三星和SK海力士涨价离谱 据报道，三星和SK海力士已经开始向客户发布未来几个季度的合约价格，价格涨幅巨大。 公众号记得加星标⭐️，第一时间看推送不会错过。 DRAM 和 NAND 闪 存 半 导 体 价 格 继 续 同 步 上 涨 。 DRAM 价 格 再 次 创 下 历 史 新 高 ， 达 到 13 美 元 。 NAND价格也飙升超过33%。 根据市场研究公司 DRAMeXchange 2 月 27 日的数据，2 月份通用 PC DRAM 产品（DDR4 8Gb 1Gx8）的平均固定交易价格为 13.00 美元，比上个月的 11.50 美元上涨了 13.04%。 自去年4月以来，通用DRAM价格已连续11个月上涨。本月平均价格创下自2016年6月开始进行该项 调查以来的最高水平。 DRAMeXchange的母公司TrendForce宣布，PC DRAM价格较上一季度上涨110%-115%，远超去年 第四季度的涨幅（38%-43%）。TrendForce分析指出："大多数DRAM供应商和PC制造商已于2月份 完成了第一季度的价格谈判。"并补充道："由于大部分合同已经签订，预计3月份的价格水平将与2月 份持平。 ...

Core Insights - The market for non-volatile memory (NVM) is evolving due to advanced SoC design requirements, AI, new sensor technologies, and stringent quality standards, leading to increased exploration of alternative solutions [2] - Over 80% of respondents in a recent survey are using or evaluating embedded NVM technology, indicating a mature and active market with significant room for development among competing technologies [2] Group 1: Awareness and Familiarity - Embedded flash memory remains the dominant choice, with over 80% of respondents familiar with it, reflecting its long-standing position as the default option [4] - Awareness of alternative NVM technologies such as FRAM, MRAM, and ReRAM has increased, with recognition rates exceeding 25%, indicating these technologies are becoming mainstream [4] Group 2: Selection Criteria - Practicality is emphasized in the selection criteria for embedded NVM, with reliability, durability, and data retention being top priorities, all scoring above 3.0 on a 4.0 scale [5] - Process scalability and power efficiency also scored above 3.0, highlighting the challenges of extending traditional embedded NVM to advanced geometries [5] Group 3: Challenges and Pain Points - Key constraints for current NVM deployment include scalability limitations and power performance trade-offs, both scoring above 3.0, indicating they are critical issues, especially at advanced process nodes [6] - Reliability concerns and cost uncertainties follow closely, suggesting that long-term predictability and economic risks remain unresolved for many design solutions [6] Group 4: Market Trends and Future Outlook - The survey indicates that design pressures are increasing, with more teams evaluating alternatives not out of curiosity but due to critical constraints on scalability, power consumption, and long-term predictability [8] - The transition from traditional storage technologies to new NVM is entering a crucial phase, with a significant number of teams expected to make specific IP choices within the next year [9] - External forecasts suggest that the market for emerging embedded NVM could reach $3.3 billion by 2030, driven by the adoption of technologies like MRAM, PCM, and ReRAM in next-generation MCUs and SoCs [10]

Core Viewpoint - TSMC is expanding its production capabilities both domestically and internationally, with significant developments in its Arizona subsidiary, which turned a profit in 2025 after previous losses, while facing challenges in its Kumamoto plant [2][4]. Group 1: Financial Performance - TSMC's Arizona subsidiary reported a profit of 16.14 billion NTD in 2025, recovering from a cumulative loss of over 39.4 billion NTD from 2021 to 2024, where the 2024 loss alone was 14.298 billion NTD [4]. - The company received government subsidies totaling 76.258 billion NTD in 2025 and 75.164 billion NTD in 2024 for local operations, aimed at offsetting costs related to real estate, facilities, and production operations [2]. Group 2: Capacity Expansion - TSMC is accelerating capacity expansion in Arizona, with the first wafer fab having commenced production in 2024 and a second fab expected to begin operations in the second half of 2027 due to strong customer demand [5]. - In Taiwan, TSMC plans to build up to 10 new fabs across major science parks, focusing on advanced processes, including 2nm technology, with significant investments in Tainan and Kaohsiung [10][11]. Group 3: Customer Dynamics - TSMC's largest customer is projected to shift from Apple to Nvidia, with Nvidia expected to account for nearly 20% of TSMC's revenue in 2025, surpassing Apple's 17% [7][8]. - The demand for AI and high-performance computing is driving this shift, as Nvidia's revenue from AI accelerators has increased significantly, reflecting the changing landscape of the semiconductor market [8]. Group 4: Market Context - Despite TSMC's investments in the U.S., the slow construction timelines suggest that Taiwan will remain the primary hub for TSMC's advanced process and packaging investments [12].

