shengyi Electronics (吉安生益电子有限公司) in Jiangxi; an R&D facility; and working capital.⁵ The broader Jiangxi manufacturing complex represented a RMB 2.5 billion commitment, with Phase I alone budgeted at RMB 1.385 billion for 700,000 square metres of annual capacity — nearly matching the company's total 2019 production footprint of roughly 800,000 square metres.⁸ In effect, management committed to doubling factory capacity at the peak of the 5G telecom cycle.

Public markets initially overlooked the cyclical risk. Based on 2020 net profit of RMB 439.2 million, the first-day closing price implied an earnings multiple far above traditional contract manufacturing benchmarks.⁴ Investors priced the subsidiary as a high-growth asset precisely as end-market demand began to plateau.

Governance of this expansion fell to a board blending state investment ties, technical expertise, and operational history. Chairman Deng Chunhua (邓春华), born in 1971, joined the leadership team from Dongguan's state sector, having led the municipal electronics industry corporation, chaired Dongguan Guohong Investment, and served as deputy general manager of the Dongguan science and technology investment group while sitting on the parent's board.⁵ He serves as the primary bridge between the company, municipal state investors, and the parent entity.

Director Chen Renxi (陈仁喜), born in 1967, serves concurrently as chairman of parent company Shengyi Technology. An organic chemistry graduate of South China University of Technology, Chen built his career in substrate manufacturing, advancing through production, operations, and executive leadership at Shengyi Technology before becoming chairman.⁵ His background reflects the group's engineering-focused orientation and its emphasis on raw material formulation.

General Manager Zhang Gongjing (张恭敬), born in 1972, directs daily operations. His career includes operational roles at EPCOS Zhuhai and Shanghai Meadville Electronics (上海美维电子) before joining Shengyi Electronics, where he managed planning, customer service, production, and served as general manager of the Ji'an manufacturing facility.⁵ His background running the Jiangxi operations directly shaped how the company subsequently addressed capacity allocation challenges at the plant.

Pre-IPO ownership structures also reflected long-term alignment strategies. During the 2016 corporate restructuring, the parent transferred minority equity stakes to four Xinyu-registered employee holding vehicles — Xinyu Chaoyi (新余超益), Xinyu Tengyi (新余腾益), Xinyu Yixin (新余益信), and Xinyu Lianyi (新余联益) — each acquiring between 3.0% and 3.4% of the enterprise.¹ Established five years prior to listing, these vehicles aligned management and employee compensation with equity value well before public capital markets provided liquidity.

Board continuity was further reinforced by Director Tang Qingnian (唐庆年), a Hong Kong resident born in 1961 who also serves as a director of Wai Wah Electronics (伟华电子有限公司) — an entity tied to the Hong Kong industrial group that participated in founding the joint venture in 1985.⁵

Despite the successful public debut, underlying financial metrics began shifting during 2020 and 2021. In 2020, revenue reached RMB 3.634 billion with net profit of RMB 439.2 million — flat compared to 2019 — while return on equity declined to 24.72%.⁴ The operational disconnect expanded dramatically in 2021. Circuit board production rose 48.67% to 1.2221 million square metres, and sales volume grew 44.67% to 1.1736 million square metres, yet annual revenue edged up by just 0.38%.⁴

This divergence signaled severe pricing pressure. While shipment volumes expanded by nearly 45%, flat top-line revenue reflected a drop of approximately 30% in average selling prices. Net profit fell 39.83% to RMB 264.3 million, and return on equity dropped to 7.48%.⁴

The initial public listing secured substantial growth capital, but deployed it at the turning point of the telecom cycle. While production volume expanded rapidly, declining average selling prices quickly squeezed profit margins, setting up the severe operational and capital allocation challenges that unfolded over the following two years.

↑ Back to Top

IV. The 5G Bust & The Ji'an Expansion Trap: Capital Allocation Falsified (2022–2023)

By early 2024, the full accounting of the downturn arrived in a formal regulatory inquiry. The Shanghai Stock Exchange's STAR Market management department issued inquiry letter 上证科创公函【2024】0070号, demanding that Shengyi Electronics quantify why it had swung to a loss, explain whether its trajectory diverged from industry peers, and disclose whether operational margins were continuing to deteriorate.⁹ On May 18, 2024, the company's auditor, 华兴会计师事务所 (Huaxing CPAs), filed a detailed response. The filing stands out as the most informative document in the company's public record, as regulatory scrutiny forced the disclosure of operational details that corporate reports typically gloss over.

Beyond the slowdown in domestic infrastructure spending, a broader geopolitical shift was taking hold. In its 2021 annual report, the company cited escalating U.S.–China trade friction as a primary headwind for China's telecommunications sector, which directly hit the circuit board suppliers feeding major domestic equipment makers.⁴ This distinction altered the nature of the industry shock. While a standard cyclical dip in base-station orders can be waited out, export restrictions that handicap a dominant customer's ability to build equipment require a structural pivot, as lost order volumes may never return in their original form.

In response, management began adjusting its product mix as early as 2021, reallocating resources toward automotive electronics and enterprise servers.⁴ Consequently, the diversification that laid the groundwork for the company's current business was not a sudden 2024 realization about artificial intelligence; it was a defensive shift initiated in 2020 and 2021 to counter external pressures on its primary telecommunications clientele.

The financial impact culminated in 2023: full-year revenue dropped 7.40% to RMB 3.273 billion, resulting in a net loss attributable to shareholders of RMB 24.99 million—a sharp reversal from the RMB 313 million net profit earned in 2022.⁹ The underlying financial mechanics reveal how the downturn unfolded.

Despite the drop in revenue, total shipment volume actually expanded, with the company selling 12.36% more surface area of circuit boards in 2023 than in the prior year. However, average selling prices plummeted by RMB 547.37 per square metre, an 18.08% decline.⁹ Lower raw material costs—including cheaper laminate, prepreg, copper foil, and chemicals—reduced unit production costs by RMB 180.26 per square metre, or 7.56%.⁹ Yet because prices fell more than twice as fast as input costs, total gross profit shrank by RMB 374.6 million.⁹ As a result, the gross margin for circuit board manufacturing contracted to 11.13%.⁹

A self-inflicted capacity surge exacerbated these pricing pressures. The Dongcheng Phase IV expansion—the primary facility funded by IPO capital—came online, introducing 258,400 square metres of annual capacity into a soft market. Although production rose 11.73%, physical capacity expanded even faster, causing factory utilization to fall by 3.85 percentage points.⁹ This lower utilization increased per-unit depreciation and energy expenses, turning fixed-cost operating leverage into a drag on earnings.

