Chinese semiconductor thread II

tokenanalyst

Brigadier
Registered Member

Huilong Semiconductor High-end Process Project with a total investment of 320 million yuan has started construction​

The project is located on the east side of Xicheng Canal, south of Zhenyang Road and west of Xiaoqingyang Road in Qingyang Town, with a total investment of 320 million yuan. The project plans to build a seven-story comprehensive building and four three-story supporting workshops. After completion, it will have an annual production capacity of 2.075 million sets of smart heaters and 300,000 sets of smart temperature control assemblies. It is committed to becoming a global leader in semiconductor heating and heat dissipation solutions.
1755528964288.png1755528981713.png
1755529019541.png1755529050114.png
According to data, Huilong has been deeply engaged in thermal control technology for over 20 years, achieving remarkable results: in 2022, it was recognized as a Jiangsu Province Specialized, Advanced, and Innovative Small and Medium-sized Enterprise, in 2023, was selected for inclusion in the Jiangsu Province Listing Cultivation Program, and in 2024, it was awarded the national "Little Giant" designation for Specialized, Advanced, and Innovative Enterprises. The company's comprehensive product line of semiconductor heaters, temperature controllers, and intelligent control systems has achieved breakthroughs in several key technologies, including uniform heating, ultra-clean insulation, CMS system control, and multi-stage temperature control. These breakthroughs effectively address heating and temperature control challenges in multiple processes, including semiconductor deposition, etching, cleaning, and coating and development. Leveraging these technological achievements, Huilong can provide semiconductor companies with comprehensive solutions from components to systems, meeting the temperature uniformity, stability, and safety requirements of semiconductor chip manufacturing processes. This enables domestic substitution of related technologies and provides strong support for the accelerated rise of "Chinese chips."

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Huahai Qingke strategically invests in Suzhou Bohongyuan​


On August 18th, Huahai Qingke announced on its official WeChat account that it recently completed a strategic investment in Suzhou Bohongyuan Equipment Co., Ltd., marking the official start of in-depth cooperation between the two leading companies in precision manufacturing equipment. This strategic partnership will leverage the complementary strengths of both parties in technology research and development, market channels, and brand influence to jointly build a one-stop planarization equipment platform covering precision thinning, grinding, polishing, and other processes, providing more comprehensive solutions for global customers.

1755529362698.png

Huahai Qingke, a leading domestic company in high-end integrated circuit equipment, boasts significant advantages in core technologies such as CMP equipment. Suzhou Bohongyuan has long specialized in the research and development of high-precision grinding and polishing equipment for hard and brittle materials such as compound semiconductors and sapphire, and its products have successfully entered the international market. Through this strategic partnership, the two companies will achieve deep integration of technological research and development, share supply chain resources, and leverage Suzhou Bohongyuan's overseas sales channels to accelerate their global expansion and further enhance their competitiveness in niche markets.
1755529407765.png

With the rapid development of power semiconductors and advanced packaging technologies, market demand for related substrate materials and thinning and polishing equipment used in the packaging process continues to grow. As a major global semiconductor consumer market, China is experiencing growth opportunities in the upstream equipment industry. Huahai Qingke's strategic investment not only strengthens its industrial chain presence in key semiconductor equipment, but will also promote technological advancement and industrial upgrading of domestically produced high-end precision planarization equipment.

In the future, the two parties will take this opportunity to establish a long-term and stable strategic partnership. Through in-depth collaboration in R&D innovation and market expansion, they will jointly enhance the competitiveness of domestically produced equipment in the global market. This cooperation will not only help meet the urgent demand for high-end equipment in domestic and international markets, but will also inject new impetus into the independent and controllable development of China's semiconductor equipment industry.

