By the end of October 2026, all 76 chip design and related semiconductor companies listed on the SSE STAR Market had released their Q3 2026 financial reports. If summarized in one sentence, the overall picture is this: Building on the sustained high prosperity driven by the global AI wave and domestic supply chain autonomy in recent years, China's chip design industry has fully transitioned from an earlier phase of "high-intensity R&D investment and capacity ramp-up" to a new stage of "large-scale profit realization and high-margin harvest," jointly driven by AI infrastructure construction, advanced storage upgrades, and edge-side intelligence.
In terms of aggregate data, among the 76 sample companies, 67 achieved year-on-year growth in operating revenue for the first three quarters, accounting for nearly 90% (88.16%); 53 companies reported positive net profits attributable to shareholders, representing nearly 70% (69.74%). Notably, 16 companies successfully turned losses into profits, with several computing power and high-end chip firms that previously incurred heavy R&D costs reaching a single-quarter profitability inflection point in Q3.
Beyond the headline growth rates, the evolution of industrial logic behind this deepening prosperity warrants closer attention. Through the financial reports of these 76 companies, we can clearly observe three profound shifts occurring in China's chip industry: the core engine of profit growth has expanded from a single segment to a synergistic explosion across "computing power + storage + optical interconnects"; upstream and downstream players are establishing high certainty of prosperity through long-term agreements (LTAs) and strategic stockpiling; and the business models of leading chip design firms are accelerating their shift from pure circuit design toward "process collaboration and quasi-IDM."
I. How Robust Is the Prosperity? Nearly 90% of Companies See Revenue Growth—Which Core Segments and High-Growth Tiers Are Leading?
The most intuitive characteristic of the chip design sector on the SSE STAR Market in the Q3 2026 reports is "strong leadership by market giants and comprehensive volume expansion in core segments."
Taking CXMT (Changxin Technology), a storage giant newly included in SSE STAR Market statistics this year, as an example, its operating revenue for the first three quarters exceeded CNY 269 billion (a massive year-on-year increase of 738.51%), with net profit attributable to shareholders reaching CNY 132 billion, and Q3 single-quarter profits continuing to grow significantly quarter-on-quarter. Even excluding the outsized impact of CXMT, the remaining 75 chip design companies still demonstrated strong growth resilience—26 companies across the entire sector saw year-on-year revenue growth exceeding 50% in the first three quarters, and 11 companies doubled their revenue (growth rate >100%). From cloud AI computing power, enterprise-grade storage, and high-speed optical communications to smart wearable SoCs, automotive image sensors, and server power management chips, a batch of leading enterprises with rapidly expanding performance have emerged across various niche segments.
To more intuitively assess the true robustness of prosperity and growth magnitude across segments, we broke down the performance of the 76 companies into six core tracks (Note: In the table, "Number of Companies with Positive Growth/Profit" represents the number and coverage ratio of companies within that track achieving year-on-year positive revenue growth or positive net profit; "Head Company YoY Revenue Growth" and representative company data reflect actual year-on-year growth rates for the first three quarters of 2026):
Amid the generally positive fundamentals, breakthroughs by two types of companies hold particular significance as industry bellwethers:
The first type comprises "demand-driven" companies where main businesses and cash flows reversed simultaneously. Taking DoNAND (Dongxin Semiconductor), GalaxyCore (image sensor leader), Aojie Technology (cellular baseband chip leader), and Loongson (domestic CPU representative) as examples, these companies not only turned losses into profits on paper but also achieved comprehensive improvements in product shipment volumes, the proportion of high-end products, and operating cash flow recovery. For instance, GalaxyCore, leveraging the mass production of high-end smartphone image sensors, saw its net profit for the first three quarters increase nearly six-fold year-on-year, with net operating cash inflows exceeding CNY 1.1 billion.