公众号记得加星标⭐️，第一时间看推送不会错过。 出人意料的是，英特尔晶圆代工负责人凯文·奥巴克利（Kevin O'Buckley）在公司任职仅两年后便决 定跳槽至高通。今后，英特尔晶圆代工将由纳加·钱德拉塞卡兰接任，他此前负责前端工艺技术开发 和制造。 英特尔发言人告诉Tom's Hardware："我们感谢凯文·奥巴克利对晶圆代工服务的贡献，并祝愿他在公 司外寻求发展机会时一切顺利。英特尔晶圆代工仍然是英特尔最重要的战略重点之一，在纳加·钱德 拉塞卡兰的领导下，该部门专注于严谨的执行和为客户提供优质服务。" 作为首席技术和运营官以及英特尔晶圆代工负责人，Chandrasekaran 现在负责监督先进工艺技术的 开 发 以 及 英 特 尔 全 球 制 造 网 络 的 日 常 运 营 ， 这 是 英 特 尔 此 前 所 缺 乏 的 职 位 ， 因 为 在 他 之 前 的 O'Buckley 和 Stuart Pann 都负责日常运营和扩张工作，但并不负责先进制造技术的开发。 相比之下，钱德拉塞卡兰负责下一代硅逻辑节点、先进封装解决方案和测试技术的研发和推广，同时 还监管全球范围内的前端晶圆制造和后端组装封装业务 ...

ASML控股公司的一位高管告诉路透社，下一代芯片制造机已经准备就绪，制造商可以开始将其投入 大规模生产，这对芯片行业来说是一大步。 这家荷兰公司生产全球唯一商用极紫外光刻（EUV）设备，这是芯片制造商的关键设备。这款新设备 将帮助台积电以及英特尔等芯片制造商.数据显示，通过消除芯片制造过程中几个成本高昂且复杂的 步骤，下一代EUV光刻机可以生产出功能更强大、效率更高的芯片。 ASML首席技术官马可·皮特斯周三告诉路透社，ASML计划于周四在圣何塞举行的技术会议上发布这 些数据，这代表着一个重要的里程碑。 ASML 花费数年时间开发出昂贵的下一代工具，因为芯片制造商一直在努力确定何时开始将其用于 大规模生产才具有经济意义。 但鉴于当前一代 EUV 工具在制造复杂 AI 芯片方面已接近技术极限，下一代机器（称为高数值孔径 EUV 工具）对于 AI 行业至关重要，可以改进 OpenAI 的 ChatGPT 等聊天机器人，并帮助芯片制造 商按时交付其 AI 芯片路线图，以满足激增的需求。 新工具造价约为 4 亿美元，是原 EUV 机器造价的两倍。 据皮特斯称，ASML的数据显示，高数值孔径极紫外光刻机（High-NA ...