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Huawei launches UCM technology, potentially reducing reliance on HBM​


Currently, artificial intelligence has entered a deep water stage of development, and AI reasoning is becoming a key stage for the next explosive growth. Reasoning experience and reasoning cost have become the golden yardstick for measuring the value of models.
However, problems such as "unable to push", "slow to push" and "expensive to push" (computing costs are too high) have become difficulties encountered in the AI reasoning process.
Huawei recently launched its innovative AI inference technology, Unified Cache Manager (UCM), aiming to upgrade the AI inference experience, improve the cost-effectiveness of inference, and accelerate the positive cycle of AI business.
It is worth noting that according to media reports, UCM technology is expected to alleviate the current task stagnation and response delay problems caused by the shortage of HBM (high-bandwidth memory) resources, provide a smoother experience for AI applications, and thus reduce the industry's excessive dependence on HBM.
UCM technology is centered around KV Cache and integrates multiple types of cache acceleration algorithm tools to hierarchically manage the KV Cache memory data generated during the inference process. It aims to expand the inference context window, achieve a high-throughput, low-latency inference experience, and reduce the inference cost per token.

1755529618434.png
According to Huawei's official introduction, UCM technology has intelligent hierarchical caching capabilities, which can achieve on-demand flow in storage media such as HBM, DRAM, and SSD based on memory popularity; at the same time, it integrates multiple sparse attention algorithms to achieve deep collaboration between storage and computing, increasing TPS (tokens processed per second) by 2-22 times in long sequence scenarios, significantly reducing the inference cost per token.
At the same time, relying on UCM's hierarchical, adaptive global prefix caching technology, the system can directly access KV cache data, avoiding repeated calculations and reducing first-token latency by up to 90%.
Furthermore, UCM offloads ultra-long sequence caches to external professional storage in a layered manner. Through algorithmic innovation, it overcomes model and resource limitations and achieves a 10-fold expansion of the inference context window to meet the needs of long text processing.


Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Hua Hong Semiconductor acquires Huali Microelectronics, aiming for a 12-inch wafer production milestone of 200,000 wafers.​


Hua Hong Semiconductor Co., Ltd. ("Hua Hong Semiconductor"), a leading domestic specialty wafer foundry, announced that it is planning to acquire a controlling stake in its sister company, Shanghai Huali Microelectronics Co., Ltd. ("Hali Micro"), through a combination of a share issuance and cash payment. The announcement states that the target asset for this acquisition is the equity interest in Hua Hong Fab 5, an asset operated by Shanghai Huali Microelectronics Co., Ltd. that competes with Hua Hong in the 65/55nm and 40nm processes. The target asset is currently in the process of being separated. The acquisition of Huali Microelectronics by Hua Hong Semiconductor was not a "sudden whim" but a planned strategic integration, the core goal of which was to resolve the historical problem of "intra-industry competition".

Huali Microelectronics initially positioned itself to fill the gap in China's 12-inch advanced logic processes. However, as market demand shifted, 65/55nm processes gradually became the mainstream choice for the Internet of Things, consumer electronics, and automotive electronics sectors. Consequently, as both companies grew, some overlap in their businesses became apparent, particularly at the 65-55nm and 40nm nodes.

Huali Microelectronics primarily owns Huahong Fab 5 and Huahong Fab 6. Huahong Fab 5, built in 2011, is China's first fully automated 12-inch production line, covering processes from 65/55nm to 40nm. Hua Hong Semiconductor primarily owns Huahong Fab 1, Fab 2, Fab 3, Fab 7, and Fab 9. Fab 7 covers processes from 90nm to 65/55nm. Fab 9 is a newly built production line, slated to enter mass production by the end of 2024, with a planned monthly production capacity of 83,000 wafers, focusing on advanced specialty processes at 40nm and below.

The industry believes that if Hua Hong Semiconductor successfully acquires Hua Hong Fab 5, it will significantly increase its 12-inch wafer foundry capacity. After the integration is complete, Hua Hong Semiconductor will have three 8-inch production lines (Hua Hong Fab 1, Hua Hong Fab 2, and Hua Hong Fab 3) and three 12-inch production lines (Hua Hong Fab 5, Hua Hong Fab 7, and Hua Hong Fab 9).

