Shenghong Technology dispels 'second-supplier anxiety': Securing its position as NVIDIA's primary PCB supplier through high-end yield rates and deep integration | Discovering Great Companies

Source: Times Business Research Institute Authors: Hao Wenran, Han Xun, Intern Yang Hongyang

Source | Times Business Research Institute

Authors | Hao Wenran, Han Xun, Intern Yang Hongyang

Edited by Han Xun

On July 13, Shenghong Technology (300476.SZ) issued a clarification announcement, firmly denying online rumors alleging that it had failed NVIDIA’s quality inspections and process audits. The company emphasized that its advanced manufacturing processes are mature and shipments remain normal, and for the first time disclosed that 'certain customers have already released long-term demand forecasts for 2027–2028.' Nevertheless, the company’s share price still declined significantly over the following few trading days.

How could a single rumor trigger such severe market volatility? The underlying reasons are far from simple. What the market fears is not merely an isolated quality control incident, but the escalating 'second-supplier anxiety': if NVIDIA were to introduce a second supplier due to Shenghong falling behind technologically, has the countdown already begun for the erosion of Shenghong’s 'half-the-kingdom' market share?

As the undisputed market leader holding over 50% of the global AI server PCB market share (including approximately 70% in UBB universal baseboards), Shenghong Technology reported revenue of RMB 19.292 billion in 2025 (up 79.77% year-over-year) and net profit attributable to shareholders of RMB 4.312 billion (up 273.52% year-over-year). Correspondingly, its stock price surged more than sevenfold during 2025 and reached a record high of RMB 402.60 in May 2026. However, this elevated valuation has made the market highly sensitive—and prone to overreaction—to any hint of 'multi-sourcing.' A closer look at the supply chain fundamentals and business logic reveals that the so-called 'second-supplier threat' has been significantly overstated.

The limits of process technology: Why advanced HDI poses an insurmountable barrier for latecomers?

Within the PCB industry’s technological spectrum—from through-hole boards and standard HDI to advanced HDI (six layers or more)—manufacturing complexity rises exponentially. The PCBs required for AI servers (such as the GB200/GB300 series and subsequent platforms) represent cutting-edge hardware, integrating large panel sizes, ultra-high layer counts (24+ layers), and ultra-advanced HDI (six or more sequential build-up layers with any-layer interconnect). This imposes extreme technical demands, with manufacturers universally grappling with physical limits such as interlayer alignment precision, reliability of high-aspect-ratio micro-vias, and dimensional stability control of ultra-low-loss materials during thermal expansion and contraction.

Shenghong Technology's core advantage lies precisely here. The company is among the few globally that have achieved mass production of 6-step 24-layer, 8-step 28-layer, and any-layer interconnect HDI PCBs. It also possesses the capability to manufacture high-multilayer PCBs with over 100 layers and is actively advancing R&D and certification for 10-step 30-layer HDI. Currently, enterprises worldwide capable of mass-producing 6-step 24-layer HDI boards remain extremely scarce, and Shenghong Technology’s high-end HDI and high-multilayer products are leading the market by two to three years in terms of mass-production readiness.

An even more critical industry divergence lies in yield rates. The 15%–30% yield gap between Shenghong Technology and its competitors translates directly into financial terms—and is sufficient to determine whether a business model is viable at all.

PCB manufacturing is a capital-intensive, high-depreciation industry where raw material costs account for over 60% of total costs. AI server PCBs use exceptionally expensive M7/M8-grade copper-clad laminates (CCL) and ultra-thin electronic glass fabric. In 2025, Shenghong Technology reported revenue of RMB 19.292 billion and operating costs of RMB 12.497 billion, of which raw material costs amounted to RMB 8.237 billion, representing 65.91% of total operating costs.

Below is a sensitivity analysis model based on industry yield-rate survey data, designed to illustrate the nonlinear impact of yield on profitability: assuming Shenghong Technology’s gross margin of 35.22% is achieved at an 85% yield rate, with raw material costs fixed at 65.91% of operating costs. If the yield drops to 50%, raw material consumption would need to increase by 1.7 times; considering fixed costs such as factory space and equipment within non-material costs, those costs are assumed to rise by 1.35 times. Under these assumptions, the gross margin would plummet from 35.22% to -2.39%, turning negative.

This model uses Shenghong Technology’s overall 2025 gross margin of 35.22% as the baseline profit margin at an 85% yield rate. However, this actually understates the true profitability of its high-end HDI products—the reported gross margin is a company-wide weighted average, dragged down by lower-margin legacy businesses like consumer electronics, whereas gross margins for high-end AI server PCB models significantly exceed this level. Therefore, the actual margin gap is likely even wider than the model suggests.

A 50% yield means half of the boards are scrapped outright during lamination and drilling stages, turning expensive raw materials into waste. For competitors, a 50% yield isn’t merely a question of earning less—it traps them in a financial meat grinder where every board produced incurs a loss, squeezed by both high depreciation and high scrap rates. Until they achieve a technological and yield breakthrough, secondary suppliers—even if certified—will lack the capacity to capture Shenghong Technology’s market share.

Certification Cycles and 'Asymmetric Risk'

Capital markets often equate 'the entry of a second supplier' with 'an immediate 50-50 split of market share,' overlooking the 'asymmetric risk aversion' characteristic inherent in high-end computing supply chains.

