Core Viewpoint - The global AI technology is shifting from training to inference, driving a new growth opportunity in the hardware supply chain [2] Summary by Category Overall - The transition from training-dominated AI to inference-driven AI is significantly increasing the demand for inference computing power, driven by the iteration of multimodal large models like Google's Gemini 3 Pro and OpenAI's Sora 2 [2] - Major cloud service providers (CSPs) are expected to increase capital expenditures, with a forecast of $431 billion by 2025, a 65% year-on-year increase, and potentially reaching $602 billion by 2026 [2] - Sovereign AI initiatives are being launched globally, such as the U.S. "Gateway to the Stars" plan with an investment of approximately $500 billion and the EU's plan to invest $21.5 billion in AI super factories, contributing to a high-growth phase in global AI infrastructure [2] - By 2030, global AI data center capacity is projected to reach 156 GW, accounting for 71% of total data center demand [2] Cloud Side - PCB: AI servers are bringing clear value increases, with Nvidia's DGX H100 single GPU corresponding to a PCB value of $211, a 21% increase from the previous generation; the GB200 NVL72 raises the single GPU value to $346 [3] - The domestic high-end PCB capacity is expected to be released in 2026 to support downstream demand, driving upgrades in upstream materials [3] - Storage: The structural supply-demand imbalance due to AI demand has led to significant price increases in DRAM and NAND Flash, with a shift in investment focus towards high-value products expected in 2026 [3] - KVCache technology is accelerating the replacement of HDDs with QLC SSDs, with a projected 30% penetration rate in the enterprise SSD market by 2026 [3] Optical Interconnect - Optical interconnect technology is entering a new era as a key component of AI computing clusters, with optical switches meeting the interconnection needs of large-scale AI clusters due to their high bandwidth, low latency, and low power consumption [4] - The MEMS-based technology route currently dominates, with domestic manufacturers actively engaging in various segments of the global supply chain [4] End Side - AI Phones: The AI phone market is expected to maintain moderate growth in 2025, with competition shifting towards end-side AI capabilities [5] - The operating systems of mobile phones are evolving from "application launchers" to "system-level intelligent agents," with flagship chips from Apple and Android continuously enhancing NPU computing power [5] - AR Glasses: The integration of AI and AR in smart glasses is seen as the future of wearable devices, with the market experiencing rapid growth [5] - The optical imaging module solutions for AR glasses are expected to favor light guide technology due to its advantages in clarity and size, while LCOS remains the mainstream for consumer products [5] Recommendations - The company suggests focusing on sectors benefiting from the shift to inference computing and hardware upgrades, including: - PCB and upstream materials: Shenghong Technology, Huitian Technology, Jingwang Electronics, Guanghe Technology, Dongcai Technology [6] - Storage and equipment: Beijing Junzheng, Zhaoyi Innovation, Jucheng Co., Jingzhida [6] - Optical interconnect: Yintan Zhikong, Saiwei Electronics [6] - End-side AI: GoerTek, Luxshare Precision, Baiwei Storage, Longqi Technology, Crystal Optoelectronics, Zhongke Lanyun, Howey Group, Sunny Optical Technology [6]

OpenAI发言人证实了这一调整，称基于用户反馈，免费及Go用户更偏好默认的流畅聊天体验。尽管 OpenAI仍计划在技术完善后重新推出该功能，且付费用户仍可使用，但此次回撤表明，科技巨头在将 高成本、高延迟的尖端模型整合进大规模消费级产品时，正面临"性能与速度"博弈的严峻挑战。 OpenAI悄然逆转了其针对大众用户的核心产品策略，撤回了旨在提升回答精度的自动"模型路由器"系 统，转而通过更快的响应速度来挽留用户。这一举措突显了在竞争激烈的AI聊天机器人市场中，先进 的"推理"能力与即时响应的消费者需求之间存在着难以调和的矛盾。 据WIRED报道，OpenAI已在免费版及每月5美元的订阅层级中取消了"模型路由器"功能，用户将默认使 用响应速度更快、服务成本更低的GPT-5.2 Instant模型。此前，该系统旨在自动分析用户提问，并将复 杂问题分流至更强大但处理较慢的"推理"模型，而现在普通用户必须手动切换才能使用这些高级功能。 知情人士透露，这一逆转主要源于其对业务指标的冲击：路由系统的延迟对日活跃用户产生了负面影 响。尽管推理模型代表了AI性能的前沿，但其处理复杂问题时可能耗时数分钟，导致不愿等待的大众 用 ...

