01

Policy Trend（June.30）

South Korea Plans Approximately KRW 800tn Investment to Build Four Chip Fabs in Southwest Region, with Samsung and SK Hynix Each to Build Two

On June 29, South Korea’s Minister of Trade, Industry and Energy Kim Jung-kwan said that four chip fabs are expected to be built in the country’s southwest region, with total investment of approximately KRW 800tn, equivalent to around RMB 4.2tn. Samsung Electronics and SK Hynix each plan to build two fabs. At the same time, the South Korean government said investment in the semiconductor sector over the next 15 years is expected to reach at least KRW 30tn, covering areas including next-generation memory, edge AI, and defense applications. Under the plan, the central region will focus on advanced packaging, while the southeast region will develop semiconductor materials, components, equipment, and next-generation power semiconductor clusters. Investment in the chip packaging cluster in the Chungcheong region is expected to reach KRW 81tn.

Commentary: South Korea’s newly announced KRW 800tn chip fab investment plan represents another mega-project for the country’s semiconductor industry following the Yongin semiconductor cluster, with a scale far exceeding any previous single industrial park plan. The four new fabs in the southwest, together with the Yongin cluster, which includes four SK Hynix fabs, the central advanced packaging cluster, and the southeastern materials and equipment cluster, form an interconnected “four-cluster” industrial layout. This marks South Korea’s attempt to build, through national strategic will, a globally leading semiconductor industry cluster that covers the full value chain from materials, equipment, design, and manufacturing to packaging. The plan for Samsung Electronics and SK Hynix to each build two fabs reflects strong national support for capacity expansion by the country’s two major memory giants, while also securing strategic reserves in HBM and advanced DRAM capacity at the national level.

From a geopolitical competition perspective, South Korea’s move is a strong response to the U.S. CHIPS Act, Japan’s Rapidus initiative, and China’s semiconductor self-reliance push. As AI chips become a core bargaining chip in great-power technology competition, South Korea is seeking to reinforce its memory dominance through ultra-large-scale investment and full-chain industrial deployment. This creates dual pressure for China’s semiconductor industry. On one hand, accelerated capacity expansion by Korean manufacturers could further widen the scale and cost gap in DRAM and NAND. On the other hand, South Korea’s push into advanced packaging, materials, and equipment is extending its competitiveness upstream, directly overlapping with areas where China is seeking breakthroughs. China’s domestic semiconductor industry therefore needs to accelerate the development of irreplaceable capabilities in differentiated segments such as mature-node specialty processes, power semiconductors, and the Chiplet ecosystem. At the same time, China should leverage its large domestic AI compute demand to drive coordinated upgrading across the local semiconductor supply chain.

02

Supply-Demand Trend（July.1）

Yageo Raises Prices Across Full Capacitor Portfolio, Spot Prices Surge Nearly Tenfold as AI Servers Intensify Passive Component Supply-Demand Imbalance

Starting July 1, global passive component leader Yageo raised prices across its full capacitor product portfolio, covering tantalum capacitors, MLCCs, aluminum electrolytic capacitors, solid aluminum electrolytic capacitors, film capacitors, and supercapacitors. The affected products account for approximately 50% of Yageo’s total revenue. This round of price increases includes direct customers such as EMS and OEM clients for the first time. According to confirmation from Shanghai Securities News, factory prices have increased by approximately 50%, while spot market prices have risen even more sharply. Since May, spot prices of high-end capacitor products have surged by nearly tenfold at the highest point. Trading activity for MLCCs has recently become unusually active in Shenzhen’s Huaqiangbei electronics component market, with some quotations being updated every half hour. Delivery lead times for certain products have also extended sharply from 3–6 weeks to more than 20 weeks.

On the demand side, AI servers are the direct driver behind this round of price increases. According to estimates by China Securities, an average AI server requires approximately 28,000 MLCCs, about 13 times higher than a standard configuration. Beyond AI infrastructure, demand from humanoid robots, electric vehicles, and autonomous driving applications is also growing rapidly, continuing to absorb industry capacity. Institutions expect tight supply of high-capacity and general-purpose MLCCs to remain difficult to ease in the short term. In addition, AMD has notified graphics card partners including Sapphire and ASUS that it will raise supply prices for GPU chips bundled with GDDR memory by approximately 10%.

Commentary:Yageo’s price increase across its full capacitor portfolio, with direct-supply customers included for the first time, indicates that supply-demand conditions in the passive component industry have shifted from “structurally tight” to broadly constrained. The fact that a single AI server requires 13 times more MLCCs than a standard configuration has fundamentally reshaped the market logic for passive components. These products are no longer merely cyclical “commodity-like” components, but have become critical scarce components in the AI compute supply chain. Spot prices surging nearly tenfold within a month and delivery lead times extending to more than 20 weeks reflect the spread of panic-driven inventory buildup among downstream customers. As the global leader in tantalum capacitors, with over 40% market share, and the third-largest MLCC supplier, with over 15% market share, Yageo’s price increase carries strong industry signaling value. Japanese and Korean suppliers such as Samsung Electro-Mechanics, Murata, and Taiyo Yuden are expected to follow with price adjustments.

