Company Trend（June.8）

Control battle over Nexperia: Chinese firm files RMB 8bn lawsuit, Dutch “lithography giant” sees revenue halved

In June 2026, China’s Wingtech Technology filed a lawsuit with the Intermediate People’s Court of Dongguan, Guangdong Province, against Nexperia’s Netherlands headquarters and its affiliated entities, seeking RMB 8bn in damages and requesting the restoration of control over Nexperia’s semiconductor manufacturing operations in the Netherlands. Previously, the Dutch government reportedly invoked the 1952 Goods Supply Act to forcibly take over Nexperia, cutting off information systems between its China and Netherlands operations, resulting in a supply chain disruption. Wingtech recorded a loss of RMB 8.748bn in 2025, followed by a further loss of RMB 189mn in Q1 2026. On May 6, 2026, the company was placed under a delisting risk warning (ST status).

At the same time, Dutch lithography equipment maker ASML is also facing pressure. In 2025, China accounted for 33% of ASML net sales, equivalent to approximately EUR 8.01bn, making it the company’s largest market. However, in Q1 2026, China’s share of ASML’s revenue declined to 19%, with import value falling by roughly RMB 12bn QoQ. ASML’s share price weakened, while Jefferies projected a 2% decline in the company’s 2026 full-year revenue. Although the Dutch government reportedly invested EUR 20bn to retain ASML’s strategic position, the company remains under pressure from U.S. export controls under the “MATCH Act,” including restrictions on DUV lithography equipment shipments and maintenance services to China.

Comments:Wingtech Technology has initiated domestic litigation in China under the Anti-Foreign Sanctions Law, seeking RMB 8bn in compensation, and has also launched international arbitration under the China–Netherlands Bilateral Investment Treaty, claiming USD 8bn. This marks a precedent in which a Chinese company uses domestic legal frameworks to counter foreign sanctions. The Dutch government’s invocation of the 70-year-unused Goods Supply Act to intervene in commercial operations of a private enterprise is viewed as procedurally contentious. Meanwhile, the China–Netherlands Investment Treaty does not include a “national security exception,” which may limit the Dutch government’s defense space in arbitration proceedings.

Following the disruption of operations in the Netherlands, Nexperia China reportedly achieved new chip mass production within six days, with monthly output exceeding 1 billion units, and made breakthroughs in 12-inch wafer technology. Shanghai Micro Electronics Equipment (SMEE) has reached a 90% yield rate for its 28nm lithography equipment. Domestic substitution is shifting from a reactive response to proactive expansion. However, uncertainty remains over the enforceability of the USD 8bn claim, as international arbitration typically takes 2–3 years to conclude. The decisive factor is not the courtroom outcome, but the industrial chain itself. As China transitions from the world’s largest buyer to a potential substitute supplier, the Netherlands risks losing not only orders but also pricing power. This development highlights a broader implication for countries imposing restrictions on China: the cost of technological decoupling is ultimately borne by domestic firms and taxpayers.

02

Market Trend（June.12）

From Wafer to Packaging: Shift in Semiconductor Industry Driven by AI Compute Demand Reshaping the Entire Value Chain

In 2026, the global semiconductor industry is undergoing a structural shift from wafer manufacturing toward advanced packaging, with the industry focus gradually moving along the value chain. Wafer segment: widening divergence between advanced nodes and mature processes. TSMC’s 3nm capacity remains fully utilized, while 2nm is expected to enter production in 2025. However, rising costs are pushing customers such as Apple and Nvidia toward chiplet-based architectures combined with advanced packaging solutions. Meanwhile, demand for mature nodes at 28-nanometer and above is increasing due to strong growth in automotive electronics and industrial control applications, leading to sustained capacity tightness. In China, foundry players including SMIC and Hua Hong Semiconductor are accelerating capacity expansion, with combined capital expenditure exceeding USD 10bn in 2025. Localization of mature-node production continues to increase steadily.

Packaging segment: moving from a supporting role to a core driver. As Moore’s Law slows, advanced packaging has become a key path to extending compute performance growth. TSMC’s CoWoS capacity is expanding by more than 200% in 2025. Nvidia’s Blackwell architecture adopts CoWoS-L packaging, with packaging accounting for over 30% of the cost per chip. Intel’s Foveros and Samsung’s X-Cube 3D packaging technologies are also being deployed in parallel. In China, companies such as JCET, Tongfu Microelectronics, and Huatian Technology are accelerating catch-up in 2.5D and 3D packaging. The domestic advanced packaging market size exceeded RMB 150bn in 2025.

Industry restructuring: from “manufacturing-led” to “packaging-led” value creation. AI compute demand is driving tighter integration across chip design, manufacturing, and packaging, with the Chiplet ecosystem accelerating its maturity. Packaging is increasingly no longer a back-end process, but a key determinant of performance, power efficiency, and cost.

Comments:As transistor scaling approaches physical limits, further performance gains driven solely by process node advancement are increasingly insufficient to meet the exponential growth in AI compute demand. Advanced packaging, through heterogeneous integration and chiplet interconnect, enables performance scaling without relying on the most advanced lithography equipment, creating a strategic opportunity for China’s semiconductor industry to pursue a “technology bypass” path.

