公众号记得加星标⭐️，第一时间看推送不会错过。 台积电悄然宣布已开始量产采用其N2（2nm级）工艺的芯片。该公司并未发布正式新闻稿宣布量产启 动，但此前曾多次表示N2工艺有望在第四季度实现量产，因此该计划已达成。 台积电在其 2nm 技术网页上发表声明称："台积电的 2nm (N2) 技术已按计划于 2025 年第四季度开 始量产。 " N2工艺采用第一代纳米片晶体管技术，在性能和功耗方面实现了全节点提升。台积电还开发了低电 阻重分布层（RDL）和超高性能金属-绝缘体-金属（MiM）电容器，以进一步提升性能。 台积电N2技术在密度和能效方面都将成为半导体行业最先进的技术。凭借领先的纳米片晶体管结 构，N2技术将带来全节点性能和功耗优势，以满足日益增长的节能计算需求。通过我们持续改进的 战略，N2及其衍生技术将进一步巩固台积电的技术领先地位，并使其在未来很长一段时间内保持领 先地位。" 从性能提升的角度来看，N2 的设计目标是在相同功耗下实现 10%–15% 的性能提升，在相同性能下 降低 25%–30% 的功耗，并且对于包含逻辑、模拟和 SRAM 的混合设计，晶体管密度比 N3E 提高 15%。对于纯逻辑设计 ...

公众号记得加星标⭐️，第一时间看推送不会错过。 中芯国际重磅收购，迎来新进展。 12月29日晚间，中芯国际公告称，公司拟向国家集成电路基金等5名中芯北方股东发行5.47亿股股份 购买其所持有的中芯北方49%股权，交易价格406.01亿元。公告显示，这五名股东分别为国家集成电 路基金、集成电路投资中心、亦庄国投、中关村发展、北京工投。 据悉，标的公司中芯北方成立于2013年，是中芯国际与北京市政府共同投资设立的12寸先进制程集 成电路制造厂，是中芯国际控股子公司。公司主要从事半导体（硅片及各类化合物半导体）集成电路 芯片的制造（含线宽28纳米及以下大规模数字集成电路制造）、针测及测试、光掩膜制造、测试封 装，是中芯国际重要的12英寸晶圆厂。 中芯国际作为中芯北方的控股股东与核心技术供给方，全面主导并负责中芯北方的生产运营管理，为 其技术落地与产能释放提供了坚实保障。 对于此次收购，公司表示，本次交易有利于进一步提高上市公司资产质量、增强业务上的协同性，促 进上市公司的长远发展。交易前后上市公司的主营业务范围不会发生变化。 公开资料显示，中芯国际拥有全方位一体化的集成电路晶圆代工核心技术体系，快速有效地帮助客户 ...

Core Viewpoint - NVIDIA's acquisition of Intel shares for $5 billion has turned into a strategic financial operation, now valued at $7.58 billion, following the approval from the Federal Trade Commission [1] Group 1: Acquisition Details - NVIDIA CEO Jensen Huang and Intel CEO Pat Gelsinger agreed on a purchase price of $23.28 per share for Intel stock, totaling 214 million shares [1] - The acquisition was completed on December 26, with Intel's stock closing at $36.68 per share on the following Monday [1] Group 2: Joint Development Agreement - NVIDIA and Intel will collaborate on developing "multi-generation" chips for data centers and personal computers, aiming to capture market share across their customer base [1] - Intel will customize x86 CPUs for NVIDIA, which will integrate these CPUs into its AI infrastructure platform [1] Group 3: Technical Advancements - The partnership will utilize NVIDIA's NVLink technology, providing a bandwidth of up to 1.8 TB/s per GPU, which is 14 times the bandwidth of PCIe 5.0 x16 slots [1] - Intel will also manufacture x86 system-on-chip (SoC) integrated with NVIDIA's RTX GPU chips, enhancing the performance of integrated CPU and GPU PCs [2] Group 4: Regulatory Context - The agreement is reminiscent of NVIDIA's previous attempt to acquire Arm for $40 billion, which faced regulatory scrutiny and was ultimately abandoned [2] - The Federal Trade Commission had previously expressed concerns that such mergers could stifle competition and innovation in the semiconductor industry [2]

公众号记得加星标⭐️，第一时间看推送不会错过。 据报道，Weebit Nano 已将其 ReRAM 技术授权给 TI。 根据协议条款，Weebit 的 ReRAM 技术将集成到 TI 的先进嵌入式处理器工艺节点中。 该协议包括知识产权许可、技术转让、Weebit ReRAM 在 TI 工艺技术中的设计和认证。 Weebit ReRAM 是一款低功耗、高性价比的非易失性存储器，声称在高温下具有优异的保持性能， 并已通过 AEC-Q100 150°C 运行认证。 德州仪器 (TI) 嵌入式处理高级副总裁 Amichai Ron 表示： "我们很高兴与 Weebit Nano 合作，将 ReRAM 存储器技术集成到我们的工艺技术和产品中。TI 与 Weebit Nano 的合作将使我们的客户能 够获得业界领先的非易失性存储器 (NVM) 技术，该技术在性能、规模和可靠性方面均具有优势，这 将有助于我们巩固作为领先的嵌入式处理器供应商的地位。" " 我 们 很 高 兴 能 与 Weebit Nano 合 作 ， 将 ReRAM 存 储 器 技 术 集 成 到 我 们 的 工 艺 技 术 和 产 品 中，"Weebit ...