For China’s domestic industry, this round of price increases represents both pressure and opportunity. In the short term, rising prices of passive components represented by MLCCs will directly push up BOM costs for domestic electronics manufacturers, especially in high-consumption segments such as servers and new energy vehicles. However, from a medium- to long-term perspective, the shortage of high-end passive components creates a rare qualification window for domestic suppliers. As overseas leaders face capacity crowding and extended delivery cycles, domestic system manufacturers will have stronger incentives to adopt products from local MLCC suppliers such as Fenghua Advanced Technology and Three-Circle Group. The key question is whether domestic suppliers can seize this window to achieve substantive breakthroughs in high-end categories such as high-capacitance, high-voltage, and high-reliability products. If successful, they could move from secondary supplier status to primary supplier status and secure a stronger position in the restructuring of the global passive component supply chain.

03

Company Trend（July.2）

Infineon Completes Acquisition of ams OSRAM’s Non-Optical Analog/Mixed-Signal Sensor Business, Strengthening Leadership in Industrial and Automotive Sensors

Infineon Technologies announced that it has completed the acquisition of ams OSRAM Group’s non-optical analog/mixed-signal sensor business portfolio. The transaction was announced in February 2026 and has now received all necessary regulatory approvals. Through the acquisition, Infineon will further strengthen its leading position in industrial and automotive sensors by leveraging the highly complementary product portfolios of both companies, while also expanding its product footprint in healthcare applications. The acquired business is expected to generate approximately EUR 230mn in revenue in calendar year 2026. Following completion, the transaction is expected to be immediately accretive to earnings per share. Around 230 employees will join Infineon, and the company will add three new operational sites in Spain, Switzerland, and India.

The acquired position and temperature sensor assets will enhance Infineon’s technical capabilities in high-precision positioning, capacitive sensing, and temperature sensing. Applications include automotive chassis position sensing, hands-off steering wheel detection, robotic angle and position sensing, and blood glucose monitoring. The mixed-signal product business also adds leading healthcare imaging and sensor solutions, including products for CT and digital X-ray imaging, as well as sensor-specific ASICs. The acquired portfolio is highly aligned with Infineon’s newly established endpoint intelligence business unit, which integrates sensing, computing, connectivity, and security. The unit is focused on building integrated system-level solutions for edge applications.

Commentary: This acquisition is a precise reinforcement of Infineon’s “sensors + edge intelligence” strategy. ams OSRAM’s non-optical sensor business has established technical barriers in niche areas such as automotive position sensing and medical imaging, and forms a complete “sensing–processing–actuation” closed loop with Infineon’s established strengths in automotive power semiconductors and MCUs. As automotive E/E architecture evolves toward centralized computing and demand for high-precision sensing in industrial automation surges, sensors are being upgraded from standalone components into a critical entry point for system-level solutions. Through this acquisition, Infineon is rapidly filling its gap in non-optical sensing, aiming to secure greater control over the data entry point in the era of edge intelligence.

From an industry consolidation perspective, semiconductor M&A in 2026 shows a clear pattern of being driven by AI deployment scenarios. Unlike the horizontal scale expansion seen in previous years, current acquisitions are more focused on completing capability stacks around specific application scenarios such as automotive, industrial, and healthcare. For domestic sensor and analog chip companies, the accelerated move by international giants to build system-level solution capabilities means that business models based solely on supplying discrete components will face greater competitive pressure. Upgrading toward integrated solutions covering sensing, signal chain, and processing has become an inevitable direction. At the same time, Infineon’s statement that the acquired business will be “immediately accretive to earnings per share” also reflects a pragmatic requirement from industrial capital for profitability and integration efficiency in acquisition targets.

04

Company Trend（July.3）

Samsung Electronics Confirms 1.4nm Process to Enter Mass Production in 2029, Two Years Later Than Originally Planned

On July 1, Samsung Electronics reiterated at SAFE Forum 2026 that its 1.4nm process, SF1.4, is progressing steadily and is expected to enter mass production for major customers in 2029. The enhanced SF1.4+ process is scheduled for 2030. This marks Samsung’s latest confirmation of its revised roadmap, following last year’s decision to postpone the original 2027 mass-production target for 1.4nm to 2029. To further improve yield, Samsung has shifted toward a design-technology co-optimization (DTCO) approach. While maintaining its existing manufacturing infrastructure, the company aims to improve power efficiency, performance, and integration density through closer coordination between design and process technology. Samsung previously applied this technology to its 2nm GAA process, achieving a 26% reduction in power consumption. Strategically, Samsung is pursuing a dual-track approach: continuing R&D on 1.4nm while also allocating substantial resources to its third-generation 2nm GAA node, which is expected to enter mass production in 2027–2028.