The second type consists of "scale inflection" companies that crossed the break-even line in Q3 alone. Chip design is a typical R&D-front-loaded industry; once product volumes cross the critical point where R&D expenses are diluted, profits release rapidly. In Q3 this year, Moore Threads, a representative domestic GPU firm, surpassed CNY 1.7 billion in single-quarter revenue and achieved CNY 363 million in net profit attributable to shareholders, officially marking a large-scale profitability inflection point. Jiehuate, a leader in analog power management, also turned non-recurring adjusted profits positive in Q3, driven by the surge in AI server power supplies. Meanwhile, VeriSilicon Holdings, a leader in chip customization and IP, saw its backlog of orders rise to a historic high of CNY 4.057 billion by the end of Q3, laying a solid foundation for future earnings conversion.
II. Where Does Gross Margin Elasticity Come From? Why Can Storage and Optical Interconnect Chips Achieve Gross Margins Above 80% and Resonate with Computing Power Chips?
A careful comparison of profitability across segments in this year's Q3 reports reveals a fascinating industrial phenomenon: In the hardware landscape of AI infrastructure, not only do core computing power chips maintain high prosperity, but "storage and high-speed optoelectronic interconnect chips" responsible for data storage and transmission have achieved astonishing gross margins ranging from 75% to 90%.
According to Q3 report data, the gross margins of four major domestic computing power companies—Hygon Information, Cambricon, Moore Threads, and MetaX—remain stable in the high-quality range of 53%–57%. Simultaneously, CXMT's gross margin for the first three quarters approached 90%, Yuanjie Technology (optical chip leader) exceeded 83% in Q3, DoNAND reached nearly 80% in a single quarter, and Montage Technology (memory interface leader) achieved a 75% gross margin for its interconnect chips. This pattern of "storage and interconnect margins leading, computing power margins steadily high" is highly consistent with the resonance observed in the global semiconductor market this year (where Micron and SK Hynix both breached 80% gross margins, leading alongside NVIDIA and TSMC).
Why do storage and optical interconnect chips exhibit such strong profitability elasticity in this AI wave? The underlying logic can be understood from three levels:
First, the shift in AI application focus toward "long-context reasoning and Agents" has made storage and communication key determinants of system efficiency.
If AI computing chips are likened to the "brain," then high-bandwidth memory chips are the "workbench" for storing data, while high-speed optical chips and interconnect chips are the "highways" connecting these brains. As large model applications evolve from early-stage training to processing hundreds of thousands of words of long documents, video understanding, and multi-step reasoning, every word generated requires re-reading vast historical memories from the "workbench." At this stage, simply increasing the calculation speed of the "brain" is insufficient; the size of the workbench and the speed of the highways directly determine the real operational efficiency of entire ten-thousand-card computing clusters. Consequently, when cloud providers build intelligent computing centers, the demand for high-end storage and high-speed optical interconnect chips becomes extremely urgent, and price sensitivity drops significantly.
Second, storage and high-end optical chips possess strong "scale cost leverage."
The initial equipment investment for wafer manufacturing and high-end optical chip production lines is enormous, with depreciation costs relatively fixed. Once process yield crosses the maturity inflection point, combined with surging downstream demand driving up product prices, incremental revenue requires almost no additional cost, converting largely directly into gross and net profit. Meanwhile, technical barriers further amplify this profitability elasticity: Montage Technology holds international standard-setting power in DDR5 memory interfaces, and Yuanjie Technology has achieved technological breakthroughs in high-power silicon photonics light sources required for 800G/1.6T optical modules.
Third, while maintaining steady gross margins, domestic computing power chips are heavily investing capital in advanced architectures and software ecosystems.
The reason computing power chip gross margins stabilize around 55% is partly because high-end accelerator cards require expensive advanced packaging and high-bandwidth memory integration, and partly because leading domestic computing power firms are in a critical period of ecosystem building. In the first three quarters of this year, Hygon Information, Cambricon, Moore Threads, and MetaX collectively invested CNY 6.29 billion in R&D, focusing on Chiplet advanced architectures and software stack ecosystems. It is precisely this complementary pattern of "storage and interconnect releasing high elasticity, logic computing power deeply building ecosystem walls" that drives the overall leap of China's AI hardware chain.