Core Viewpoint - TSMC is set to increase advanced process pricing from 2026 to 2029 due to high demand driven by AI applications, with a projected price increase of 3% to 10% for advanced processes starting January 2026 [1][2][3]. Group 1: Pricing Strategy - TSMC has communicated with clients about raising prices for advanced processes over the next four years, reflecting increased production costs and high demand [1][3]. - Despite the price increase, clients are actively reserving advanced process capacity, indicating strong demand in the AI sector [2][3]. - The expected price increase for 2026 is in the single-digit percentage range, with variations based on client procurement levels [1][3]. Group 2: Capital Expenditure - TSMC's capital expenditure for 2026 is projected to reach a historical high of between $42 billion and $45 billion, maintaining a strong investment trend [3]. - The company has already reported a capital expenditure of $29.39 billion for the first three quarters of 2023, with expectations for the fourth quarter to reach between $10.61 billion and $12.61 billion [3]. - Global semiconductor manufacturing capital expenditure is estimated to reach $160 billion in 2023, primarily supported by TSMC and Micron's investments [4]. Group 3: Market Dynamics - The semiconductor industry is experiencing a seller's market for advanced processes, with rising costs reflected in pricing strategies across various segments, including wafer foundry and advanced packaging [5][6]. - The strong demand for AI chips positions TSMC at the core of the AI market, enhancing its bargaining power [6]. - The anticipated price increases in advanced packaging and memory also indicate a trend towards "chip inflation," which may impact consumer electronics demand [5][6].

Core Insights - The recent semiconductor technology roadmap by the Korean Society of Semiconductor Engineers (ISE) predicts advancements down to 0.2nm, indicating a shift in the competitive landscape of the semiconductor industry rather than just a focus on smaller process nodes [1][3]. Device and Process Technology Roadmap - The roadmap outlines a 15-year vision from 2025 to 2040, focusing on nine key semiconductor technology trends, including AI semiconductors, optical interconnects, and quantum computing [1][3]. - The evolution of logic devices aims to maintain performance and power efficiency while reducing size, with a shift from Design-Technology Co-Optimization (DTCO) to System-Technology Co-Optimization (STCO) [4][5]. Logic Technology Trends - Logic device nodes are projected to progress from 2nm in 2025 to 1nm by 2031, and approach 0.2nm by 2040, with key variables including gate length and 3D integration capabilities [5][12]. - The transition from FinFET to Gate-All-Around (GAA) structures is expected, with further innovations like CFET (Complementary FET) enhancing performance through 3D stacking [8][10]. Metal Interconnect Technology - Metal interconnects are becoming a critical performance bottleneck, necessitating innovations in materials and processes to achieve lower resistance and higher reliability [14][15]. - Backside Power Delivery Networks (BSPDN) are anticipated to be introduced around 2028, improving power efficiency and area utilization [14][15]. Memory Technology Trends - The semiconductor industry is shifting focus from computation to memory, with AI and HPC driving demand for high-capacity, high-bandwidth, low-latency, and low-power memory solutions [16][17]. - DRAM technology is evolving towards vertical channel transistors and stacked architectures, with significant advancements expected in the coming years [19][21]. Non-Volatile Memory (NVM) Developments - 3D NAND technology is projected to increase in layer count, reaching up to 2000 layers by 2040, while facing challenges in manufacturing processes [23][25]. - New NVM technologies like PCM and ReRAM are being explored, with PCM seen as having balanced scaling potential [26][27]. AI Semiconductor Roadmap - The AI hardware market is expected to grow significantly, with AI-related computing projected to account for about 20% of global computing demand by 2025 [29][30]. - Performance for training and inference hardware is expected to improve dramatically, with TOPS/W metrics increasing significantly by 2040 [30][31]. Optical Interconnect Technology - Optical interconnects are viewed as a key solution to the limitations of traditional copper interconnects, with applications expanding in data centers and AI-driven systems [33][36]. - The roadmap anticipates the introduction of Co-Packaged Optics (CPO) technology, which integrates optical and electronic components to enhance data transmission capabilities [42][44].