Opportunities for Chinese companies are expanding. With restrictions on advanced nodes, a combined strategy of mature-node manufacturing plus advanced packaging is becoming a viable pathway. Companies such as JCET with its XDFOI platform and Tongfu Microelectronics with its VISIONS solution have achieved mass production capability in 2.5D and 3D packaging. Domestic Chiplet standards are also accelerating. However, it should be noted that high-end packaging equipment, including high-precision die bonders and temporary bonding tools, as well as key materials such as epoxy molding compounds and TSV copper pillars, still rely heavily on imports, with domestic substitution rates remaining below 20%.

Advanced packaging capacity is expanding rapidly. Global CoWoS capacity is expected to increase by approximately 150% in 2026, but potential volatility in downstream AI server demand may lead to periods of oversupply. In addition, U.S. semiconductor restrictions on China are expanding from equipment and materials into packaging technologies, with extraterritorial enforcement risks continuing to rise.

03

Policy Trend（June.12）

Indium Phosphide Export Controls Tighten: China’s Targeted Countermeasures Trigger “Material Bottleneck” Across Global AI Optical Communications Supply Chain

In June 2026, China’s indium phosphide (InP) export controls continue to intensify, causing significant disruption across the global AI semiconductor and optical communications supply chain. Indium phosphide is a critical substrate material for 800G and 1.6T high-speed optical modules and electro-absorption modulated laser (EML) chips, with no direct substitutes currently available. Since China placed indium phosphide under export licensing controls in February 2025, global second-largest supplier AXT (Beijing Tongmei) has accumulated over USD 60mn in backlog orders, while delivery lead times have extended from 2–3 months to 6–9 months.

The global indium phosphide substrate market remains highly concentrated, with Japan’s Sumitomo Electric accounting for approximately 43%, AXT for 35%, and Japan’s JX Metals for 13%, together controlling over 90% of global capacity. Following the implementation of export controls, prices for 6-inch indium phosphide wafers have surged by approximately 250%, reaching USD 5,000 per wafer, while the global supply gap for high-end substrates is estimated at 70%. Order visibility for major customers including Nvidia and Broadcom has extended into 2028. Although Lumentum has expanded capacity by four times, supply remains insufficient to meet demand. In Taiwan, suppliers such as Visual Photonics Epitaxy and LandMark Optoelectronics are also facing potential supply disruptions due to delays in export approvals from AXT.

Comments:This export control regime can be characterized as a form of “precision strike,” in which China does not impose an outright export ban, but instead leverages licensing approval timing to effectively influence the global optical communications supply chain. China accounts for approximately 70% of global indium production, while indium phosphide serves as a critical “feedstock” for AI data center optical interconnects. Without it, 800G and 1.6T optical modules cannot be manufactured at scale. This “material bottleneck” strategy may prove more impactful than traditional chip export restrictions, as downstream players may have strong fabrication capacity, but constraints at the upstream materials level can stall the entire supply chain.

However, it should be noted that domestic self-sufficiency in high-end 6-inch InP substrates remains below 5%, with gaps still existing in key equipment and process capabilities. Export controls therefore act as a double-edged sword: in the short term they may accelerate domestic substitution efforts, but if overly restrictive in the long term, they could also incentivize accelerated investment in alternative technologies abroad, such as thin-film lithium niobate. The key issue lies in whether the current window period can be used to genuinely build durable technological barriers, rather than relying solely on policy-driven advantages.

04

Company Trend（June.12）

Nvidia Vera CPU “Unblocked” for China: High-End Chip Export Breakthrough or “Special Edition” Trap?

In early 2026, Nvidia’s most advanced AI chip platform Vera Rubin entered full-scale production. However, the China market remains in a paradoxical situation, where access to high-end CPUs is possible, while top-tier GPUs remain effectively restricted. In May 2026, Nvidia CEO Jensen Huang confirmed that the company’s direct AI accelerator sales to China have dropped from 95% to 0%, with business activities in the segment fully halted.

However, Nvidia has not exited the China market. In January 2026, supply chain sources indicated that Nvidia is actively seeking approval to sell the H200 chip to China, including the GH200 Grace CPU + Hopper GPU superchip, with cumulative orders exceeding 2 million units and a unit price of approximately USD 27,000. Huang also stated during an earnings call that there is a “real possibility” that the U.S. government may allow exports of more advanced Blackwell processors to China.

At the same time, he reiterated that the United States should not supply China with its latest and most advanced flagship chips, a position Nvidia “fully supports.” This implies that even if the H200 receives approval, it would likely be a downgraded version, with reduced interconnect bandwidth and compute performance, creating a material gap versus flagship products.

Comments:Nvidia’s strategy of bundling the Grace CPU with the Hopper GPU into the GH200 superchip enables partial compliance with export regulations, as CPU components may meet regulatory requirements while GPU components remain restricted. This “split shipment” approach allows Nvidia to preserve revenue exposure in the China market while remaining within U.S. export control boundaries. However, for Chinese customers, the effective product is a constrained system with incomplete compute capability. Without full GPU performance support, the architectural advantages of the Grace ARM-based CPU cannot be fully utilized, significantly weakening overall cost-performance efficiency.