Core Viewpoint - The global semiconductor market is projected to reach $792 billion by 2025, marking a 25.6% growth compared to 2024, driven primarily by the surge in artificial intelligence (AI) demand [2] Group 1: Market Growth and Drivers - The growth in the semiconductor market is the strongest since 2021, with AI being the main driver, exemplified by Nvidia's revenue growth of 65% [2] - Major memory manufacturers, including Samsung, SK Hynix, and Micron, reported a collective revenue increase of 29%, attributing this to AI [2] - In Q4 2025, memory companies are expected to see revenue growth between 21% and 34%, while Nvidia anticipates a 20% revenue increase [2][3] Group 2: Company Performance - Nvidia's revenue for Q4 2025 is projected at $68.1 billion, with a 20% increase compared to the previous quarter [3] - Samsung's revenue is expected to reach $30 billion, reflecting a 33% increase, driven by AI [3] - Micron anticipates a 21% revenue increase, with a strong demand for AI driving price increases [3] Group 3: Future Outlook and Challenges - For Q1 2026, Micron expects a significant revenue increase of 37%, while SanDisk anticipates a 52% growth, and Kioxia expects a 64% increase [4] - However, companies like Intel, Qualcomm, and AMD are forecasting revenue declines due to seasonal factors and memory shortages impacting the PC and smartphone markets [4][5] - IDC predicts a decline in smartphone and PC shipments in 2026, with a potential drop of 0.9% to 2.4% [5] Group 4: Predictions for 2026 - Forecasts for 2026 growth rates vary, with lower estimates around 9.5% to 12%, while higher estimates range from 23% to 30% [6] - The semiconductor market is expected to continue benefiting from strong AI growth, with projections indicating over 20% growth for the year despite challenges in the smartphone and PC sectors [9]

Core Viewpoint - Despite Nvidia's strong quarterly performance and guidance, its stock price has declined, attributed to profit-taking and market rotation concerns [2][5][7]. Group 1: Financial Performance - Nvidia reported a 73% year-over-year revenue growth, with Q4 revenue reaching $68 billion, exceeding market expectations by nearly $2 billion [3][12]. - The adjusted earnings per share for the quarter were $1.62, surpassing expectations by $0.8 [3]. - Nvidia's revenue forecast of $78 billion for the upcoming quarter is higher than most investors' expectations, indicating strong demand [3]. Group 2: Analyst Ratings and Predictions - Goldman Sachs maintains a "buy" rating for Nvidia, citing increased capital expenditures from hyperscale data center operators and rising chip procurement from non-traditional buyers [2]. - Cantor Fitzgerald reaffirms a "buy" rating with a target price of $300, highlighting a $3.1 billion revenue beat in Q4 [4]. - UBS and TD Cowen also maintain "buy" ratings, emphasizing strong demand and an improving capital expenditure environment [3][4]. Group 3: Market Concerns - Concerns about the sustainability of capital expenditures from hyperscale data center operators have been raised, contributing to Nvidia's stock price decline [2][5]. - The stock has faced selling pressure due to profit-taking and a shift in market focus from software to semiconductors [2][5]. - There are worries regarding Nvidia's partnership with OpenAI, with uncertainty about the completion of a $100 billion investment and collaboration agreement [7]. Group 4: Competitive Landscape - Analysts suggest that Nvidia's dominance in AI may be challenged as the industry shifts from training-intensive to inference-driven workloads, increasing competition from other chip manufacturers [9]. - Concerns have been expressed about Nvidia's ability to maintain its competitive edge in the evolving computing landscape [9]. Group 5: Investor Sentiment - Investor sentiment appears cautious, with many analysts noting that Nvidia's stock price reflects high expectations, leading to potential volatility in the coming quarters [5][9]. - The Philadelphia Semiconductor Index has seen a decline, with major AI and semiconductor stocks, including Nvidia, experiencing significant drops [5].

