Chinese semiconductor thread II

tokenanalyst

Lieutenant General
Registered Member

Toujing Technology: Full order book, rising storage prices may drive manufacturers to continue expanding production.​

Toujing Technology reported strong Q3 2025 financial results, with revenue rising 124.15% year-on-year to 2.266 billion yuan and net profit up 225.07% to 462 million yuan. This growth is attributed to successful mass production launches of advanced semiconductor equipment for cutting-edge process nodes, along with improved economies of scale and reduced operating expenses.

The company has a full order book, driven by rising demand in the memory market and technological advancements beyond Moore’s Law. Key products—such as PECVD Stack (ONO), ACHM, PECVD Bianca, ALD SiCO, and others based on new platforms (PF-300TPlus/M) and reaction chambers (pX, Supra-D)—have passed customer certification and entered mass production.
1764861800956.png1764861830291.png

Chairman Lu Guangquan emphasized that the industry is entering a post-Moore’s Law era, with innovations like gate-all-around (GAA) transistors, backside power delivery, high-K metal gates, and 3D integration driving demand for advanced equipment. The rapid growth of data has accelerated adoption of high-bandwidth memory (HBM) and multi-layer 3D NAND flash, increasing the need for thin-film deposition and bonding technologies.

Toujing is deepening its expertise in thin-film deposition (PECVD, ALD, HDPCVD, etc.) and 3D integration equipment, including wafer-to-wafer and chip-to-wafer hybrid bonding. The company has received repeat orders for hybrid bonding tools and successfully shipped next-generation high-speed, high-precision systems; chip-to-wafer bonding verification is ongoing. It also completed development of laser lift-off technology after permanent bonding.

With existing industrial bases in Shenyang and Lingang (Shanghai) supporting over 700 units/year production capacity, Topwin is constructing a second plant in Shenyang to further expand output—ensuring it can meet rising demand.

Please, Log in or Register to view URLs content!
 

tokenanalyst

Lieutenant General
Registered Member
Everyone talks about building power plants for AI datacenters but nobody talk about developing technology that lower energy usage.​

The gallium nitride power module developed by Jiufengshan Laboratory could save 300 million kilowatt-hours of electricity annually for ultra-large AI computing centers.​


The Jiufengshan Laboratory has developed a breakthrough gallium nitride (GaN) power module that could save 300 million kilowatt-hours of electricity annually in ultra-large AI computing centers—equivalent to about 240 million yuan in cost savings. Each module, no bigger than a fingernail, can be scaled up to one million per 1-gigawatt AI rack.

Currently, power modules consume 1 billion kWh/year (11% of total data center usage) by converting high-voltage electricity into low-voltage for processors. Replacing traditional silicon-based chips with GaN technology reduces energy use by 30%, shrinks module size by 30%, and cuts costs in half, offering a major leap in efficiency.

The lab, composed of seven PhDs, overcame nanoscale manufacturing challenges to build an end-to-end GaN supply chain—from material growth to packaging. Their innovation has already attracted commercial orders exceeding 10 million yuan and partnerships with domestic power suppliers.

With global data center electricity demand expected to double by 2030, driven by AI, the technology is timely. The team is now entering pilot testing, aiming for mass production within 3–5 years, targeting a market worth hundreds of billions of yuan.

The project also won recognition in China’s National Postdoctoral Innovation Competition and attracted strong investor interest. Led by Dr. Li Sichao, the lab aims to accelerate deployment, delivering high-efficiency power solutions critical for sustainable AI infrastructure.

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 
Last edited:
Top