In terms of production capacity, according to the official websites of Hua Hong Group and Huali Microelectronics, Hua Hong Fab 5 currently has a stable monthly production capacity of 38,000 wafers, while Fab 7 currently has a monthly capacity of 40,000 wafers (with a planned capacity of 94,000 wafers). The newly built Wuxi Phase II 12-inch production line (i.e., Hua Hong Fab 9) has a planned monthly capacity of 83,000 wafers. Without considering the new capacity of Fab 9, Hua Hong Semiconductor's 12-inch production capacity is expected to increase to 78,000 wafers. From a more forward-looking perspective, as the capacity of Fabs 7 and 9 gradually increases, Hua Hong Semiconductor's projected 12-inch production capacity will reach 215,000 wafers. This scale is expected to firmly establish Hua Hong Semiconductor as a leading global specialty foundry and foster strong economies of scale.

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Suzhou Yilan Micro-Silicon Photonics Heterogeneous Integrated Thin-Film Lithium Niobate Pilot Line is online​


On August 15, 2025, Suzhou Elan Micro Semiconductor Technology Co., Ltd. (hereinafter referred to as "Elan Micro") successfully held a launch ceremony for its silicon photonic heterogeneous integration thin-film lithium niobate pilot line at Suzhou Nano City. As one of the few companies internationally focused on silicon photonic heterogeneous integration, the successful launch of this pilot line marks a significant milestone for Elan Micro, signaling its ability to provide high-speed optical interconnect solutions for cutting-edge industries such as optical communications, AI computing, and quantum information technology.

At the launch ceremony, Dr. Chen Wei, Chairman of Suzhou Yilan Micro, delivered a speech, saying: "The launch of the pilot line is an important milestone for Yilan Micro from technology research and development to industrialization. We firmly believe that silicon photonics heterogeneous integration technology will become the key to breaking the technological barriers of optical chips, and the large-scale mass production of silicon photonics heterogeneous integration thin-film lithium niobate chips will help China occupy a more important position in the global optical communications industry."

As a top expert in the optical communications industry, Liu Bo, deputy mayor of Suzhou Municipal Government, highly recognized the continuous progress of Yilan's micro-heterogeneous integration platform, and expressed great appreciation for the company's courage to undertake the key special project "Metropolitan Single Carrier 1.2Tb/s Coherent Optical Transmission Integrated Chips and Modules" of the National Key R&D Program "Information Photonics Technology". He believes that the company will play a greater role in the 3.2T/6.4T next-generation artificial intelligence data center application scenarios, which is of vital strategic significance for my country to break through the bottleneck of key core technologies of silicon-based heterogeneous integration and overcome the "bottleneck" problem of ultra-high-speed optical chips.

--------------------------------------------------------------

Some products:

Data Center 1.6T-DR8 Silicon Photonics Chip

Silicon photonic chips, developed using CMOS processes and based on silicon-based optoelectronics technology, offer the advantages of high bandwidth, high speed, low energy consumption, and low cost.
A Mach-Zehnder interferometer (MZI) is used to convert phase modulation to intensity modulation: a low-drive-voltage, high-bandwidth modulator (MZM) based on a Mach-Zehnder interferometer (MZI) modulates the phase of light by changing the refractive index of the waveguide under the influence of an electric field.

1755532551714.png1755532608198.png

Data center 400G/800G/1.6T silicon-based heterogeneous integrated photonic chip

Silicon photonic chips, developed based on CMOS technology and centered on silicon-based optoelectronics technology, have the advantages of high bandwidth, high speed, low energy consumption and low cost.

1755533550898.png

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

RF chip company Zhenlei Technology profits increased​


Zhejiang Zhenlei Technology Co., Ltd. (Zhenlei Technology) recently released its 2025 semi-annual report. The announcement revealed that the company achieved operating revenue of 205 million yuan in the first half of the year, a year-on-year increase of 73.64%. Net profit attributable to parent company shareholders was 62.3197 million yuan, a year-on-year increase of 1006.99% .

Zhenlai Technology's operating income was 132 million yuan, a year-on-year increase of 106.2% ; net profit attributable to shareholders of the parent was 39.85 million yuan, a year-on-year increase of 252.5%; net profit attributable to shareholders of the parent after deducting non-recurring items was 36.10 million yuan, a year-on-year increase of 616.3% .

Core revenue comes from power management chips

According to its official website, Zhenlai Technology, founded in September 2015, specializes in terminal RF front-end chips, high-density packaged microwave modules, and microsystems. It is a private high-tech enterprise integrating design, development, manufacturing, and sales. It has established four leading domestic platforms: terminal RF front-end chips, phased array T/R components and microsystem design, high-density integrated packaging, electrical performance testing, and a reliability center.