The peer comparisons requested by the Shanghai Stock Exchange provided clear counterevidence to the parent-synergy thesis. In 2023, competitors posted significantly higher gross margins: 深南电路 (Shennan Circuits) recorded 23.43%, 崇达技术 (Chongda Technology) reached 25.94%, and 沪电股份 (WUS Printed Circuit) achieved 31.17% while growing revenue by 7.23% and net profit by 11.09%.⁹ By contrast, Shengyi Electronics posted an overall gross margin of just 14.56% and swung to a net loss.⁹ WUS Printed Circuit lacked a parent laminate manufacturer, but possessed a critical commercial edge: a customer portfolio heavily aligned with enterprise networking and North American data center demand rather than Chinese telecom base stations.

These results challenge the premise that vertical integration with a laminate parent guarantees superior profitability. Upstream ownership failed to protect gross margins against an unfavorable end-market mix, allowing an unintegrated competitor with a superior product mix to outperform Shengyi Electronics by more than 1,600 basis points in gross margin. The tangible benefit of parent alignment rests not in permanent cost advantages, but in development velocity for next-generation materials—an edge in speed evaluated in subsequent sections.

Disclosures from the regulatory response indicate that operational reallocation was already underway during the downturn. Telecommunications revenue fell 26.56% to RMB 1.524 billion, dropping from 60.92% of total sales to 48.60%. Meanwhile, server-related revenue rose 25.32% to RMB 765.0 million, expanding to 24.39% of the mix, and automotive electronics sales grew 41.75% to RMB 540.0 million, representing 17.22%.⁹ Additionally, the auditor's response revealed that the company had qualified major global server clients, including Amazon, and had placed AI motherboard and accelerator card programs into mass production.⁹ Even during a loss-making year, foundational work for the subsequent compute expansion was taking shape.

The capital allocation strategy at the Ji'an facility in Jiangxi presented a more complex challenge. At the time of the IPO, management justified the expansion based on Prismark market forecasts predicting an 8.09% compound annual growth rate for automotive electronics from 2019 to 2023, outpacing the broader circuit board market.⁵ Subsequent disclosures detailed a different trajectory: actual market growth averaged just 3.98% annually from 2021 to 2024 before contracting 4.96% in 2024, while the automotive circuit board segment shrank 3.33% in 2023 and grew only 1.69% in 2024, accompanied by intense price competition due to lower technical barriers.⁵ In response, management adjusted the facility's scope in 2023, reallocating a portion of planned capacity toward enterprise servers.⁵

While project adjustments continued into 2025, the initial capital commitment reflected significant miscalculation. A multi-year plan designed for 1 million square metres of annual capacity, funded at a cyclical peak based on third-party demand projections, entered an industry-wide downturn. Regulatory filings acknowledged that maintaining the original construction plan could have led to low initial utilization rates, triggering heavy overhead and depreciation expenses.⁵ This concession mirrored the operational drag already experienced at the Dongcheng Phase IV facility.

The nature of the 2023 downturn requires careful contextualization. A net loss of RMB 24.99 million on RMB 3.27 billion in revenue represented a narrow operational deficit rather than a liquidity crisis. The company faced no going-concern doubts, and the auditor's response focused on operational margin compression rather than financial restatements.⁹ Throughout the contraction, management maintained research spending and continued client qualification programs.

The primary impact was sustained pressure on corporate returns rather than balance sheet distress. A company that achieved a return on equity near 30% in 2019 generated no positive return in 2023, even while managing an expanded asset base and elevated depreciation costs. For capital-intensive manufacturers, this dynamic illustrates a central risk: capacity added during peak demand creates fixed overhead burdens that persist through subsequent downcycles.

Management addressed the downturn through major governance adjustments. In April 2024, the board approved an equity incentive plan issuing 41,591,058 restricted shares—representing 5.00% of total share capital—to 512 employees at an initial grant price of RMB 5.01 per share, later adjusted to RMB 4.76 following a dividend distribution.⁵

The scale of the equity grant prompted shareholder debate, marked by independent director 陈文洁 (Chen Wenjie) publicly soliciting proxy votes prior to the extraordinary general meeting on May 16, 2024, where the measure passed.¹⁰ While issuing options equal to 5% of equity near an operational trough offered management upside from an impending cyclical recovery, subsequent performance conditions introduced accountability. On June 9, 2025, the board cancelled 1,083,700 granted-but-unvested shares, and the June 2026 vesting registered 5,770,059 shares to 490 participants—fewer than the 512 in the original grant.⁵¹

The 2023 net loss marked the trough of the post-IPO downcycle. The historical record depicts a company that misjudged both the timing and end-market demand for its major capital expansion, underwent formal exchange scrutiny over its margin compression, and simultaneously laid the operational groundwork for recovery by securing qualifications with tier-one compute clients.

↑ Back to Top

V. The AI Server Supercycle Pivot: High-Layer HDI, 800G/1.6T Switches, & Financial Resurgence (2024–Present)

On April 9, 2025, Deng Chunhua sat through the annual results briefing for a year in which revenue had risen 43.19% to RMB 4.687 billion and net profit had recovered to RMB 332 million.⁶¹¹ Server products had climbed to 48.96% of total sales, and 800G high-speed switch boards had reached batch delivery capability.¹¹ It was, by any reading, a strong turnaround.

What he actually said about the market is the more revealing part. Management described the plants as running at full capacity while simultaneously acknowledging that industry oversupply persisted and price competition was unavoidable.¹¹ No gross margin guidance was offered. For a chief executive presenting a 43% revenue recovery, that is a conspicuously unenthusiastic frame — and it is the kind of thing worth logging, because it establishes a baseline against which later, more confident statements can be measured.

Then the demand curve went vertical.

The mechanism is worth explaining in plain terms, because the leap from a telecom board to an AI compute board is not incremental. A conventional x86 server motherboard is a demanding but well-understood product: perhaps fourteen to eighteen layers, moderate signal speeds, high volumes, thin margins. An AI accelerator platform is a different animal. The GPU or ASIC modules sit on OAM cards, which plug into a universal baseboard (UBB) — a single very large board that must carry enormous current to power the chips while routing thousands of ultra-high-speed differential pairs between them without the signals degrading into noise. That pushes layer counts past thirty, forces the use of high-order HDI construction — where extra ultra-fine wiring layers are built up in stages on each side of the core, five or six stages deep, with laser-drilled microvias narrower than a human hair — and requires laminate materials that lose almost no signal energy at frequencies where ordinary resin behaves like a sponge.

The next process step, mSAP, etches even finer traces by plating them up rather than etching them down, which is how the industry keeps shrinking line widths as speeds climb toward 224 gigabits per second per lane.

The commercial consequence is that a single AI compute board set is worth a multiple of a general-purpose server board set, and only a handful of factories worldwide can make one at acceptable yield. Prismark's data captures the split: in 2025, boards of more than 18 layers grew 72.8% and HDI grew 26.0%, while the overall PCB market grew far more modestly, and the 2025–2030 compound growth forecasts for those two categories are 21.7% and 9.2% respectively.⁵ The industry did not have a demand boom; it had a demand boom in two specific product categories.

