weixin

Author page description

February 5

  • Exclusive Interview with DeepSeek Founder Liang Wenfeng

    Dark Current Journal

    ➀ DeepSeek's V2 model sparked a fierce price war in the large model industry;

    ➁ Liang Wenfeng discusses the unexpected success and pricing strategy of DeepSeek;

    ➂ The focus on model architecture and the pursuit of AGI (Artificial General Intelligence);

    ➃ DeepSeek's commitment to open-source and community contribution;

    ➄ The challenges and opportunities in the AI industry, including competition with tech giants;

    ➅ The future of AI and the importance of original innovation.

  • Rapidus Announces Trial Production of 2nm Chips on April 1st!

    XinZhiXun

    ➀ Rapidus, a Japanese wafer foundry startup, is building a 2nm wafer factory in Hokkaido with a target to start trial production in April 2025 and mass production in 2027;

    ➁ Rapidus CEO Koike Atsuyoshi stated that the factory construction is proceeding smoothly and trial production will begin on April 1st;

    ➂ Rapidus aims to achieve an estimated cumulative revenue of 18 trillion yen from mass production between 2027 and 2036;

    ➃ TSMC is leading in the mass production of 2nm chips, but Rapidus has advantages in process speed, yield, and performance, and aims to catch up with TSMC;

    ➄ Rapidus, founded in August 2022, is a joint venture of eight Japanese companies, including Toyota, Sony, NTT, NEC, SoftBank, Denso, and Mitsubishi UFJ, with a target to mass produce 2nm processes by 2027;

    ➅ Rapidus plans to use robots and AI to create an automated 2nm production line in northern Japan, aiming to reduce delivery times by a factor of three compared to competitors;

    ➆ Rapidus has become the first company in Japan to receive EUV lithography machines, with the first batch of EUV lithography equipment arriving in Japan on December 14, 2024;

    ➇ Rapidus' fully automated factory gives it a competitive edge, as the back-end processes of chip manufacturing, such as interconnection, packaging, and testing, are labor-intensive;

    ➈ CEO Atsuyoshi Koike states that this will provide higher performance and faster delivery times for the same 2nm products compared to other companies.

    AutomationRapidus
  • NXP Plans to Lay Off 1,700 Employees Globally!

    XinZhiXun - Langke Jian

    ➀ NXP announced plans to lay off approximately 1,800 employees globally due to increased tariffs in the US and market pressures.

    ➁ The decision is not directly related to concerns about potential trade wars but is related to broader market conditions.

    ➂ NXP expects its workforce to 'slightly decrease', with the number of employees reduced no more than 5% globally.

    ➃ NXP's President and CEO Kurt Sievers commented on the company's performance in 2024, noting a 5% decline in revenue and a focus on managing growth strategies.

    ➄ NXP obtained a €1 billion loan from the European Investment Bank to accelerate semiconductor innovation.

    LayoffsNXPsemiconductor industry
  • The Global Race for Diamond Semiconductor

    icbank

    ➀ Diamond semiconductors are poised to become the future king of semiconductor materials due to their exceptional physical properties like hardness, acoustic velocity, thermal conductivity, and Young's modulus.

    ➁ Diamond semiconductors are highly suitable for power semiconductors due to their excellent dielectric breakdown strength, high thermal conductivity, and high-frequency operation capability.

    ➂ Japan is leading in the research and development of diamond semiconductor materials, with companies like Orbray and Saga University making significant breakthroughs.

    ➃ Diamond semiconductors have potential applications in communication satellites, nuclear power, and quantum computing.

    ➄ The United States and Europe are also actively investing in diamond semiconductor research and development, with companies like Diamond Foundry and DIAMFAB making advancements.

    ➅ China is also making progress in diamond semiconductor research and development, with companies like Hefeng Diamond and Shangtian Xuan Semiconductor Technology Co., Ltd. leading the way.

    ChinaJapanPower SemiconductorsQuantum ComputingUSA
  • TSMC Plans to Increase Prices by 15%

    Semiconductor Industry Observation

    ➀ TSMC is planning to increase the price of its most advanced semiconductor wafers by up to 15% this year.

    ➁ The increase is attributed to rising production costs and potential U.S. tariffs.

    ➂ The price hike could further deviate the final product from the Moore's Law curve.

    ➃ TSMC is expected to start raising prices from the 7nm node, with the cost of a wafer approximately $10,000.

    ➄ The company's most privileged client, Apple, currently pays $18,000 per 3nm wafer.

    ➅ The new tariffs could push the price of 3nm wafers up to $20,000 to $23,000 per wafer.

    ➆ Higher wafer prices have already led to higher equipment prices, but performance improvements have been less impressive.