Core Viewpoint - The demand for Google's Tensor Processing Units (TPUs) is expected to be strong, leading to significant growth in the High Bandwidth Memory (HBM) supply chain, with Google projected to become a key demand source in the HBM market, second only to NVIDIA [2] Group 1: Market Dynamics - Bank of America has raised its forecast for Google's TPU shipments from 4 million to 4.6 million units for this year, which is double the previous prediction of 2.3 million units for 2025 [2] - It is anticipated that Google's share in the HBM market may exceed 30% this year, indicating structural growth in the Application-Specific Integrated Circuit (ASIC) market [2] - The shift towards ASICs like TPUs, optimized for specific tasks such as training and inference, is becoming crucial in the evolving AI landscape [2] Group 2: HBM Supply Chain Competition - Samsung Electronics and SK Hynix are competing to maintain their dominance in the HBM supply chain, with both companies supplying HBM3E to Google and working on HBM4 and HBM4E [3] - Samsung has officially announced shipments of HBM4, while SK Hynix has begun mass production of HBM4 and is optimizing with major clients [3] - There are predictions that Google may skip HBM4 and directly adopt HBM4E, which could accelerate the development of HBM4E by Samsung and SK Hynix [3] Group 3: Diversification of Demand - The demand for accelerators is diversifying, with AMD signing a contract to provide Meta with AI accelerators, which is expected to further broaden the sources of HBM demand [4] - Since last year, Samsung has been supplying 12-layer HBM3E products to AMD's flagship AI accelerator MI350 series, indicating a growing diversification in HBM demand sources [4] - The diversification of HBM demand sources is seen as positive for the memory industry, enhancing the bargaining power of Korean manufacturers [5]

Group 1 - The Japanese government will become the largest shareholder of domestic chip manufacturer Rapidus, holding approximately 10% of voting rights, with the option to increase its stake to over 50% if the company faces difficulties [2] - The government will invest 100 billion yen (approximately 641 million USD) through the Japan Information Technology Promotion Agency, with an additional 150 billion yen planned for the fiscal year 2026, bringing the total investment to 250 billion yen [2] - Private investments from over 30 companies are expected to exceed 160 billion yen, raising Rapidus's total capital to approximately 420 billion yen, with the government contributing about 60% of this amount [2] Group 2 - The Japanese government plans to provide a total support of approximately 3 trillion yen, including R&D subsidies, and will have a voting power slightly higher than the largest private shareholder [3] - Special category shares without voting rights can be converted into voting shares if the company goes bankrupt, allowing the government to increase its voting power to about 60% [3] - The golden share will grant the government veto power over significant management decisions, such as the appointment and dismissal of directors, approval of share transfers, and mergers and alliances [3] Group 3 - Concerns have been raised about the potential lack of accountability in public-private partnerships, emphasizing the importance of clear leadership between the private sector and the government [4] - Experts suggest that even with government-supported policies, effective control over economic or corporate operations ultimately depends on the companies' own decision-making [4]

Core Viewpoint - Broadcom has announced the delivery of the industry's first 2nm custom computing SoC based on its 3.5D XDSiP platform, which combines 2.5D technology and face-to-face (F2F) 3D integration, marking a significant advancement in semiconductor integration technology [2][3]. Group 1: Technology and Innovation - The 3.5D XDSiP platform is a modular multi-dimensional stacked chip platform that supports independent scaling of compute, memory, and network I/O within a compact form factor, enabling efficient low-power computing for massive AI workloads [2]. - Broadcom's XDSiP technology integrates 2nm process innovations with face-to-face 3D integration, providing unprecedented computing density and energy efficiency, crucial for next-generation AI and high-performance computing [3]. - The design of Broadcom's XDSiP utilizes hybrid copper bonding (HCB) technology, optimizing inter-chip communication and significantly enhancing interconnect speed while reducing signal routing distances [10]. Group 2: Market and Strategic Partnerships - Fujitsu is the first company to publicly adopt Broadcom's 3.5D XDSiP technology, which is a key driver for its FUJITSU-MONAKA project aimed at delivering high-performance, low-power processors [3][11]. - Broadcom expects to sell at least 1 million chips based on its stacked design technology by 2027, representing a potential revenue source worth billions of dollars [15]. - The company has collaborated with major tech firms like Google and OpenAI to develop custom processing units, contributing to significant growth in its AI chip business, which is projected to double year-over-year [16]. Group 3: Competitive Landscape - Broadcom's 3.5D XDSiP technology aims to compete with existing solutions from AMD and Intel, as it seeks to establish a standardized platform for building multi-chip processors [13]. - The company is positioned as a significant competitor to NVIDIA and AMD in the AI chip market, with its technology allowing clients to build more powerful and energy-efficient chips to meet the growing computational demands of AI software [15][16]. - Broadcom's initial design is similar to AMD's MI300X but is open for licensing to any interested parties, indicating a strategic move to broaden its market reach [8][13].