Weekly News | Supply chain trends in Semiconductor industry #172

172nd of weekly news

01

On June 23, Samsung officially released a new generation of flagship chip-Exynos 2500. This chip, based on advanced 3nm gate-all-round (GAA) processing technology, presents better power efficiency and improved thermal performance in a fan-out wafer-level package (FOWLP), significantly reducing chip thickness. The CPU, based on 1+7+2 architecture, consists of 1×3.3GHz ultra-large core, 2×2.74GHz large cores, 5×2.36GHz large cores, and 2×1.8GHz small cores, and its large core performance is up 15% over the last generation of chip. The GPU is a customized Xclipse 950 based on AMD RDNA3 architecture. With an upgraded dual shader engine architecture, this should deliver 28% higher FPS in ray tracing.       

In addition, it is also equipped with AI engine powered by 24K MAC NPU and DSP. The NPU features 59 trillion operations per second (TOPS), and the performance of AI increases by 39% compared to the previous generation of chip. It supports cameras with up to 320 million pixels and allows 8K video recording and playback at 60fps and 30fps. In terms of communications, the 5G NR modem of the Exynos 2500 supports the latest 3GPP Rel.17 standard and non-terrestrial network (NTN), allowing connections via satellites at cellular network blind spot. The chip will finally debut in Samsung’s Galaxy Z Flip 7. 

From a market perspective, the Exynos 2500, with its advanced 3nm GAA process and excellent performance, is expected to compete with the products of Qualcomm, Apple and other competitors, and will win a largershare for Samsung in the high-end chip market. The powerful AI performance and image processing capabilities of the chip will meet the growing consumer demand for taking photos, recording videos and AI applications using smart phones, driving further upgrades in smartphone functionality and user experience. 

From the perspective of technology development, the multiple innovated technologies of the Exynos 2500 not only improve its performance and energy efficiency but also provides new ideas and directions for future chip design. In addition, its support for 3GPP Rel. 17 standard and NTN reflects Samsung's forward-looking layout in the field of communications technology and laid a foundation for the future development of 5G and even 6G communications technology.

02

On June 25, ASE Technology Holding (ASE) announced at the shareholders' meeting that it is actively planning to set up a chip testing facility in the USA to meet the needs of AI-chip manufacturer Nvidia (NVIDIA). ASE chief operating officer Tien Wu said at the company’s shareholders’ meeting that the move is to comply with American policy and meet customers’ demand and Taiwan’s industrial chain has to expand its capacity in the USA. ASE will take a global deployment as its business strategy and work closely with wafer foundries and supply chain partners to expand its capacity to respond to market changes. He stressed that the industry was still confident in finding a way out despite the changes in the external environment. Wu also mentioned that ASE will actively advance the planning of the US testing fab according to customers’ demand and refer to TSMC's success story in the USA. To enter the US market requires consideration for a number of factors, including customer support and staffing. In addition, ASE has opened a second facility in San Jose, California, to expand its testing service capabilities.

TASE’s testing facility in the USA is an important layout of the semiconductor industry driven by the international situation and market demand. From the perspective of supply and demand in the industry, this move helps alleviate the urgent need for advanced semiconductor testing services in the USA, especially in the field of AI chips. With the rapid development of large facilities of tech giants such as Nvidia, the expansion of testing capacity is becoming particularly critical. This move of ASE not only makes it easier and more efficient to serve local customers but also helps to stabilize the supply chain to a certain extent and reduce supply risks due to geographical constraints. This layout of ASE further highlights the decentralization trend of the global semiconductor industry chain, which also reminds different countries and regions of paying more attention to the integrity and safety of the industry chain when formulating industrial policies, to cope with the increasingly complex international competition environment.

03

According to report by Korean media on June 24, Samsung Electronics is accelerating 2nm process layout and will bring mass production to its Taylor Foundry fab in Texas in 2026. Samsung internally discussed the introduction of 2nm production line devices from January to February 2026. The Taylor facility originally planned to adopt 4nm process but now has been prioritized for development 2nm process. The project was postponed several times due to immature technology, but the clean room renovation restarted in Q2 2025 and is expected to be completed by the end of the year to be ready for installing equipment.   

In order to ensure the mass production of 2nm chips by the end of 2026, Samsung postponed the construction of 1.4nm pilot line, originally scheduled to start at P2 plant in Pyeongtaek in Q2 2025, to the end of 2025 or early in 2026, resulting in a possible delay of the mass production of 1.4nm chips to 2028. In terms of customers, Samsung's 2nm process has won orders from a domestic company and a Japanese AI chip design company, but it has not yet received orders from global tech giant, forming a different story compared with that of TSMC. At the same time, Samsung is steadily advancing its 2nm chip and has started prototype mass production of 2nm Exynos 2600 to be used for powering Galxy s26 in 2026, aiming to achieve the yield of more than 50%. If it is successful, the mass production of the chip will be fully carried out in early 2026.

Samsung is accelerating its 2nm process layout and plans to start mass production at Taylor fab in 2026, which will have significant impact on the market and supply and demand landscape. This will drive the development of semiconductor technology to meet the demand of high-performance, low-power consumption chips, especially in areas such as AI, 5G communications, and HPC. This move will help Samsung to improve its competitiveness in the global semiconductor market and compete intensively with competitors such as TSMC. From the perspective of supply and demand, the mass production of 2nm chips will provide more advanced chip supply for the market and ease the tight supply of high end chips. Apart from helping stabilize chip prices, it will provide strong support for the development of related industry chains and boost the prosperity of the whole semiconductor ecosystem. At the same time, this may stimulate other companies to speed up tech development and capacity expansion, further intensify market competition and drive the development of the industry.