    ➇ The recent iPhone chip did not significantly outperform its predecessor, and the Nvidia RTX 50 series is one of the most disappointing graphics cards in recent years.

    ➈ TSMC plans to start producing 4nm chips in Arizona this year and aims to produce 3nm and 2nm nodes in the next decade.

    ➉ There are rumors that TSMC will build a 1nm super wafer factory in Tainan, Taiwan.

    Price IncreaseTSMCsemiconductor
  • Google's Deep Dive into TPU

    icbank

    ➀ TPU is a specialized computing core for matrix multiplication (TensorCore) connected to a stack of fast memory (HBM);

    ➁ TensorCore consists of MXU (Matrix Multiplication Unit), VPU (Vector Processing Unit), and VMEM (Vector Memory);

    ➂ TPU is very fast in matrix multiplication and has high FLOPs/s;

    ➃ TPU has various network configurations like ICI and DCN for efficient communication.

    TPU
  • Wafer Foundry Forecast: Growth of 20%

    icbank

    ➀ The global wafer foundry industry is expected to grow by 20% this year, driven by the surge in demand for advanced semiconductor processes due to the AI boom.

    ➁ TSMC, the leading wafer foundry company, is expected to outperform the industry average growth rate.

    ➂ TSMC's stock price rebounded and reached over 28 trillion New Taiwan Dollars, with revenue growth projected to be between 24% and 26% in 2025.

    ➃ The global semiconductor revenue is forecasted to grow by 13.4% this year, reaching a peak of $714 billion.

    ➄ AI applications are driving the demand for AI/HPC chips and storage, with high bandwidth memory (HBM) demand particularly strong.

    ➅ Despite the growth potential, geopolitical and trade tariff factors may pose negative impacts on the semiconductor industry.

    AI DemandTSMCWafer Foundrysemiconductor
  • AMD Issues Warning, Stock Price Plummets

    icbank

    ➀ AMD's stock price plummeted after disappointing data center business outlook;

    ➁ Despite strong overall revenue growth in the fourth quarter, data center performance overshadowed other achievements;

    ➂ Concerns about AMD's weakened momentum in the AI device field were confirmed with DeepSeek, a Chinese startup, entering the market at a lower cost, further exacerbating worries;

    ➃ AMD's total revenue grew 24% to $7.66 billion in the fourth quarter, exceeding expectations;

    ➄ AMD remains optimistic about 2025, predicting double growth in revenue and EPS.

    AIAMDData centerstock market
  • Semiconductor Industry: One Acquisition, One Sale Plan

    Semiconductor Industry Observation

    ➀ Teradyne is acquiring technology and development team of automatic test equipment from Infineon Technologies, which will allow Teradyne to continue supporting Infineon with manufacturing support services.

    ➁ The 80-person team based in Regensburg, Germany, will help Teradyne accelerate the development of power semiconductor test systems.

    ➂ Magnachip Semiconductor, after being blocked by the US government from being acquired by a Chinese private equity firm four years ago, is seeking to sell again.

    ➃ Potential buyers for Magnachip include Samsung Electronics, LG Electronics, LX Group, Doosan Group, and DB HiTek Co.

February 4

  • Who is Investing in Cambrian Era?

    远川投资评论

    ➀ Cambrian Era has been striving to become the Chinese version of NVIDIA for nine years, with investors already enjoying the stock price surge;

    ➁ Yuanhe Yuandian, an early investor, was confident in Cambrian Era's potential due to its technology and market prospects;

    ➂ Huawei's support accelerated Cambrian Era's commercialization, but the company faced challenges after losing Huawei as a major customer;

    ➃ Cambrian Era's revenue mainly comes from government-led projects, raising concerns about its sustainability;

    ➄ Despite challenges,公募基金 and ETFs have been investing in Cambrian Era, driven by the AI sector's growth and domestic substitution trends.

    AI Chipinvestmentstock market
  • DeepSeek Gains Support from Three Domestic GPU Powers: A Significant Lesson for the World

    智东西

    ➀ DeepSeek has reached a new record with over 20 million daily active users in its first 20 days;

    ➁ DeepSeek has become the fastest-growing AI app globally, surpassing ChatGPT in terms of user growth;

    ➂ Chinese tech giants Alibaba Cloud and Baidu AI Cloud have announced support for DeepSeek-V3 and DeepSeek-R1 models;

    ➃ China's chip companies such as Huawei, Tencent, and Alibaba are also supporting DeepSeek.

    AI
  • 张忠谋两万字访谈:谈黄仁勋、28nm、苹果和其他

    半导体行业观察

    ➀ Zhang Zhongmou discusses his relationship with Jensen Huang and the special relationship between TSMC and NVIDIA, including how they met and how they resolved conflicts.

