Chinese semiconductor thread II
- Thread starter vincent
- Start date
They smuggled too many chips. And Deepseek, Butedance already have very efficient training framework. So they brought too many GPUs, and now some data center are empty. Don't expect anyone to make big sales till demand increases. Deepseek put a huge dent in demand for GPUs.
Just for the record, since this hasn't been posted here: huawei is making it's own lithography machine, a 14nm Argon Fluoride(ArF) Immersion(ArFi) Lithography machine, mass production scheduled for 2026-2027 source: , literally the only piece of media on it in the english language: . Do anyone know the specs for it? How does it compare with the duvs that the dutch government blocked to china in 2023, kneecaping them(free market preacher netherlands btw, what do u expect from 14 eyes member), the NXT:1970i, NXT:1980Di, NXT:2000i, NXT:2050i, NXT:2100i, and do anyone have more info on the SMEE SSA800i-10w, and SMEE SSA900i-10w?
Also Hua Hong can make 7nm too(source: , not only SMIC, and Huawei's fabs. Thanks in advance
Also Hua Hong can make 7nm too(source: , not only SMIC, and Huawei's fabs. Thanks in advance
The 10w is not accurate. RSLaser ArFi lasers are at least between 90-60 Watts.
Chinese semiconductor industry
2000i has a higher wph and better overlay accuracy? Is there any other advantage? Someone posted a chart from ASML here once that showed the major difference between DUVi machines wasn't power source output or numerical aperture of the optics, but the accuracy of the workpiece. Given the...
www.sinodefenceforum.com
From Hongmeng Ecosystem to RF Front-end: Decoding the Industrial Chain "Development Revolution" Behind the Reorganization and Acquisition of Gokewei
As competition in the global semiconductor industry chain becomes increasingly fierce, Goke Microelectronics recently disclosed that it plans to purchase 94.366% of the shares of SMIC Ningbo held by 11 transaction parties including Ningbo Yongxin through the issuance of shares and payment of cash. This move quickly attracted industry attention.
Public information shows that Gokewei is an integrated circuit design company recognized by the Ministry of Industry and Information Technology. It has promoted domestic substitution in many core areas such as audio and video encoding and decoding, live satellite channel demodulation, digital and analog mixing, advanced security encryption, and solid-state storage controller chips.
As a key component in 5G communications, the Internet of Things, automotive electronics and other fields, filters have long been monopolized by overseas giants, with a low localization rate. Among them, BAW filters have extremely high technical barriers due to their complex thin film deposition and micromachining processes.
The transaction announcement shows that the target of this transaction, SMIC Ningbo, focuses on special process technologies such as RF front-end, MEMS sensors, and high-voltage simulation, and is one of the few domestic manufacturers that mass-produce BAW filters. SMIC Ningbo has built a globally rare "full-band coverage + full-process penetration" capability in the field of filter manufacturing, becoming one of the few domestic companies that can achieve a SUB-6GHz full-band product matrix, forming a complete solution from low-frequency IoT to high-frequency 5G communications.
Jinglue Semiconductor has completed hundreds of millions of yuan in financing and is a domestic in-vehicle interconnection and switching chip developer
Recently, Jinglue Semiconductor completed a financing of several hundred million yuan from a strategic investment from SDIC. It will be mainly used to support the acceleration of R&D innovation and mass production of in-vehicle interconnection and switching chips, and help the domestic substitution strategy achieve breakthroughs in the core links of automotive intelligence.
Jinglue Semiconductor is a rare domestic provider of in-vehicle Ethernet chips and full-stack in-vehicle high-speed network solutions. Its technical capabilities are comparable to those of first-class foreign companies. It has high-speed and high-performance analog, digital signal processing chips, system-level chips and mixed analog and digital design capabilities, and has 100% self-developed intellectual property rights and large-scale communication chip mass production experience.
The core team of Jinglue Semiconductor comes from top semiconductor companies in Silicon Valley, with rich experience in analog, digital, DSP and SoC fields, and the company's founder has long-term successful technical accumulation in the fields of Ethernet communication chips and high-speed SerDes. Jinglue Semiconductor has mass-produced various PHY and Switch chips of 100M, 1G, and 2.5G, with cumulative shipments of nearly 200 million. Recently, Jinglue Semiconductor successfully achieved the first mass production of 32G high-speed SerDes IP on the SMIC 28nm process, helping to expand the application of domestic mature processes in high-speed interconnect chip fields such as AI and data centers.
Market reaction to this news looks too harsh (5% drop for SMIC). May be because it's being referred to as SMIC Ningbo in the media. SMIC refers to this company as Ningbo Semiconductor International (NSI) rather than SMIC Ningbo. SMIC called it an associate rather than a subsidiary because SMIC only had a minority stake (about 14%) in this fab and it contributed very little to SMIC bottom line. Goke has clearly bought the bulk of it from other shareholders, as SMIC only owned 14% to begin with.
Beijing Keyi Hongyuan metrology lightsource patent was authorize.
Can someone much more knowledgeable make compilation for every category in euv like the lightsource and the subcategory in lightsource, the optics and the subcategories like ule glass, etc? And make it in a document, it's hard to get a macro view of the progress in euv, thank you.
Well, they have the only competitive and advanced AI chips and the best ecosystem that runs with it at the moment, Reason the whole world relies on them(including China to some extent). So until China can produce AI chips as good as Nvidia in all categories and with a corresponding eco-system as good as CUDA, then others will prioritise Nvidia obviously. Huawei and others first need to be able to corner the entire Chinese market before they can have a chance of challenging Nvidia around the world with better products and cheaper prices. That's the only way countries will be able to go against US sanctions to buy Huawei chips and eco-system instead of Nvidia..