One forecast worth tracking rather than trusting sits inside the company's own filings. The 2025 annual report cited TrendForce expecting global AI server shipments to grow more than 20% in 2026 and to reach 17% of all server shipments.⁵ Four months later the interim report cited an expectation of more than 28% growth for the same year.¹ Third-party forecasts for this market are being revised upward in real time, which is what a genuine boom looks like — and also exactly what the late stage of one looks like. The direction of revision is information; the level is not.

Shengyi Electronics landed in both. 2025 revenue reached RMB 9.494 billion, up 102.57%. Net profit attributable to shareholders reached RMB 1.473 billion, up 343.76%, and profit excluding non-recurring items reached RMB 1.468 billion, up 348.72%.⁵ Gross margin on the main business rose 9.20 percentage points to 28.62%.⁵ Revenue from AI compute-related products grew 242% year on year, and server products passed 60% of total sales.⁵

Two details inside that result matter more than the headline. First, the geography: export revenue reached RMB 5.750 billion, up 151.20%, at a gross margin of 30.57%, while domestic revenue reached RMB 3.395 billion, up 54.54%, at 25.31%.⁵ The overseas business is now both the larger half and the better-margin half — a complete inversion of the company's historical identity as a supplier to Chinese telecom equipment makers.

Second, the concentration. In 2025 the top five customers accounted for RMB 5.964 billion, or 65.21% of sales, with the largest single customer at RMB 3.924 billion — 42.90% of the entire company.⁵ None were related parties.⁵ The company does not name them; the annual report lists them as "Customer 1" through "Customer 5."⁵ For context, at 胜宏科技 Victory Giant Technology — the closest domestic comparable in AI accelerator HDI — the top five customers accounted for 41.98% and the largest for 14.97%.¹² Shengyi Electronics' single largest account is nearly three times as concentrated as its rival's.

That is the central analytical tension of the current cycle and it deserves stating without hedging: the same customer relationship that produced a 242% growth rate in AI compute revenue also means that roughly 43 cents of every revenue renminbi depends on the purchase orders of one buyer whose identity shareholders cannot verify. The 2019 lesson — 45% of revenue from Huawei, followed by a telecom collapse — was learned at considerable cost. The 2025 balance sheet has arrived at a structurally similar place with a different customer.

The 2026 numbers so far have extended the trend and, importantly, changed its character. First quarter revenue was RMB 2.411 billion, up 52.62%, with net profit of RMB 444.7 million, up 122.16%.¹³ First half revenue reached RMB 5.784 billion, up 53.46%, with net profit of RMB 1.111 billion, up 109.36%, on a weighted return on equity of 17.95% for the six months.¹¹⁴ Gross margin for the half worked out to roughly 34.3%, against 30.4% a year earlier.

The most informative operating statistic in the half-year report is not a financial one. Output area rose 24.73% year on year while revenue rose 53.46%.¹ Roughly half the growth came from making more boards and roughly half from making more valuable boards. That is the signature of a genuine mix shift rather than a volume boom, and it is the single best piece of evidence that the company's move up the technology stack is real rather than rhetorical. It also means the earnings are more exposed to product mix than to unit demand — which cuts both ways when the mix reverses.

The intra-year sequencing is worth following, because it shows where the inflection actually happened. In the first half of 2025 revenue reached RMB 3.769 billion, up 91.00%, with net profit of RMB 531 million, up 452.11% — the profit line moving five times faster than the revenue line, which is what operating leverage looks like when a plant that was half-empty fills up.¹⁵ A year later the ratio had normalized: first half 2026 profit grew roughly twice as fast as revenue rather than five times.¹ The easy part of the recovery — reabsorbing fixed costs — was largely complete by the end of 2025. What has driven results since is mix and price, which is a harder and less durable engine.

The quarterly path within 2026 tells the same story with more precision. Second quarter revenue worked out to roughly RMB 3.37 billion against RMB 2.41 billion in the first quarter, with net profit of roughly RMB 666 million against RMB 445 million.¹³¹ Sequential acceleration of that magnitude in a capital-intensive manufacturer usually means new capacity is landing into unfilled demand rather than replacing existing volume.

Capital returns arrived alongside the recovery. The company implemented an interim distribution for the first half of 2025 following an extraordinary general meeting, and proposed a full-year 2025 cash dividend that brought total distributions to RMB 740.89 million — 50.29% of net profit attributable to shareholders and 44.11% of the parent-company distributable reserve.⁵ Paying out half of a record year while simultaneously raising equity and debt for a RMB 5 billion-plus capacity programme is a deliberate choice rather than an obvious one, and it is the kind of thing worth watching: a company genuinely short of capital for its expansion would ordinarily retain more.

The shareholder base has changed to match the story. When management held its September 2025 investor meetings, the attendee list ran to Citigroup, BlackRock, DE Shaw, Balyasny, Neuberger Berman, Manulife Investment, Redwheel, AllianzGI, Schonfeld, Polymer Capital, Dymon Asia and Pacific Alliance Group alongside domestic houses including 申万宏源 Shenwan Hongyuan, 国金证券 Sinolink Securities, 元大证券 Yuanta Securities and 招商基金 China Merchants Fund.¹⁵ Global multi-strategy and long-only capital was doing site work on a Dongguan board maker, which is itself a marker of how the AI supply chain has redrawn who owns these companies. By the end of the first half of 2026 the register carried 54,981 shareholders of record.¹⁴

What those investors were told is as revealing as who they were. The questions logged in the meeting record were almost entirely about capacity timing, 800G and next-generation 224G development status, Thailand progress and raw material prices, and the answers were correspondingly concrete on schedules and notably non-committal on margins.¹⁵ Management's disclosed position on inputs at that point was that commodity prices had risen slightly and its own raw material costs remained broadly stable — a statement that the first half of 2026 subsequently overtook, when the company described laminate prices rising through multiple rounds.¹⁵¹ The narrative has been consistent on capacity and honest about revision on cost.

A small personnel note belongs here for completeness. 唐慧芬 Tang Huifen served as both chief financial officer and board secretary through the 2025 meetings; by the 2026 interim report the board secretary function was carried by 林江 Lin Jiang.¹⁵¹⁴ Splitting a combined CFO and board-secretary role during a period of heavy capital-markets activity is unremarkable in itself, but role changes in the finance function during a large fundraising are always worth noting.

The capacity programme behind this is aggressive, and it has four distinct legs.

The first is the Dongguan intelligent computing centre project, approved by the board on December 6, 2024 at roughly RMB 1.4 billion, built inside existing Dongguan buildings — Phase I sized at 150,000 square metres a year, Phase II at 100,000.⁵¹⁵ Phase I entered trial production in the second half of 2025 and formal production in 2026; Phase II was pulled forward and began entering service during the first half of 2026.⁵¹

The second is Ji'an. On August 14, 2025 the board approved an approximately RMB 1.9 billion intelligent manufacturing high-layer compute board project, installed on floors of the existing Ji'an Phase II building, in two stages of 350,000 square metres each.⁵¹⁵ Stage one completed construction and started production at the end of the first half of 2026; stage two is under way, for 700,000 square metres a year when complete.¹ The Jiangxi asset that had been a millstone is being repurposed — this time for servers and high-layer network boards rather than automotive.