III. What Forward-Looking Signals Exist in the Balance Sheet? How Do Upstream Long-Term Agreements and Midstream CNY 39.1 Billion Stockpiling Lock in Next Year's High Growth?
For cyclical-growth industries like semiconductors, the income statement reflects "performance over the past three quarters," while contract liabilities, inventories, and prepayments on the balance sheet hide "order and delivery confidence for the next four quarters."
This year's Q3 balance sheets convey two extremely positive forward-looking signals:
The first signal is that "long-term agreement lock-ins" by upstream storage and interconnect leaders extend into 2027.
Amid highly prosperous demand, downstream server manufacturers and major end-users are locking in long-term capacity via advance deposits. By the end of Q3, CXMT's short-term contract liabilities plus non-current liabilities representing LTA deposits totaled approximately CNY 46.754 billion in locked-in downstream prepayments, with net operating cash inflows exceeding CNY 250 billion in the first three quarters. Montage Technology's contract liabilities also surged 197.8% compared to the beginning of the year. Abundant advance receipts and operating cash flows not only validate the authenticity of downstream demand but also provide sufficient internal funding for next-generation technology R&D.
The second signal is that the four major domestic computing power leaders have jointly built a CNY 39.1 billion "capacity arsenal."
By the end of Q3, the total ending inventory of Hygon Information, Cambricon, Moore Threads, and MetaX reached CNY 25.266 billion, while prepayments to upstream wafer fabs and OSAT facilities totaled CNY 13.880 billion. Combined, these figures amount to CNY 39.146 billion.
In the semiconductor supply chain, the production cycle from wafer tape-out to completion of advanced packaging and testing often exceeds six months for high-end chips. Whoever secures scarce foundry and advanced packaging capacity in advance gains the initiative in downstream intelligent computing center bidding and deliveries. The CNY 39.1 billion in strategic stockpiling and capacity prepayments by the four computing power leaders essentially prepares ample "supplies" for large-scale deliveries in Q4 2026 and throughout 2027.
IV. How Is the Business Model Upgrading? Moving from Pure Circuit Design to "Process Collaboration and Quasi-IDM," How Are Leaders Widening Their Moats?
Beyond the prosperity dividends brought by AI, the Q3 2026 reports reflect a profound mode transformation in China's chip design industry: More excellent companies are no longer satisfied with being purely asset-light (pure Fabless) design firms but are actively extending into manufacturing, specialty processes, and advanced packaging/testing, exploring a path of dimensional upgrade toward "process collaboration and quasi-IDM."
In the past, traditional chip design companies mostly used standard process templates (public PDKs) provided by foundries to draw circuit diagrams. While offering fast starts and light assets, this approach easily led to homogeneous competition. However, in this year's Q3 reports, the companies with the fastest profit growth and most stable gross margins have almost all established some level of manufacturing or packaging/testing collaboration barriers.
Specifically, leading SSE STAR Market companies have pioneered three clear paths for model upgrading:
• Path 1: Full-Process IDM Vertical Integration (Representative Companies: CXMT, Raytron Technology, Yuanjie Technology)
In areas highly sensitive to process, such as memory, special infrared, and high-end optical chips, owning production lines means controlling yield and iteration speed. For example, Raytron Technology, a leader in special and automotive infrared, connected the entire chain from bottom-layer infrared detector chips to thermal imaging modules via self-built wafer lines, achieving over 70% net profit growth in the first three quarters to reach CNY 1.216 billion, with gross margins consistently above 50%. Yuanjie Technology also became a primary force in localized high-speed optical chip breakthroughs relying on full-process autonomous production lines from epitaxial growth to packaging and testing.