Commentary: Samsung’s delay of its 1.4nm mass-production timeline from 2027 to 2029 reflects the physical limits and engineering challenges facing advanced process technologies as they enter the angstrom era. Samsung has attributed the delay to “higher performance targets” and “increased technical complexity.” At a deeper level, however, as GAA transistor architecture advances toward more aggressive nodes, the difficulty of materials engineering, device structure design, lithography precision, and yield control is rising exponentially. Even leading manufacturers are finding it increasingly difficult to maintain a two-year node migration cadence. At the same time, Samsung’s decision to pursue a dual-track strategy and focus significant resources on its third-generation 2nm GAA node is intended to secure competitiveness at this mainstream node in 2027–2028. This also helps prevent the company from losing short- to medium-term market share by overcommitting resources to 1.4nm, as 2nm remains the most urgent node for customers today.

For TSMC, Samsung’s delay further reinforces its leadership in advanced process technologies. TSMC’s A14, or 1.4nm-class process, is planned to enter production in 2028, one year ahead of Samsung. For China’s domestic semiconductor industry, the widening gap in advanced nodes is an undeniable reality. However, Samsung’s delay also confirms one key judgment: as process technology moves to 1.4nm and below, the cost-performance benefits of relying solely on linewidth scaling decline sharply, while “beyond Moore” approaches such as Chiplet, advanced packaging, and DTCO become increasingly important. Domestic chip companies should therefore take a more pragmatic approach by deepening capabilities in mature nodes and breaking through in specialty processes, while actively building advanced packaging and heterogeneous integration capabilities. The goal should be to offset process-level lag through system-level innovation.

05

Company Trend（June.30）

CXMT Signs Long-Term Supply Agreement Worth Over RMB 20bn with Tencent, Marking Key Step in Domestic Substitution of Server DRAM

According to Guancha.cn, ChangXin Memory Technologies (CXMT) has signed a long-term supply agreement worth over RMB 20bn with Tencent, mainly covering the supply of server DRAM memory chips over the next several years and up to five years. Against the backdrop of tight global memory chip supply and rising prices, internet and cloud computing companies are increasingly using long-term agreements (LTAs) to secure supply. Reuters described the deal as a major endorsement of CXMT by Tencent. Details of the transaction remain unclear, including whether it covers HBM products. Separate reports suggest that CXMT is also in talks with other major Chinese internet companies on similar partnerships.

Historically, the global DRAM market has long been dominated by three giants: Samsung Electronics, SK Hynix, and Micron. Chinese internet companies have remained highly dependent on imported server DRAM chips. This deepened cooperation not only validates the performance and supply stability of CXMT’s DRAM products, but also helps absorb its newly added capacity. More importantly, it helps fill China’s autonomous supply gap in high-end server DRAM and continues to advance domestic substitution in the DRAM sector.

06

Company Trend（July.1）

Longsys Builds Million-Unit Monthly mSSD Delivery Capacity as SiP Packaging Reshapes Storage Module Production Model

Longsys announced that its Suzhou packaging and testing manufacturing base has successfully built stable million-unit monthly delivery capacity for mSSD, or modular solid-state drives. The company has also reserved room for continued capacity expansion, with flexibility to potentially double output in the future. In October 2025, the first batch of Longsys mSSD samples rolled off the production line at the Suzhou base, marking a key transition from R&D to trial production. By June 2026, after production line optimization and capacity ramp-up, the company successfully achieved its million-unit monthly delivery target. Longsys has also established deep partnerships with leading customers including Lenovo and ASUS, with product adoption progressing steadily.

Technologically, Longsys’ mSSD adopts innovative SiP, or system-in-package, integration technology. It highly integrates the controller chip, NAND flash, power management IC, and passive components into a single package, enabling an “Office is Factory” model from chip wafers to end products. Looking ahead, through hardware-software co-optimization, mSSD is expected to see multi-dimensional upgrades in read/write performance, capacity utilization, cost control, and intelligent scheduling. The product is designed to support various AI endpoint devices and application scenarios, including AI PCs and AI Boxes.

07

Company Trend（July.2）

Arm China Partners with Macau International Technology Industry Center to Build “Joint Incubator” and Advance Greater Bay Area AI + Semiconductor Ecosystem

On June 30, the Macau International Technology Industry Center was officially inaugurated, with Macao SAR Chief Executive Sam Hou Fai attending the ceremony. Arm Technology (China) Co., Ltd. signed on as an ecosystem partner of the center. The two sides will jointly establish a “joint incubator” and carry out multi-dimensional cooperation in areas including ecosystem incubation, R&D innovation, and industry enablement. Arm China CEO Chen Feng delivered a keynote speech at the inauguration ceremony, saying that the company will leverage Arm’s global ecosystem and its localized innovation advantages to jointly accelerate the integration of the “AI + semiconductor” industry and ecosystem development. At the 2026 Macau InLotuX Summit held during the same period, Arm China COO Bian Lu introduced the company’s full-stack technology layout across three major areas: Edge AI, Physical AI, and Cloud AI. She also said the company is willing to deepen industry-academia-research collaboration with universities and research institutions in Macao. This cooperation marks a key step for Arm China in deepening “AI + semiconductor” industrial collaboration across the Guangdong-Hong Kong-Macao Greater Bay Area.