The third is the AI computing HDI base, approved November 17, 2025 at roughly RMB 2 billion, on newly purchased land in Dongguan's Dongcheng district — about 40 mu — targeting 167,200 square metres a year of five-stage-and-above HDI including mSAP product.⁵¹⁶ Construction formally began in the first half of 2026, ahead of the original schedule.¹

The fourth is Thailand, and it is the one with the longest slip. The board approved a Thai production base on July 12, 2023 at approximately USD 100 million, and raised the budget to USD 170 million on August 19, 2024.¹⁷ The site is in the TFD industrial park in Thailand, on roughly 51 rai of industrial land bought for THB 434,664,500 — not the Rayong location sometimes cited in secondary commentary.¹⁸

Groundbreaking was in November 2024; the Board of Investment granted its investment promotion certificate in March 2025; the main structure topped out by December 31, 2025; and 2026 is the year of equipment commissioning and trial production.⁵¹⁵ That is three years from board approval to first trial output, for a plant whose entire purpose is to be available when a customer asks for non-China supply. Nothing about the sequence suggests mismanagement — greenfield builds in new jurisdictions take this long — but it should temper any assumption that "China+1" capacity can be summoned on demand.

Which brings us to the Ji'an reckoning, and to the fairest test of whether this management team corrects its own mistakes. In 2025, the company formally terminated the IPO-funded Ji'an Phase II multilayer board project. Of the RMB 637.87 million of IPO proceeds committed to it, the deployed funds were returned in full to the supervised special account by June 20, 2025, and the money was reallocated to the Dongguan intelligent computing centre project.⁵ Ji'an Shengyi would continue its Phase II build with its own and self-raised funds.⁵

How should that be weighed? Three observations, and they do not all point the same way. It is a genuine correction — the company redirected capital away from a market whose growth forecast had been falsified by its own cited data, and it did so through the proper channel of a disclosed change of use rather than by quietly letting the project drift.

It came late — the automotive PCB market contracted in 2023, and the decision was made in 2025. And it is an explicit admission that the original IPO thesis for that project was wrong, which is worth remembering the next time the same team commits RMB 5.3 billion across three projects on the strength of a demand forecast. The revised claim that survives is narrow but real: this management corrects capital allocation errors, slowly, once the evidence is unambiguous. What has not been demonstrated is that it avoids them in advance.

The funding of the current expansion introduces its own tensions. On November 17, 2025 the board approved a private placement of up to RMB 2.6 billion; on March 12, 2026 the size was trimmed to RMB 2.5295 billion, with RMB 1.0 billion for the AI computing HDI base, RMB 1.1 billion for the Ji'an compute board project, and RMB 429.5 million for working capital and bank loan repayment.¹⁶

The issue cleared the Shanghai exchange's review on May 20, 2026 and remained subject to CSRC registration, for up to 124,773,176 new shares — 15% of pre-issue capital.¹⁹ Meanwhile the company is levering up: first half 2026 finance costs rose 216.52% year on year on higher borrowings and lower FX gains, and investing cash outflow reached RMB 1.258 billion in six months.¹

The parent's behaviour alongside that is the part a skeptical investor should sit with. Shengyi Technology's stake stood at 62.93% at the end of 2025.⁵ By mid-2026 it held 59.44% directly, with 25,600,000 shares — 3.06% — parked in the guarantee and trust account backing the exchangeable bond.¹⁴

The parent is not selling shares outright, and a 0.01% coupon with an initial exchange price of RMB 97.44 is a structure that only makes sense if the issuer is comfortable parting with stock at that level.³

That is not a governance failing, and the company disclosed that no directors, senior managers or 5%-plus holders subscribed.³ It is, however, a data point: the best-informed shareholder in the company chose the moment of peak enthusiasm to write a call option on 3% of it.

↑ Back to Top

VI. Core Business Economics, Industry Structure, & Competitive Landscape

Strip away the artificial intelligence narrative, and what remains is a contract manufacturer operating across three main end markets, reliant on one dominant customer, and burdened by a cost structure where roughly two-thirds of outlay goes toward bought-in materials. Analyzing how the company generates profit requires evaluating what management can—and cannot—control.

What the company cannot control is pricing at the lower end of the market. Its own framing of the 2026 industry backdrop was unusually candid: while demand for high-layer count and high-order high-density interconnect (HDI) boards remained robust, supply in those premium tiers stayed structurally tight. Meanwhile, upstream laminate prices climbed through multiple rounds, and price competition across mid- and low-end printed circuit boards intensified under sustained capacity pressure.¹ This reflects a deeply bifurcated market rather than a rising tide across all segments. Virtually all momentum in recent results stems from the top two product tiers.

The server and high-performance computing segment has become the primary operational core. It accounted for more than 60% of revenue in 2025, up from 48.96% of orders in 2024 and 24.39% of sales in 2023.⁵⁹¹¹ This segment carries the steepest technical hurdles—30-plus layer counts, five- and six-stage HDI builds, signal integrity rated for 224 gigabits per second, and thermal management for kilowatt-class accelerator modules—and, judging by export performance, commands the highest margins. Because Shengyi Electronics does not disclose standalone segment gross margins, its profit contribution must be inferred from regional performance: export revenue surged 151% at a 30.57% gross margin, compared to domestic revenue growth of 55% at a 25.31% margin.⁵ Any precise profit split for this division remains an estimate rather than an audited disclosure.

The telecom and networking business has been refocused around higher-value applications. Rather than producing lower-margin base station boards, the plant now emphasizes 800-gigabit switch boards in volume production alongside specialized optical module boards. By mid-2026, 800-gigabit optical module boards were scaling in mass production, 1.6-terabit optical module boards were in customer sampling and testing, and 1.6-terabit switch components had entered small-batch production.¹

Management has also introduced a new product category: satellite communication circuit boards completed development and reached small-batch output in 2025, with production projected to expand through 2026.⁵ This milestone warrants a clear caveat. Satellite circuit boards remain at the stage of technical qualification rather than commercial scale. Given the company's uneven historical record in converting technical qualifications into sustained revenue, satellite offerings should be viewed as an unproven option with no demonstrated cash flows to date.

Automotive electronics now serves as operational ballast rather than a growth catalyst. The company reported initial breakthroughs into autonomous driving domain controllers and millimetre-wave radar boards, securing batch orders while operating dedicated automotive lines.⁵ Yet management's rationale for unwinding the Ji'an automotive expansion—citing low technical barriers, fierce price competition, and end-market growth lagging the broader circuit board industry—remains the clearest evaluation of the segment's underlying economics.⁵

Automotive volume keeps manufacturing lines active when compute demand ebbs, but it does not drive overall corporate margins.