In an AI computing cluster valued at several million dollars, PCB material costs typically account for less than 3% of total hardware cost—but the signal integrity and system reliability they enable represent a core risk exposure for the entire hardware platform. Minor interlayer misalignment or blind via fractures in high-end HDI boards can trigger signal degradation or even electrical surges that cause physical damage and reputational harm. Given this trade-off—low cost sensitivity versus extremely high risk sensitivity—major customers’ primary commercial logic will inevitably prioritize stability over procurement price competition.

According to Prismark data and industry surveys, NVIDIA currently has no large-scale alternative suppliers for high-end AI server PCBs. Full certification for new suppliers typically takes 12–18 months, and in some stringent scenarios can extend to 18–24 months. Moreover, production capacity outside mainland China falls far short of NVIDIA’s scale requirements. This certification period is not merely procedural—it represents a trust barrier built through extreme surge testing, thermal stress cycling, and validation of multi-million-unit data throughput. From the H100 era (where Shenghong held roughly 40% share) to the GB200 era (with approximately 70% share of UBB boards), Shenghong Technology has already shipped millions of units, achieving deep process co-development and data-loop integration with major clients. Potential secondary suppliers, by contrast, face at least 18 months of rigorous qualification hurdles.

In the AI hardware industry, roughly 18 months equates to one product generation cycle. While secondary suppliers are still in the certification phase, Shenghong Technology has already secured approximately 50%–55% of the PCB allocation for NVIDIA’s next-generation Rubin platform (2026–2028). Its share has shown a consistent upward trajectory—from H100 to GB200 and now to Rubin—without being hindered by expectations of 'secondary supplier substitution.' As JPMorgan noted in its June 2026 research report, the ultimate determinant of allocation isn’t qualification documents, but the sustained validation of 'scalable, flawless delivery capability.'

Long-term orders anchor deep integration; gross margin validates pricing power

Against this backdrop, the more strategically significant incremental information in the July 13 announcement was Shenghong Technology’s first disclosure that 'certain customers have already released long-term demand forecasts for 2027–2028.' The real implication of this signal is that Shenghong is no longer merely an 'order-based manufacturer' but is now deeply embedded in major clients’ co-development ecosystems for next- and even next-next-generation computing architectures, forming a structurally coupled relationship that is difficult to replace in the short term.

The pricing power derived from this deep integration has been thoroughly validated under recent cost pressure tests. In the first half of 2026, leading copper-clad laminate (CCL) producer T.T. Board Holdings implemented multiple rounds of price hikes, with cumulative increases exceeding 50% year-to-date. Prices for electronic glass fabric continued to climb, with G75 electronic yarn up approximately 61% cumulatively and 7628 electronic fabric rising about 78%. For traditional cyclical consumer electronics PCB manufacturers, such cost shocks would typically exert significant pressure on profitability.

Shenghong Technology, however, delivered markedly different results: in Q1 2026, the company reported revenue of RMB 5.519 billion, up 27.99% year-over-year; net profit attributable to shareholders reached RMB 1.288 billion, up 39.95% YoY; and gross margin stood at 34.46%, an increase of 1.08 percentage points YoY and 0.95 percentage points QoQ. Excluding approximately RMB 900 million in negative impacts from foreign exchange losses and additional interest expenses due to capacity expansion during the quarter, its gross margin would have further risen to approximately 34.9%.

The counter-cyclical rise in gross margin confirms a critical thesis: Shenghong’s earnings surge is not driven by traditional electronics cycles or general price increases, but by a structural volume breakout in high-end HDI products characterized by high average selling prices and high margins. In a strong seller’s market, the initial premium pricing and yield advantages of high-end AI PCBs are sufficient to fully absorb substantial upstream raw material cost increases—this is precisely

the core logic enabling the company to navigate through cyclical volatility and consistently deliver earnings growth.

Key View: The company’s high-end HDI yield barrier and deep client integration solidify its primary supplier status; concerns over secondary supplier substitution are overstated.

Disclaimer:This report is intended solely for clients of Era Business Research Institute. The receipt of this report does not constitute the recipient as a client of our company. This report is prepared based on information that we believe to be reliable and publicly available; however, we make no representation or warranty as to the accuracy or completeness of such information. The opinions, assessments, and forecasts contained herein reflect only the views and judgments as of the date of publication. We do not guarantee that the information in this report will remain up to date. We reserve the right to modify the content of this report without prior notice, and investors should independently monitor any updates or revisions. While we strive to ensure objectivity and fairness, the views, conclusions, and recommendations expressed herein are for reference only and do not constitute an offer or solicitation to buy or sell any securities mentioned. These views and recommendations do not take into account any individual investor’s specific investment objectives, financial situation, or particular needs, and therefore do not constitute personalized investment advice at any time. Investors should carefully consider their own circumstances and fully understand and utilize the content of this report, and should not rely solely on this report when making investment decisions. Neither our company nor the authors shall bear any legal liability for any consequences arising from reliance on or use of this report. To the best of our knowledge, neither our company nor the authors have any legally prohibited interests in the securities or investment targets referred to in this report. To the extent permitted by law, our company and its affiliated entities may hold positions in the securities issued by companies mentioned in this report and may engage in transactions involving such securities, or may provide or seek to provide investment banking, financial advisory, or other financial services related thereto. Copyright in this report belongs exclusively to our company. No institution or individual may reproduce, copy, publish, cite, or redistribute this report in any form without our prior written permission. Where citation or republication is permitted with our consent, it must be used within the authorized scope, clearly credited as 'Era Business Research Institute,' and must not involve any distortion, omission, or modification inconsistent with the original intent. We reserve the right to pursue legal responsibility for any infringement. All trademarks, service marks, and logos used in this report are the property of our company.

125K Views