At the same time, Nvidia continues to emphasize that while high-end chip exports to China are restricted, the company should not lose access to the China market. This reflects its strategic objective of maintaining dominance over the CUDA ecosystem in China. If Chinese enterprises are forced to fully transition to domestic alternatives such as Huawei Ascend and Cambricon, Nvidia’s decade-long CUDA ecosystem moat could face structural erosion. DeepSeek-V4 has already been adapted to eight domestic chip platforms, while Cambricon’s order backlog extends to 2027, indicating that domestic substitution is accelerating through a “software-defined hardware” pathway.

05

Company Trend（June.9）

China Electronics Technology Group Reports Over 5 Million Units of Silicon-Based Gallium Nitride RF Chip Mass Production: Breakthrough in Domestic 5G Terminal “Core” and Accelerated RF Front-End Localization

China Electronics Technology Group (CETC) announced that cumulative production of its silicon-based gallium nitride (GaN-on-Si) RF chips for smart terminal applications has exceeded 5 million units, marking large-scale commercialization of domestically developed RF front-end chips in the consumer electronics sector. The product uses a silicon substrate instead of traditional and more costly silicon carbide (SiC) or sapphire substrates, and applies heterogeneous epitaxial technology to achieve the high-performance characteristics of gallium nitride, while delivering higher power efficiency, lower power consumption, and lower cost.

The RF front-end is a core module in smartphones and IoT devices and has long been dominated by U.S. companies such as Qorvo, Skyworks, and Broadcom. This development fills the gap in domestic GaN-on-Si RF chips for smart terminal applications in China. The chip is mainly used in 5G smartphone power amplifiers (PA) and Wi-Fi 6/7 RF modules. It has reportedly passed validation by leading smartphone OEMs and has been deployed in multiple mainstream 5G handset models. GaN-on-Si is regarded as a disruptive technology route, as silicon substrates cost only about one-tenth of silicon carbide and are compatible with existing 8-inch and 12-inch CMOS production lines, significantly reducing manufacturing barriers. The milestone of over 5 million units indicates that GaN-on-Si RF chips have moved from laboratory validation into commercial-scale mass production and shipment stages.

06

Company Trend（June.11）

Biwin Storage Secures USD 1.861bn Enterprise-Grade NAND Flash Procurement Contract: Domestic Memory Module Vendor Expands into Enterprise Market as Supply Chain Competition Intensifies

Biwin Storage announced the signing of a USD 1.861bn procurement contract for enterprise-grade NAND Flash components, marking the largest single procurement deal in China’s storage industry to date. The contract primarily covers NAND Flash wafers used for enterprise solid-state drive (SSD) production, targeting high-end application scenarios including data centers, cloud computing infrastructure, and AI servers.

Enterprise storage represents the highest technical barrier and most profitable segment in the memory value chain and has long been dominated by global manufacturers such as Samsung Electronics, SK Hynix, and Micron Technology. The large-scale procurement indicates that domestic memory module vendors are accelerating their transition from the consumer segment toward enterprise-grade storage markets. The company stated that revenue from its enterprise storage business increased by more than 200% YoY in 2025, driven by rapid expansion in AI infrastructure demand.

Notably, the procurement contract adopts a long-term fixed-price mechanism. Against a backdrop of high NAND Flash price volatility, this structure helps stabilize supply availability while locking in cost advantages. Biwin Storage also disclosed that its in-house enterprise SSD controller chip has entered the tape-out stage, with future plans to achieve full-stack vertical integration across NAND components, controller ICs, and module-level products.

07

Company Trend（June.12）

Empyrean Technology Advances Advanced Packaging EDA Platform: Domestic Chip Design “Last Mile” Bottleneck No Longer Dependent on Foreign Tools

Empyrean Technology, a leading domestic EDA provider, announced a major breakthrough in its advanced packaging EDA platform, filling a critical gap in China’s semiconductor design toolchain. The platform covers the full workflow of 2.5D and 3D packaging, supporting key capabilities including chiplet-based heterogeneous integration, through-silicon via (TSV) routing and placement, and thermal-electrical-mechanical multi-physics co-simulation. This marks a transition of domestic EDA in advanced packaging from point tools to a full-process platform.

Advanced packaging has become a key pathway for extending Moore’s Law, with technologies such as TSMC’s CoWoS and Intel’s Foveros highly dependent on EDA tool support. The segment has historically been dominated by Cadence, Synopsys, and Siemens EDA, with limited domestic presence. This breakthrough indicates that Chinese chip design companies may reduce reliance on foreign tools in the packaging stage, thereby lowering supply chain disruption risks.

The platform has reportedly completed process validation with leading domestic OSAT companies, including JCET Group and Tongfu Microelectronics, and supports multiple advanced packaging architectures such as CoWoS-like and EMIB-like solutions. Against the backdrop of surging AI compute demand, the deployment of domestic advanced packaging EDA tools is expected to accelerate the maturity of the Chiplet ecosystem and provide a more self-reliant design foundation for AI and HPC chips in China.