A structural shift in customer dynamics alters how the company's competitive moat must be evaluated. Historically, Shengyi Electronics sold to telecommunications equipment vendors such as Huawei, ZTE, and Ericsson, who designed complete systems and purchased matching boards. In the artificial intelligence compute cycle, design authority has shifted: chipmakers publish platform reference blueprints, cloud hyperscalers and server OEMs modify them, and circuit board fabricators build to specifications established before bidding opens. Securing early involvement—while layer stack-ups and material formulas remain fluid—allows a supplier to shape design parameters around its own manufacturing strengths. Management highlights this strategy, pointing to joint development projects with major domestic and global AI server providers and cloud hyperscalers.⁵¹

Whether such early collaboration creates a durable competitive advantage or simply qualifies the vendor for subsequent bidding rounds cannot be proven from disclosures alone. However, early co-development represents the principal mechanism in contract manufacturing capable of generating pricing power—a position that must be re-established with every new platform architecture.

Evaluating the company's positioning against peer financial performance provides a clear test of its competitive standing.

WUS Printed Circuit (沪电股份) remains the primary industry benchmark. In 2025, WUS generated RMB 18.945 billion in revenue, up roughly 42%, and RMB 3.822 billion in net profit, up 47.74%, yielding a circuit board gross margin of 36.91%.²⁰ In the first half of 2026, it delivered RMB 13.689 billion in revenue—a 61.17% increase—and RMB 2.923 billion in net profit, up 73.72%, with a 35.14% gross margin.²⁰ WUS operates at double Shengyi's scale and has historically anchored the industry's margin ceiling.

Shennan Circuits (深南电路) represents the scale leader with the broadest offering, including integrated circuit packaging substrates. Its 2025 revenue rose 32.05% to RMB 23.647 billion, while net profit jumped 74.47% to RMB 3.276 billion, with a 28.32% gross margin. In the first half of 2026, revenue grew 46.33% to RMB 15.296 billion, producing net profit of RMB 2.251 billion.²¹ As the largest fabricator in the peer group, Shennan's gross margin closely tracks Shengyi Electronics' overall levels, demonstrating that manufacturing volume alone does not dictate profitability.

Victory Giant Technology (胜宏科技) stands as the direct competitor in graphics processing unit accelerator HDI boards. Its 2025 results reflected a sharp up-market mix shift: revenue surged 79.77% to RMB 19.292 billion and net profit rose 273.52% to RMB 4.312 billion, yielding a 35.22% gross margin—achieved even as total physical sales volume contracted 2.72%.¹² Overseas shipments generated 76.83% of its top line.¹² Victory Giant generated over RMB 1 billion more in net profit than Shennan Circuits on roughly 18% less revenue, illustrating where industry profits currently concentrate.

Within this competitive landscape, Shengyi Electronics remains the smallest of the four by total revenue but the fastest growing. Crucially, the margin deficit that marked its 2023 downturn has narrowed substantially. In 2023, Shengyi's board gross margin lagged WUS by approximately 2,000 basis points.⁹ By 2025, that gap in core business gross margin had closed to roughly 830 basis points.⁵²⁰ In the first half of 2026, the margin spread shrank to just 80 basis points, with Shengyi posting 34.3% against WUS's 35.14%.¹²⁰

These numbers require balanced interpretation. The premise that parent ownership of substrate manufacturing guarantees superior profitability remains unsupported; Shengyi Electronics has never surpassed WUS in gross margin, including in the first half of 2026. However, the narrower proposition—that group-level substrate development accelerates entry onto next-generation high-speed platforms—is supported by recent performance. A manufacturer that lagged the market leader by 2,000 basis points during an industry trough closed that gap to near-parity within ten quarters while more than doubling top-line revenue. Disclosed data cannot isolate whether this recovery stems from substrate co-development, timely plant additions, customer concentration, or broader capacity constraints across high-layer boards. While the narrowing margin gap is clear, the underlying cause remains unproven. The forward test will be explicit: if the margin spread widens significantly as new industry capacity comes online in 2027, the substrate-synergy thesis will be disproven.

The company's forward trajectory depends on two operational outcomes.

Shengyi Electronics will maintain its momentum if demand for high-layer count and high-order HDI boards outpaces industry-wide capacity expansion, and if its four ongoing construction projects come online efficiently with high manufacturing yields.

Concrete operational evidence supports this positive case. First-half 2026 figures show total surface area output grew 24.73% while revenue surged 53.46%, confirming a structural shift toward higher-value boards rather than simple volume growth. Furthermore, the Dongcheng Phase IV facility—which created significant overhead drag in 2023—has transitioned into a major profit contributor, producing HDI, optical module, and rigid-flex boards at scale.⁵ Finally, research and development expenditures expanded 72.02% in 2025 to RMB 487.9 million, targeting next-generation switch architectures, accelerator modules, universal baseboards, optical interfaces, and AI compute platforms.⁵

Conversely, the business faces three downside risks, each grounded in corporate history.

The first risk is a recurrence of the 2021 downturn, where expanding shipment volumes coincided with steeper price declines, transforming new plants into heavy depreciation burdens. The 2023 pattern—where average selling prices dropped 18% while unit production costs fell only 8%—serves as a clear template. Across four active projects, management is adding roughly 1.2 million square metres of annual capacity, exceeding the 1.0 million square metre Ji'an plan that executive leadership previously scaled back as overbuilt.

The second risk stems from customer bargaining power. Cloud hyperscalers and server OEMs maintain qualification cycles lasting 12 to 18 months, which offers short-term supplier stickiness. However, corporate buyers facing tight supply and elevated component prices routinely qualify secondary and tertiary manufacturing sources. This dynamic played out during the 2022–2023 telecommunications slowdown, when equipment vendors broadened their supplier bases, leaving 16- to 24-layer manufacturing capabilities unable to prevent an 18% drop in average selling prices. Present disclosures offer no evidence that buyer strategies have changed; current pricing strength reflects temporary supply tightness rather than customer lock-in.

The third risk involves raw material costs. Bought-in materials comprise the largest expense category, and laminate prices experienced multiple price increases during the first half of 2026. Management has acknowledged that failure to offset rising input costs through manufacturing efficiencies or downstream price adjustments directly harms profitability.⁵¹

This cost structure creates an inherent tension across group ownership. Parent company Shengyi Technology benefits directly from higher substrate prices, even as those price increases pressure the subsidiary's operating margins. While related-party transactions are audited and conducted on commercial terms, this structural dynamic provides a financial hedge for parent shareholders rather than public investors in the subsidiary.

↑ Back to Top

VII. Playbook: Business & Investing Lessons

Strip the Shengyi Electronics story of its specifics and three transferable lessons remain — each visible in the company's operational records rather than inferred from theory.