    ➁ He shares the behind-the-scenes story of TSMC's cooperation with Apple, the bet on 28nm technology, and the success of TSMC's 28nm technology.

    ➂ He talks about the challenges faced by TSMC during the economic crisis in the early 2000s, the impact of layoffs, and the importance of research and development.

  • Cerebras CEO: DeepSeek Triggers Surge in Enterprise Demand for AI Chips!

    XinZhiXun (芯智讯)

    ➀ Cerebras Systems' wafer-scale AI chip (WSE-3) outperforms the fastest GPUs by 57 times in executing DeepSeek-R1 models with 70 billion parameters.

    ➁ Cerebras CEO Andrew Feldman states that enterprise clients are highly enthusiastic about DeepSeek's new R1 inference model, with a surge in demand within ten days of its launch.

    ➂ The WSE-3 chip, made on a 12-inch wafer with TSMC's 5nm process, has 4 trillion transistors, 900,000 AI cores, 44GB on-chip SRAM, and a total memory bandwidth of 21PB/s, with a peak performance of 125 FP16 PetaFLOPS.

    ➃ DeepSeek-R1 offers performance comparable to advanced inference models from OpenAI at a low training cost and has been open-sourced, allowing tech firms to build AI applications and chip manufacturers to optimize for the model.

    ➄ Andrew Feldman emphasizes that while DeepSeek poses some risks, users should exercise basic judgment, as seen with the use of electric saws.

    AI ChipCerebras

January 28

January 27

  • DeepSeek Founder Interview: China's AI Can't Always Follow, Someone Must Stand at the Forefront of Technology

    深度学习与NLP

    ➀ DeepSeek V3 model became popular for being open-source with a training cost 1/11th that of Llama 3.1 405B while outperforming it.

    ➁ The founder, Liang Wenfeng, discusses how DeepSeek became a price disruptor and how they prioritize research over commercialization.

    ➂ DeepSeek is committed to original innovation and sees the future of AI in China requiring someone to stand at the forefront of technology.

  • How to Properly Use DeepSeek: A Comprehensive Guide Before the Spring Festival

    未知

    ➀ DeepSeek's most important secret is to abandon prompt templates and focus on real scenarios and specific needs.

    ➁ The 'Say Human Language' feature can make DeepSeek's responses more accessible and understandable.

    ➂ DeepSeek has a powerful depth thinking skill that can be enhanced with specific prompts like critical thinking and reverse consideration.

    AI
  • Deepseek Advanced Usage Guide

    槽边往事 和菜头 出品

    ➀ Using Deepseek, an AI model for inference, to generate poetry in different styles (e.g., Li Bai, Su Dongpo).

    ➁ Utilizing Deepseek to create bedtime stories for children.

    ➂ Explaining complex concepts like 'reflexivity' using simple language with Deepseek's help.

January 25

January 19

  • HaiNengDa Pleads Guilty to Stealing Motorola's Trade Secrets

    芯智讯

    ➀ HaiNengDa pleads guilty to stealing Motorola's trade secrets related to digital mobile radio technology.

    ➁ HaiNengDa will be fined up to $60 million and must pay additional compensation to Motorola Solutions.

    ➂ The civil litigation continues, with HaiNengDa seeking to overturn previous judgments and appeal to the Seventh Circuit Court of Appeals.

  • The Bittersweet Symphony of AI Chip Companies on the Eve of IPOs

    雷锋网

    ➀ An engineer named Hao Tian gave up his 8 million yuan worth of stock options and quit before his company's IPO.

    ➁ Senior engineer Jun Yuan is optimistic about the company's future and believes that the company will achieve a good financial status after going public.

    ➂ Sales representative Han Lin faces significant pressure to sell AI chips, which must be operational for the company to generate revenue for IPO approval.

    AI芯片IPO

January 18

January 17

  • The Behind-the-Scenes Capital Tug-of-War Behind Zhao Ming's Exit from Honor

    旭日大数据

    ➀ Zhao Ming resigned as CEO of Honor due to health reasons, with Li Jian taking over;

    ➁ Honor completed its shareholding restructuring and is preparing for an IPO;

    ➂ Li Jian, the new CEO, has extensive experience in strategic management and is seen as a response to capital and market needs;

    ➃ Honor is focusing on overseas markets and high-end markets with the appointment of new leaders;

    ➄ Zhao Ming's leadership was crucial for Honor's independent growth and strategic adjustments;

    ➅ Frequent personnel changes and concerns about the IPO process have raised market concerns;

    ➆ The exit of Zhao Ming and the capital struggle may be due to issues with the distribution of interests and strategies after the shareholding reform;

    ➇ The smartphone industry capital logic and the future of Honor are discussed.