04

On June 25, SEMI released the latest 300mm Fab Outlook Report, showing that the global semiconductor manufacturing industry will keep a strong growth momentum. From the end of 2024 to 2028, global semiconductor production is projected to grow at a compound annual growth rate (CAGR) of 7%, reaching a record high of 11.1 million wafers per month (wpm). A key driver of this growth is the continued expansion of advanced process capacity (7nm and below), which is expected to increase by approximately 69% – from 850,000 wpm in 2024 to 1.4 million wpm in 2028 – representing a CAGR of around 14%, double the industry average.

According to SEMI, the advanced process capacity is expected to achieve a significant milestone in 2026, surpassing one million wafers for the first time, with capacity reaching 1.16 million wpm. In addition, the deployment of 2nm and below capacity shows even more aggressive scaling throughout the forecast period, with production capacity expanding dramatically from less than 200,000 wpm in 2025 to over 500,000 wpm by 2028, reflecting the strong market demand driven by AI applications in advanced manufacturing.

From the perspective of industry landscape, the rapid expansion of advanced process capacity will intensify competition in the global semiconductor industry. Leading tech giants such as TSMC, Samsung and others will continue to hold leading positions in the field of high-end chip manufacturing while other companies will have to invest more in R&D and upgrade their technology level to compete for market share. This may further increase the industry concentration and shows a situation that the strong one gets even stronger. 

As for the market, the increase in production capacity will ease the tight supply in the current semiconductor market, especially to meet the strong demand for high-performance chips in emerging tech areas like AI. With more production capacity released, chip prices will be stabilized gradually and even go down, which will promote the wider application and rapid development of AI and the promote the prosperity of related industries. 

05

On June 26, 2025, TGVTECH, a subsidiary of Wg Tech (Jiangxi) Group Co., Ltd., officially signed a special development agreement with Polar Bear Tech to jointly promote the development of a highly integrated AI computing chip with heterogeneous Chiplets and multi-layer (full) glass substrate stacks. This cooperation aims to accelerate the commercialization of glass substrate used in Chiplet packaging and semiconductor sector and promote domestic chip industry to achieve independent, controllable production. 

Polar Bear Tech was incubated by the Institute for Interdisciplinary Information Sciences (IIIS) of Tsinghua University. Guided by Andrew Chi-Chih Yao, Academician of Chinese Academy of Sciences and winner of Turing Award, Polar Bear Tech self-developed AI accelerating chip and successfully industrialized it. The team of the Company mainly develops three core technologies: neural processing unit (NPU), Chiplet, and algorithm, to create a system-level solution for Chiplet. TGVTECH, a domestic leader of glass substrate industry and has developed global leading TGV full-process capability relying on the 16-year glass finishing and application technology of its parent company Wg Tech. With the cooperation, the two sides will gave play to their tech advantages, jointly overcame cote tech challenges such as glass substrate HPC chip packaging and interconnection of large sized chips. 

06

On June 26, Wuhan Mengxin Technology Co., Ltd, a leading domestic company engaged in Beidou navigation chips, has unveiled a new generation of high-precision SoC chip, Zhumeng®MX2740A(full system and full frequency points) and Qimeng IV MX2730A(full system and multiple frequency points) and launched two Beidou 2.0 chips (MX2740 and MX2730) for the first time in China. 

This chip has achieved multiple tech breakthroughs and is the smallest full-system full-frequency point chip. Compared with similar products, its size is reduced by 50%, with power consumption down 40%. Moreover, its speed of capturing Beidou signals is increased by 20 folds, its cold start time is less than 5 seconds and its index has reached the international leading level. In addition, the chip has full-frequency point independent capture and tracking capabilities, supports enhanced signals of mainstream low- and high-orbit satellites, and its original Smart Suppress®multi-level anti-interference technology sharply enhances its reliability and adaptability. 

Based on the MX2740 and MX2730 series chips, Mengxin Technology released modular schemes for 24 major areas at the same time, fully covering scenarios of industrial applications and mass applications, such as intelligent lawnmower, drones, precision agriculture, intelligent drive, disaster monitoring, etc., and has achieved high precision, high reliability, always online, real-time response, anti-interference, and low power consumption integration.

07

On June 26, Loongson Zhongke held “2025 Loongson Product Release & User Conference” in Beijing and officially released a new generation homemade general-purpose processor Loongson 3C6000. This processor adopts an independently designed Loongson architecture instruction set, which does not need foreign license. It uses LA664 architecture core and has six-issue pipelines, so its general performance is doubled over the last-gen chip. The Loongson 3C6000 is a 16-core, 32-thread CPU and can be packaged through self-developed Loongson Link interface multi-silicon package to form 32-core 64-thread 3C6000/D and 60/64-core 120/128-thread 3C6000/Q. In terms of performance, the performance of single-core 3C6000/S is comparable to that of 16-core Xeon Silver 4314 released by Intel in 2021, the performance of multi-core 3C6000/D is comparable to that of 32-core Xeon Gold 6338, and the performance of 64-core 3C6000/Q exceeds that of 40-core Xeon Platinum 8380, and its comprehensive performance reaches that of the mainstream products in the market in 2023. 

The series of processors also incorporate RAS security redundancy mechanism and SM acceleration engine and have passed the level II certification of “Safety and Reliable Test Results Announcement”, the highest level of the certification, securing application security in critical areas. In addition, the Loongson 3C6000 series has formed a high-performance autonomous server solution, which can meet the computing needs of general computing, smart computing, storage, industrial control, workstation and other scenarios.