Lesson one: in capital-intensive manufacturing, build timing matters more than build quality. The Dongcheng Phase IV facility that dragged the company into an operating loss in 2023 is the same plant described two years later as running high-density interconnect (HDI), optical module, and rigid-flex production at scale with materially higher output, value, and profit contribution.⁵

Nothing about the physical asset changed. What changed was whether market demand existed to fill its capacity. That dynamic encapsulates the mechanics of operating leverage in contract manufacturing: identical capital equipment generates a loss at low utilization and high profits at full capacity, with the swing between the two measured in hundreds of basis points of operating margin.

The reverse scenario is equally instructive. In the first half of 2026, the company reported total assets of RMB 14.08 billion — a 25.32% expansion in six months — against attributable net assets of RMB 6.51 billion.¹⁴ An asset base expanding at that pace will either be validated by sustained order volumes or become an overhead drag during the next cyclical contraction. The balance sheet accommodates both possibilities, as fixed assets impose overhead burdens regardless of market conditions.

For investors, this structural reality carries clear implications. Announcements of capacity expansion provide limited foresight into future capital returns, because ultimate profitability depends on end-market demand two to three years out — a timeline that management cannot reliably forecast. What does provide analytical value is the operational sequence: whether a fabricator expands before or after customer purchase commitments are secured. Shengyi Electronics constructed Dongcheng Phase IV and planned the Ji'an facility ahead of confirmed 5G deployment schedules, absorbing heavy margin compression when demand plateaued. The company is currently executing four major expansion projects into active compute demand. Whether that timing proves prescient or repeats the 2021 overbuild will remain unverified until the new production lines come online.

A secondary layer to this lesson involves industry-wide behavior. Expansion decisions across the sector tend to cluster, as competing fabricators respond to identical demand signals from major technology clients. The 2023 downturn resulted not from Shengyi Electronics expanding in isolation, but from broad-based industry overcapacity. Regulatory filings revealed that competitors including Shennan Circuits and Chongda Technology similarly experienced revenue contractions and margin erosion.⁹ Consequently, corporate capital discipline by a single firm provides only partial protection against collective industry overexpansion.

By contrast, WUS Printed Circuit maintained higher profitability during that downcycle not through capacity restraint, but through strategic alignment with enterprise networking and data center end markets rather than domestic telecom infrastructure.

Lesson two: technical capability secures market qualification, not pricing power. The 2022–2023 telecommunications contraction served as a clear test of technical moats. Shengyi Electronics possessed fabrication capabilities for 16- to 24-layer boards that few global factories could match, yet those technical specs did not prevent an 18% single-year decline in average selling prices.⁹ When multiple qualified suppliers possess a given manufacturing capability, corporate buyers retain the leverage to negotiate prices downward. Technical difficulty alone does not equal supply scarcity, and only scarcity yields pricing power.

The corresponding dynamic in the current artificial intelligence cycle demonstrates that peak margins occur in the window between when an advanced technology becomes essential and when secondary suppliers achieve mass-production yields. In 2025, that gap widened substantially, as global demand for circuit boards exceeding 18 layers grew 72.8% while high-tier fabrication capacity remained constrained.⁵ Major industry participants are now building out production lines to capture that premium. Historically, operational windows of this nature compress from the supply side on timelines dictated by factory construction and equipment qualification schedules.

Beyond corporate strategy, the historical record highlights an important analytical resource. The most detailed operational disclosures from Shengyi Electronics were prompted by regulatory oversight. The Shanghai Stock Exchange's 2023 inquiry letter required the company to disclose granular metrics — including unit selling prices, per-metre production costs, incremental capacity releases, utilization fluctuations, and line-by-line peer comparisons — that standard financial summaries typically omit.⁹ Regulatory inquiry responses of this type are published across Chinese equity markets and provide critical empirical data for evaluating margin dynamics.

Lesson three: a domestic spin-off restructures corporate governance without altering fundamental supply-chain economics. China's A拆A spin-off framework achieved its intended corporate finance objectives: creating a dedicated equity currency that enabled the company to issue a 5% equity grant to 512 key employees, establishing an independent management board focused exclusively on circuit board fabrication, and securing direct access to capital markets, highlighted by a placement targeting up to RMB 2.5295 billion.⁵¹⁶

However, public listing did not insulate the subsidiary from broader industry cycles or provide structural cost relief through its parent alignment. Related-party procurement between Shengyi Electronics and Shengyi Technology remains a strictly regulated, arm's-length transaction requiring formal board approvals and independent oversight. These governance rules protect minority shareholders by ensuring fair market pricing, but they equally prevent the parent company from absorbing the subsidiary's operational losses during industry downturns.

The parent company's capital allocation actions offer additional context. Parent entity Shengyi Technology's issuance of exchangeable bonds secured by a 3.08% equity stake in the subsidiary demonstrates how controlling shareholders can monetize equity value at cyclical peaks while retaining operational control. Investors evaluating a listed subsidiary must monitor both the subsidiary's operational execution and the parent entity's portfolio management, as each reflects a distinct perspective on asset valuation.

↑ Back to Top

VIII. Analysis & Bear vs. Bull Case (7 Powers, Risk Radar, & Falsification Tests)

Running Shengyi Electronics through Hamilton Helmer's seven sources of durable advantage is a useful discipline precisely because most of the boxes come back empty, and the ones that do not are conditional.

Scale economies: present but bounded. A high-end HDI line is a genuine fixed-cost hurdle — the AI computing HDI base alone carries a total project investment of RMB 2.032 billion for 167,200 square metres of annual capacity, and the Ji'an compute board project RMB 1.937 billion for 700,000 square metres.¹⁶ Chemistry baths, laser drills, automated optical inspection and scrap recovery all reward throughput. But the scale that matters is per-technology-node, not company-wide, and Shengyi Electronics is the smallest of the four major listed Chinese players by revenue. Shennan Circuits is roughly two and a half times its size and earns a near-identical gross margin, which suggests aggregate scale is not the binding variable.²¹⁵

Process power: the strongest of the seven, and the hardest to verify from outside. Registration accuracy across thirty-plus layers, blind and buried microvia formation, impedance control at 224G, and yield on five- and six-stage HDI are accumulated organizational knowledge that cannot be bought as equipment.

The affirmative evidence is real: mass production of OAM and UBB boards, batch production of 800G optical module boards, small-batch production of 1.6T switch products, and an R&D programme that rose 72.02% in spend during 2025.⁵¹

The limitation is that yields are not disclosed, so an outside investor cannot distinguish process power from the industry's general shortage of advanced capacity. That distinction will be settled by what happens to margins when the shortage ends.

Switching costs: real but asymmetric. Server OEM and hyperscaler qualification runs roughly twelve to eighteen months, which imposes a genuine cost on a buyer who wants to move a board. It does not impose a cost on a buyer who adds a second source at the next platform revision — which is the actual competitive threat, and the one that materialized in telecom.