The company's terminal RF front-end chips, including power amplifiers, low-noise amplifiers, and RF switches, are widely used in wireless communication terminals such as ad hoc networks, radios, digital intercoms, and navigation.
Its microsystems and modules can be used in various phased array payload systems, including those onboard spacecraft, ground vehicles, and ships. The company has become a core supplier of RF chips, microsystems, and modules in the domestic communications and radar sectors.

The company's main product sales in the first half of the year included RF transceiver chips, high-speed and high-precision ADC/DAC chips, power management chips, microsystems and modules , etc., providing chip and microsystem products and technical solutions from antennas to signal processing.

1755534187686.png

1755534228255.png
1755534249979.png



The semi-annual report shows that RF transceiver chips and high-speed, high-precision ADC/DAC chips generated revenue of 80.8637 million yuan, power management chips generated 103 million yuan, and microsystems and modules generated 51.0474 million yuan. Currently, power management chips are the core revenue source, followed by RF transceiver chips and high-speed, high-precision ADC/DAC chips . The remaining businesses account for a relatively small proportion.

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
 

Fertorstar

Just Hatched
Registered Member
Well a least in the high end gaming space there is no alternatives, YET, that area looks like most of the "smuggling", if you can called it that, is happening, in the AI space looks like homegrown GPUs makers like Hisilicon, Biren and Moore are making progress with a lot of then introducing FP8 circuitry and better software stack, that nascent ecosystem is what the Chinese government is trying to protect against NVIDIA monopolistic practices.
I mean, sure, but that largely confirms what I said. There being market segments that China cannot backfill is evidence of Nvidia being better.

We don’t live forever - we would like products to solve today’s problems because we can’t wait many years for hypothetical products in the future. But it’s exactly that tension between software outputs today vs. hardware inputs later that’s causing China to be confused as to whether it wants to trade its rare earth card for H20s or whether it wants to jawbone H20s into being banned (even ignoring how the B200 has 30x more compute and 2x more memory)
 

tokenanalyst

Brigadier
Registered Member
I mean, sure, but that largely confirms what I said. There being market segments that China cannot backfill is evidence of Nvidia being better.

We don’t live forever - we would like products to solve today’s problems because we can’t wait many years for hypothetical products in the future. But it’s exactly that tension between software outputs today vs. hardware inputs later that’s causing China to be confused as to whether it wants to trade its rare earth card for H20s or whether it wants to jawbone H20s into being banned (even ignoring how the B200 has 30x more compute and 2x more memory)
The rare earth "card" was a part of a bigger trade deal than just the H20.

The issue is what apart from Gaming, who drivers are the main issue, I don't think that what NVIDIA is offering now is much better than what Chinese players like HiSilicon is offering, at hardware level. Cuda has a big impact more than the hardware due all the legacy tools already made around it, but here is the kicker, the reason why Nvidia doesn't want to abandon the Chinese market that is willing to pay a 15% bribe to enter that market is that they know that due NVIDIA absent that an domestic AI ecosystem is brewing in China, from GPUs to CUDA like solutions, in just a matter of few years they could find themselves no only out of the Chinese market but also competing with Chinese companies that were let grow with only competing among themselves. NVIDIA monopoly can only exist if they are there to compete. The "smuggling" is not going to help them in the short run and much less in the long run, and worse is mostly happening in gaming GPUs and older AI GPUs.
 

tphuang

General
Staff member
Super Moderator
VIP Professional
Registered Member
I mean, sure, but that largely confirms what I said. There being market segments that China cannot backfill is evidence of Nvidia being better.

We don’t live forever - we would like products to solve today’s problems because we can’t wait many years for hypothetical products in the future. But it’s exactly that tension between software outputs today vs. hardware inputs later that’s causing China to be confused as to whether it wants to trade its rare earth card for H20s or whether it wants to jawbone H20s into being banned (even ignoring how the B200 has 30x more compute and 2x more memory)
what did I tell you about being a newbie and read a little bit before you posting on this thread?

You are banned from this thread for a week.
 
Top