• Path 2: Integrated R&D and Packaging/Testing (Representative Company: BIWIN Storage Technology)
In AI smartphones, AI glasses, and server storage, stacking dozen-plus layers of ultra-thin memory chips into tiny spaces while ensuring high-speed heat dissipation relies on advanced packaging/testing as the core barrier. BIWIN Storage Technology formed an integrated closed loop of "main controller chip design + firmware algorithms + self-built advanced packaging/testing" by constructing high-specification wafer-level advanced packaging bases. This allows rapid response to custom needs of top-tier terminal clients, boosting net profit margin to 28.55% in the first three quarters and generating CNY 8.626 billion in net profit attributable to shareholders.
• Path 3: Fab-Lite and Deep Customization of Bottom-Layer Processes (Representative Companies: GalaxyCore, SmartSens Technology, Jiehuate)
Even without pursuing full self-built factories, leading design firms are delving into bottom-layer processes via "Fab-Lite" or "Virtual IDM" models. GalaxyCore leveraged self-built specialty process lines to achieve mass production of 50-megapixel high-end smartphone main camera sensors. SmartSens Technology deeply co-customized bottom-layer pixel processes with foundries, rapidly gaining market share in high-end smartphones and automotive cameras, achieving CNY 1.153 billion in net profit in the first three quarters. Jiehuate, a leader in analog chips, introduced its own process platform into foundries, successfully developing high-current power management ICs adapted for AI servers, driving over 50% revenue growth in the first three quarters.
Meanwhile, in the edge AI field, companies like Amlogic, Bestechnic, Espressif Systems, and Allwinner Technology have achieved dual wins in revenue and profit in smart wearables, AI glasses, and smart home markets by deeply combining advanced low-power processes with self-developed edge AI architectures.
V. Outlook for 2027: After Dense Single-Quarter Profitability Inflections, What Are the Core Growth Engines for China's Chip Design Industry Next Year?
Looking ahead from the vantage point of the Q3 2026 reports, as numerous companies demonstrate strong single-quarter sequential growth momentum and profitability inflections in Q3, China's chip design industry has officially entered a "period of large-scale profit realization and high-quality growth harvest."
Looking toward Q4 2026 and full-year 2027, four core growth engines are worth the industry's focused anticipation:
1. Accelerated Conversion of CNY 39.1 Billion Computing Power Strategic Stockpiles into Delivery Performance: As domestic telecom operators, cloud providers, and intelligent computing center projects enter dense delivery and acceptance cycles, the ample wafer and semi-finished inventory reserves accumulated by computing power leaders will continue converting into tangible operating revenue and cash flow over the next four quarters.
2. Maturing Domestic Advanced Packaging Expansion and Computing Software Ecosystems: Continuous improvement in domestic advanced packaging supply chain capabilities will further raise the capacity ceiling for local AI chips. Simultaneously, rapid progress by domestic computing power firms in software compilers, communication libraries, and large model adaptation is making Chinese computing clusters increasingly "user-friendly and easy to use," strengthening long-term repurchase stickiness among downstream customers.
3. Continuous Structural Upgrades in High-End Storage and High-Speed Optical Interconnect Products: As requirements for memory bandwidth and optical communication speeds in AI servers continue to double, the shipment proportions of enterprise-grade DDR5 memory, enterprise SSDs, high-bandwidth memory, and 800G/1.6T high-speed optical chips will keep rising, providing enduring profitability support for relevant leading companies.
4. Comprehensive Penetration of "Quasi-IDM" and Edge AI into Automotive, Industrial, and New Hardware: Chip companies with bottom-layer process collaboration capabilities will encounter broader growth spaces in waves including AI server power supplies, intelligent driving perception chips, and next-generation smart terminals like AI glasses and AI PCs.
Overall, the Q3 2026 reports of SSE STAR Market chip design companies not only delivered impressive financial results but also revealed a clear trajectory toward maturity and confidence for China's semiconductor industry: Breaking computing bottlenecks through storage and high-speed interconnect advances, fortifying the computing foundation through strategic stockpiling and software deepening, and widening technical moats through process collaboration. China's chip power is opening new value spaces amidst the global AI hard-tech wave.