Cornered resource, counter-positioning, network effects, branding: absent. Contract PCB manufacturing has no meaningful brand equity with end consumers, no network dynamics, no business model that incumbents cannot copy, and no exclusive input. The parent relationship is the closest thing to a cornered resource, and Section II established why it is not one: the parent supplies 14.51% of procurement on arm's-length terms.⁵

Porter's five forces produces a similarly sober picture. Supplier power is meaningful and currently rising, with laminate prices up through multiple rounds in the first half of 2026 and materials dominating the cost structure.¹

Buyer power is high and concentrated — one customer at 42.90% of 2025 revenue, five at 65.21%, none named, against a domestic rival whose largest account is 14.97%.⁵¹² Rivalry is intense at the low end and currently muted at the high end, but every credible rival is expanding high-end capacity simultaneously.

Substitutes are limited in the near term, since there is no alternative to a physical interconnect substrate, though advanced packaging that pulls interconnect from the board into the package is a genuine long-run share threat at the highest densities. Barriers to entry are high in capital and higher in process knowledge, but they are barriers against newcomers, not against the three well-capitalized incumbents already inside.

The current risk radar has three items that matter and several that do not.

AI capex digestion is the dominant risk, and the company says so itself. The 2025 annual report's own performance-decline risk factor states that if investment in AI data centre infrastructure slows, growth in demand for AI server and high-speed network PCBs could decelerate or fall, and — combined with macro volatility, deteriorating trade conditions, intensified competition, customer and product development shortfalls, or raw material price swings — could slow the company's growth or reverse it.⁵ That is management's own language, in a filing, in a year of record profit. It belongs directly alongside any claim that the AI mix shift is structural.

The historical base rate is unhelpful to the bulls: the 5G cycle produced a volume boom that peaked in 2020, a price collapse in 2021, and a loss by 2023. Infrastructure build-outs digest.

Customer concentration is the second, and it is worse than it looks. The 42.90% single-account exposure is not merely a revenue risk; it is a pricing risk, because a buyer of that size faces almost no cost in qualifying a second vendor and considerable incentive to do so. The company flags the risk of dependence on a single customer, a single field or overseas markets in its own operating risk disclosure.⁵ The forward test is whether revenue outside the top two compute accounts grows as a share of the total, and whether the top-five figure disclosed each April declines from 65.21%.

Geopolitical and supply-chain exposure is the third. Export revenue reached 63% of the 2025 main business and grew 151%, which makes the company structurally exposed to any restriction on advanced AI accelerator shipments into or out of Asian assembly chains, and to the tariff and origin rules governing where a board must be made.⁵ The Thailand plant is the mitigation, and it will not be in commercial production until 2026 at the earliest.⁵ Currency is a related second-order exposure; the first half of 2026 saw finance costs rise substantially in part on lower FX gains.¹

Now the falsification tests, stated as claims with verdicts rather than as a balanced list.

Claim: parent CCL integration guarantees superior gross margins. Verdict: rejected, and replaced with a narrower claim that is now supported but not attributed. The 2023 record settles the original claim — an 11.13% PCB gross margin against 31.17% at WUS in the same year, with the parent relationship fully in place.⁹

What has since happened is that the gap closed to roughly 80 basis points by the first half of 2026.¹²⁰ The surviving claim is therefore about speed to new platforms, not cost, and its attribution to the parent is unproven.

The forward test: if the gross margin gap to WUS re-widens beyond roughly 500 basis points once industry high-end capacity catches up in 2027, the material-access explanation fails and the closure was simply a scarcity rent shared by everyone with a qualified line.

Claim: high-layer HDI capability is a structural moat. Verdict: rejected as stated, narrowed to a cyclical scarcity rent. The telecom collapse is the direct counterexample, run on the same company, with the same management, in the same decade. There is no evidence yet that AI compute buyers behave differently from telecom buyers once supply loosens.

The forward test is price realization: watch whether revenue per square metre of output continues to rise. In the first half of 2026 output area grew 24.73% against revenue growth of 53.46%, meaning realized value per unit rose sharply.¹ The moment that spread inverts — area growing faster than revenue, as it did in 2021 — is the moment the cycle has turned, and it is observable a year before it reaches the profit line.

Claim: the Ji'an expansion demonstrated capital discipline. Verdict: rejected on the original decision, partially rehabilitated on the correction. The project was funded at a cycle peak on an automotive forecast that the company's own subsequent filing shows was wrong by more than half, and it was terminated as an IPO project with proceeds returned to the supervised account in June 2025 and redirected to Dongguan.⁵

The rehabilitation is that the correction was made properly and the Jiangxi asset is now being repurposed for compute.

The forward test is not a promise but a number: whether the return on the roughly RMB 5.3 billion committed across the AI computing HDI base, the Ji'an compute board project and the Thailand plant clears the company's cost of capital across a full cycle, and specifically whether those plants hold utilization when demand next pauses.

Claim: AI compute demand will permanently insulate earnings. Verdict: rejected outright, by management's own risk disclosure and by the company's own history. No cyclical hardware manufacturer selling into an infrastructure build-out has permanent earnings insulation, and this one has a documented precedent of exactly the opposite within the past five years.

Finally, the activist stress test — what a skeptical investor would put to the board. First, disclosure: the company discloses no segment gross margins and no customer names, which means shareholders cannot independently assess where profit is earned or how exposed they are to a single buyer.

Second, dilution: a 5% restricted stock plan struck near a cyclical trough, followed by a placement of up to 15% of pre-issue capital, is meaningful cumulative dilution for existing holders, even if both are individually defensible. Third, leverage and cash conversion: operating cash flow of RMB 917 million in the first half of 2026 against investing outflows of RMB 1.258 billion and a 216.52% rise in finance costs means the expansion is being funded partly with debt into a cyclical peak.¹

Fourth, working capital: receivables stood at RMB 2.831 billion at the end of 2025, 53.41% of current assets and 25.19% of total assets, with inventory a further RMB 1.526 billion — a concentrated credit exposure to a customer base the company will not name.⁵

Fifth, and most pointedly, the related-party question is not whether pricing is fair — it is governed and disclosed — but whether a subsidiary buying an input from its own controlling shareholder during a period of rising input prices has the same negotiating latitude as a wholly independent buyer. That question does not have a public answer.

↑ Back to Top

IX. Epilogue & Key KPIs to Watch

The company's corporate arc is remarkably clear. A 1985 joint venture with USD 3.8 million of capital in a Dongguan county town grew into a telecommunications board specialist, rode a national 5G build-out to a 28.62% return on equity, served as the pioneer for China's domestic spin-off framework, raised capital at the peak of its market cycle, spent those proceeds into an industry collapse, posted a net loss, faced regulatory inquiries from the stock exchange, quietly qualified tier-one North American compute clients during its worst financial year, and then — as the artificial intelligence boom took hold — more than doubled its top line in a single year and distributed half its earnings to shareholders.⁵

The historical record does not support a narrative where vertical integration and raw technical excellence engineered this recovery. Operational disclosures from 2023 demonstrate that neither dynamic prevented an operational loss.