根据新闻稿，智能集成电路基板提供集成电路芯片与外部读卡器之间的物理和电气接口。当智能卡插 入ATM机或被护照读卡器扫描时，电子信号通过基板表面形成的触点进行传输。该公司指出，传统 上，为了确保低接触电阻、耐腐蚀性和机械耐久性，需要在基板表面镀上金、钯等贵金属。 公众号记得加星标⭐️，第一时间看推送不会错过。 LG Innotek宣布开发出全球首款"下一代智能IC基板"，该基板是信用卡、电子护照和USIM卡等智能 卡的核心组件。该公司表示，这款新型基板在提高产品耐用性的同时，还能将生产过程中产生的碳排 放量减少一半。 这项公告意义重大，因为它重点介绍了一种材料和封装方面的创新，该创新直接影响信号可靠性、长 期磨损以及 ESG 合规性。这些都是从事安全集成电路、卡接口和大批量电子产品制造的工程师需要 考虑的关键因素。 LG Innotek推出全球首款"下一代智能IC基板" LG Innotek于 12 月 22 日宣布，该公司已成功开发出"下一代智能 IC（集成电路）基板"，该基板具 有更高的性能，并且生产过程中产生的碳排放量减少了一半。 智能IC基板是安装IC芯片的关键组件，这些芯片用于存储智能卡（例如信用卡、 ...

Core Viewpoint - TSMC is set to benefit from the booming AI applications, leading to a tight supply of advanced process capacity below 3nm, with plans to raise prices for advanced processes from 2026 to 2029 [1][2] Group 1: Price Increase and Demand - TSMC has communicated with clients about a price increase for advanced processes, expected to be in the single-digit percentage range starting January 2026, reflecting higher production costs and strong demand [1][3] - Analysts predict that TSMC's advanced process pricing could increase by 3% to 10% in 2026, with variations depending on client procurement levels [1][2] Group 2: Capacity and Client Engagement - Major clients like NVIDIA and AMD are actively expanding AI applications, contributing to the sustained demand for TSMC's advanced processes [1][2] - TSMC's clients are eager to secure advanced process capacity to maintain a competitive edge in technology [2] Group 3: Capital Expenditure - TSMC's capital expenditure for 2026 is projected to reach a historical high of between $42 billion and $45 billion, maintaining a flexible approach to ensure revenue growth outpaces capital spending [3] - The actual capital expenditure for TSMC in the first three quarters of this year reached $29.39 billion, with expectations for the fourth quarter to contribute an additional $10.61 billion to $12.61 billion [3] Group 4: Industry Context - Global semiconductor manufacturing capital expenditure is estimated to reach $160 billion this year, with a 3% year-over-year growth, primarily driven by TSMC and Micron's significant investments [4]

公众号记得加星标⭐️，第一时间看推送不会错过。 为满足人工智能（AI）普及带来的存储芯片需求增长，三星电子正计划扩大其位于平泽的半导体工厂 的产能。在决定恢复平泽5号生产线（P5）的建设后，该公司已启动关键基础设施的招标程序，并且 有迹象表明，平泽4号生产线（P4）的设备引进时间正在提前。 29日，业内人士透露，三星电子正积极推进平泽P5工厂气体和化学品供应设备的招标流程。气体和 化学品工艺设备是生产线建设中至关重要的组成部分。通常情况下，设备订单会在框架搭建完成后才 下达，但此次为了加快批量生产进度，该公司据称将采取"快速通道"策略，即框架搭建、设备订购和 安装工作同步进行。 具体来说，P5是三星电子在平泽第二园区正在推进的一条新生产线。内部决策机构近期决定启动框架 建设，目标是在2028年投产。然而，鉴于电子和IT行业正面临存储芯片严重短缺的困境，且预计从 明年开始供应短缺情况将进一步恶化，因此有人预期P5的投产时间可能会提前。 据报道，全球工业气体公司和韩国韩国设备企业正准备参与竞标。潜在竞标者包括林德集团、液化空 气集团、汉阳气工（默克）、沃尼克控股、汉阳工程和STI。总投资额预计至少数千亿韩元，最高 ...

Core Viewpoint - NVIDIA aims to dominate the inference stack with its next-generation Feynman chip by integrating LPU units into its architecture, leveraging a licensing agreement with Groq for LPU technology [1][18]. Group 1: NVIDIA's Strategy and Technology Integration - NVIDIA plans to integrate Groq's LPU units into its Feynman GPU architecture, potentially using TSMC's hybrid bonding technology for stacking [1][3]. - The LPU modules are expected to enhance inference performance significantly, with Groq's LPU set to debut in 2028 [5]. - The Feynman core will utilize a combination of logic and compute chips, achieving high density and bandwidth while maintaining cost efficiency [6]. Group 2: Inference Market Dynamics - The AI industry's computational demands have shifted towards inference, with major companies like OpenAI and Google focusing on building robust inference stacks [9]. - Google’s Ironwood TPU is positioned as a competitor to NVIDIA, emphasizing the need for low-latency execution engines in large-scale data centers [9][10]. - Groq's LPU architecture is designed specifically for inference workloads, offering deterministic execution and on-chip SRAM for reduced latency [10][14]. Group 3: Licensing Agreement and Market Position - NVIDIA's agreement with Groq is framed as a non-exclusive licensing deal, allowing NVIDIA to integrate Groq's low-latency processors into its AI Factory architecture [18][21]. - This strategy is seen as a way to circumvent antitrust scrutiny while acquiring valuable talent and intellectual property from Groq [19][21]. - The transaction is viewed as a significant achievement for NVIDIA, positioning LPU as a core component of its AI workload strategy [16][21].