    Smartphone Industryhonor

January 12

  • Who is Investing in China's Version of NVIDIA and AMD?

    WeChat Official Account

    ➀ Two Chinese startups, known as the 'Chinese versions of NVIDIA' and 'AMD', are striving for IPOs.

    ➁ Both companies have raised over 5 billion yuan in funding, with top-tier investors like Sequoia China, Tencent, and Deep Venture Capital involved.

    ➂ White Road Capital, founded in 2015, has invested in over 100 projects, with a focus on semiconductor and smart manufacturing.

    investmentsemiconductorstartups

January 8

  • Semiconductor Giant Reshuffles Staff, Postpones Salary Increase!

    微信公众号:半导体技术天地

    ➀ Renesas Electronics, a Japanese semiconductor manufacturer, plans to lay off hundreds of employees this year due to weak demand for chips.

    ➁ The company, which serves major automotive manufacturers like Toyota and Nissan, has canceled its salary increase plan for employees scheduled for this spring.

    ➂ Renesas expects its revenue to decline by 9% to approximately 1.33 trillion yen (US$ 13.4 billion) in 2024, with an operating profit margin expected to drop by 5 percentage points to 28.9%.

    Renesas Electronics
  • Canon's NIL Technology Competes with EUV Lithography

    IEEE Spectrum

    ➀ Canon has delivered the first commercial version of a technology that could one day revolutionize the manufacturing of the most advanced silicon chips.

    ➁ NIL (nanoimprint lithography) can draw circuit features as small as 14 nanometers, matching the level of processors produced by Intel, AMD, and NVIDIA.

    ➂ Canon's approach is different from the EUV lithography system produced exclusively by ASML of the Netherlands.

    EUV Lithography
  • 8 New SiC Collaborations: Guangqi/Shanghai Auto Orders, 8-inch Product Delivery

    Unknown

    ➀ Xilinx Integration and Guangqi Aion have signed a strategic cooperation agreement to establish a joint laboratory for automotive semiconductor technology research and development.

    ➁ XiZhi Technology has achieved a delivery milestone of over 20,000 units of its 800V platform high-end electric drive SiC-DCM power modules.

    ➂ Jingchi Machinery has successfully passed customer verification for its 8-inch SiC resistive crystal growth furnace.

    SiCTechnology Collaboration

January 5

  • Academician Huang Ru No Longer Serving as President and Deputy Secretary of the Communist Party Committee of Southeast University

    Not specified in the provided content

    ➀ Academician Huang Ru has ceased to serve as the President and Deputy Secretary of the Communist Party Committee of Southeast University;

    ➁ Huang Ru is a female Hui ethnic Chinese born in November 1969, from Nan'an, Fujian Province, a member of the Communist Party of China, with a doctorate in science, a professor, an academician of the Chinese Academy of Sciences, and an academician of the Academy of Sciences for the Developing World. She is a member of the Central Committee of the 20th Party Congress.

    ➂ Huang Ru has a bachelor's and master's degree from the Department of Electronic Engineering at Southeast University, and a doctorate from the Department of Computer Science and Technology at Peking University.

    Leadership Changeeducation
  • The Secret Weapon Behind Broadcom's Breakthrough to Trillion

    优分析

    ➀ Broadcom, with decades of experience in communication chips, is making a significant shift towards AI, with AI-related revenue accounting for 41% of its semiconductor business.

    ➁ Broadcom focuses on customized AI accelerators (XPUs) and related network connection technologies, avoiding direct competition with NVIDIA in the general AI acceleration chip market.

    ➂ Broadcom's XPU platform covers a full range of solutions from AI design process, AI IP, to SoC packaging, enabling rapid response to customer needs.

    AI ChipsBroadcom
  • CPU Market Gears Up for Two New Competitors

    icbank

    ➀ NVIDIA and MediaTek are expected to enter the PC CPU market this year, increasing competition.

    ➁ Intel's CEO acknowledges the increased competition and emphasizes the need to stay alert.

    ➂ Arm's CEO predicts that Arm-based CPUs will challenge Intel and AMD's x86 architecture in the coming years.

    ➃ Reports suggest that NVIDIA and MediaTek are developing new Arm architecture CPUs for PCs.

    ➄ The entry of these companies is driven by the rise of AI PCs and the growing demand for more efficient and powerful processors.

    ➅ Microsoft's role in encouraging the development of new CPU architectures for Windows PCs is highlighted.

    MediaTekNVIDIA