A more defensible assessment is that Shengyi Electronics remained a capable, capital-intensive fabricator operating with suitable technology in an oversupplied end market, until the artificial intelligence compute expansion shifted industry demand directly toward its pre-existing capabilities. Management's contribution lay in qualifying compute customers early, redirecting capital once initial expansion assumptions proved invalid, and scaling production rapidly when demand materialized. These represent solid operational skills rather than a permanent competitive moat.

Shengyi Electronics currently operates at maximum operating leverage in both directions. Four expansion projects are scheduled to come online between 2026 and 2027. A private placement targeting up to RMB 2.5295 billion has cleared exchange review. The Thai facility remains roughly a year away from commercial output. A single customer accounts for more than 42% of revenue. Meanwhile, the controlling parent has structured exchangeable bonds backed by 3.08% of the subsidiary's shares at an initial conversion price of RMB 97.44. Each of these elements magnifies the impact of any shift in AI infrastructure capex.

A central operational asymmetry governs this structure. The company's cost base is fixed and expanding — marked by four production facilities building or ramping capacity, rising depreciation schedules, and overseas workforce onboarding — while its top line remains concentrated, highly cyclical, and subject to annual pricing negotiations. That structural tension produced the highest profits in corporate history in 2025 and an operating loss in 2023 using essentially the same physical asset base and executive team.

Neither extreme reflected a fundamental change in underlying business quality — a critical baseline when evaluating future financial projections.

Three operational metrics will determine the company's trajectory.

First, the spread between production volume growth and revenue growth. Disclosed in annual and interim filings, this metric serves as the clearest indicator of average selling price and product mix shifts, leading reported gross margins by several quarters. In 2021, when circuit board shipment area expanded by 44.67% while total revenue rose just 0.38%, average selling prices were collapsing a full year before a net loss appeared.⁴ Conversely, in the first half of 2026, when output area grew 24.73% while revenue surged 53.46%, high-value product mix was driving performance.¹ Investors should track the direction of this growth spread far more closely than headline revenue.

Second, the gross margin gap relative to WUS Printed Circuit (沪电股份). Rather than tracking absolute gross margin — which fluctuates with broader industry supply and demand — analysts should watch the relative margin spread. This differential indicates whether Shengyi Electronics is closing the operational gap with the industry benchmark or merely benefiting from temporary high-layer board shortages. The margin gap stood at roughly 2,000 basis points during the 2023 trough and narrowed to approximately 80 basis points in the first half of 2026.⁹¹²⁰ Where this spread settles after industry-wide high-end capacity comes online in 2027 will test the validity of the competitive moat.

Third, annual customer concentration metrics. Disclosures detailing top-five customer revenue shares each April offer the only public view into whether management is diversifying beyond single-buyer dependencies. A declining revenue share for the largest client alongside top-line expansion would signal a broader, more resilient franchise. A rising concentration metric during an industry boom would confirm that the structural customer risk of 2019 has simply re-emerged under a new buyer.

References

1. Shengyi Electronics Co., Ltd. 2026 Half-Year Report (生益电子股份有限公司2026年半年度报告) — Shanghai Stock Exchange filing, 2026-08-15 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

2. Shengyi Electronics set to become the first successful "A-split-A" case as spin-offs crowd the STAR Market — 21st Century Business Herald, 2021-01-22 ↩↩↩↩

3. Announcement on completion of the controlling shareholder's private placement of exchangeable bonds (公告编号：2026-005) — Shengyi Electronics via CNInfo, 2026-02-03 ↩↩↩

4. Shengyi Electronics Co., Ltd. 2021 Annual Report Summary (2021年年度报告摘要) — Shanghai Stock Exchange filing, 2022-03-25 ↩↩↩↩↩↩↩↩↩

5. Shengyi Electronics Co., Ltd. 2025 Annual Report (生益电子股份有限公司2025年年度报告) — CNInfo, 2026-04-24 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

6. Shengyi Electronics Co., Ltd. 2024 Annual Report (生益电子股份有限公司2024年年度报告) — Shanghai Stock Exchange filing, 2025-03-27 ↩↩

7. Shengyi Technology Co., Ltd. official corporate website — SYST ↩

8. PCB maker Shengyi Electronics formally files for the STAR Market; Ji'an RMB 2.5 billion project phase one expected to start in the second half — Cailian Press, 2020-05-28 ↩↩↩↩↩↩

9. Huaxing Certified Public Accountants reply to the Shanghai Stock Exchange inquiry letter on Shengyi Electronics' 2023 annual report (华兴专字[2024]23011560079号) — CNInfo, 2024-05-18 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

10. Announcement on granting the reserved portion of the 2024 Restricted Stock Incentive Plan (公告编号：2025-021) — Shengyi Electronics via Shanghai Stock Exchange filing, 2025-04-18 ↩

11. Shengyi Electronics: server sales share rises to nearly 50%; intelligent computing centre high-layer HDI project targeted for Q4 trial production — Cailian Press, 2025-04-09 ↩↩↩↩

12. Victory Giant Technology's bumper harvest — Economic Observer, 2026-03-13 ↩↩↩↩

13. Shengyi Electronics Co., Ltd. 2026 First Quarter Report (2026年第一季度报告) — CNInfo, 2026-04-28 ↩↩

14. Shengyi Electronics Co., Ltd. 2026 Half-Year Report Summary (2026年半年度报告摘要) — CNInfo, 2026-08-15 ↩↩↩↩↩

15. Shengyi Electronics Investor Relations Activity Record No. 2025-004 (投资者关系活动记录表) — Shanghai Stock Exchange filing, 2025-09-12 ↩↩↩↩↩↩↩↩

16. Announcement on adjusting the 2025 plan for issuing A shares to specific investors (公告编号：2026-013) — Shengyi Electronics via CNInfo, 2026-03-13 ↩↩↩↩

17. Announcement on increasing the investment quota for the Thailand subsidiary (公告编号：2024-044) — Shengyi Electronics via Shanghai Stock Exchange filing, 2024-08-21 ↩

18. Shengyi Electronics announcement on progress of the new production base investment in Thailand — Sina Finance, 2023-12-19 ↩

19. Shengyi Electronics private placement clears Shanghai Stock Exchange review; plans to raise up to RMB 2.53 billion — Eastmoney, 2026-05-21 ↩

20. WUS Printed Circuit (002463) earnings profile and financial summary — Tonghuashun (10jqka) F10, 2026-08-29 ↩↩↩↩↩↩

21. Shennan Circuits (002916) earnings profile and financial summary — Tonghuashun (10jqka) F10, 2026-08-29 ↩↩