Core Insights - Apple is deepening its "vertical integration" strategy by developing its first self-designed AI server chip, codenamed "Baltra," in collaboration with Broadcom, aiming to reduce reliance on Nvidia chips [1][2] - The "Baltra" chip is specifically designed for "AI inference," focusing on executing tasks rather than training large-scale AI models, with Apple opting to rent Google's custom Gemini model for $1 billion annually [1] - The architecture of "Baltra" will differ significantly from traditional training chips, emphasizing low latency and high concurrent throughput, with a focus on optimizing low-precision mathematical operations to reduce energy consumption and enhance user response speed [2] Company and Industry Summary - The collaboration with Broadcom is crucial for overcoming core network transmission technology challenges, with the chip expected to be operational by 2027 [1] - The chip is likely to utilize TSMC's advanced 3nm "N3E" process, with design work anticipated to be completed within the next 12 months [2] - The strategic focus on inference rather than training aligns with industry trends towards optimizing AI performance for user-facing applications [2]

摩根大通预计存储芯片迎史上最长景气周期：头部厂商市值今年逼近万亿美元，2027年将飙至1.5万亿，涨幅超50%。 HBM需求持续挤占传统DRAM产能，AI推理对内存消耗是训练3倍，供需缺口将延续至2027年。预测2026财年DRAM价 格暴涨53%，企业级市场强劲将完全抵消消费端压力。 摩根大通在最新研报中指出，当前头部存储芯片制造商的总市值已接近1万亿美元。基于历史估值中枢推算，到2027年这 一数字将飙升至1.5万亿美元，意味着头部厂商仍有超过50%的上涨空间。 市场正在经历"双轨制"定价。B2B（企业级/AI）需求强劲支撑价格高位，而B2C（消费级）则面临周期性压力。但整体而 言，服务器端的需求上行将完全抵消消费端的下行风险。 估值重塑：向1.5万亿美元进军 摩根大通在报告中直击投资者痛点： 在存储股过去三个月大幅上涨并逼近1万亿美元市值大关后，下一步怎么走？ 摩根大通给出的答案非常明确： 继续做多。 基于"市值/市场规模（TAM）"的估值框架，摩根大通预测2027年存储市场规模将达到约4200亿美元。取2018年和2021 年周期的市销率（P/S）中值3.5倍计算，头部存储及内存制造商的合计市值有望 ...

软件ETF（515230）跟踪的是软件指数（H30202），该指数从市场中选取涉及软件开发、销售和服务等 业务的上市公司证券作为指数样本，覆盖操作系统、应用软件、网络安全等领域的代表性企业，以反映 软件行业相关上市公司证券的整体表现。该指数具有显著的成长性和技术导向性，能够较好地体现软件 行业的市场趋势和发展动态。 东吴证券表示，AI推理要求小数据块、高并发和大存储容量，现有以CPU为中心的传统架构难以满足需 求。因此，提出将控制和数据路径移至GPU，通过GPU直接连接SSD来提高存储量和传输速率，并利用 SCADA软件架构控制存储I/O。这促使底层硬件变革带动软件重构，形成以GPU为中心的新架构。预计 2027年初将推出首款搭载HBF技术的AI推理系统，同时Hammerspace和CloudianHyperStore已开始优化软 件性能。数据库产业有望因GPU直连SSD架构的创新迎来新的发展机遇。 （文章来源：每日经济新闻） ...

Core Viewpoint - Morgan Stanley's latest report indicates that the total market capitalization of leading storage chip manufacturers is approaching $1 trillion, with projections suggesting it could rise to $1.5 trillion by 2027, indicating over 50% upside potential for these companies [1][4]. Market Dynamics - The current cycle is expected to be the longest and strongest storage upcycle in history, with concerns about oversupply in DRAM due to new capacities in 2027 deemed unnecessary by Morgan Stanley's data models [2][5]. - The market is experiencing a "dual-track" pricing system, where strong B2B (enterprise/AI) demand supports high prices, while B2C (consumer) faces cyclical pressures [2][9]. Valuation Insights - Morgan Stanley predicts that the storage market size will reach approximately $420 billion by 2027, with a projected price-to-sales (P/S) ratio of 3.5x based on historical cycles, leading to a potential market capitalization of $1.5 trillion for leading storage manufacturers [3][4]. Supply and Demand Analysis - Despite concerns about new wafer fabs leading to oversupply, Morgan Stanley's analysis suggests that DRAM supply growth will lag behind demand growth due to structural demand from AI and HBM [5][6]. - The supply-demand gap is expected to remain, with a projected shortfall of 3% to 5% in 2027, despite new capacities coming online [9]. Pricing Trends - Strong CSP (Cloud Service Provider) demand is expected to lead to a significant allocation of capacity to HBM, increasing its share of total DRAM capacity from 19% in 2025 to 28% in 2027 [9]. - DRAM average selling prices (ASP) are projected to surge by 53% in FY26, with NAND ASP increasing by approximately 30% [18]. AI and Technology Drivers - The rise of AI is driving demand for HBM and enterprise SSDs (eSSD), with AI servers requiring three times the SSD capacity of regular servers [17]. - HBM supply is expected to remain constrained, with a projected shortfall of 8% to 12% continuing through 2027 and possibly into 2028 [19]. Capital Expenditure Trends - While storage manufacturers are announcing capacity expansion plans, actual bit supply growth is expected to be challenged by physical migration issues [24]. - Equipment spending for storage wafer fabs is anticipated to grow significantly, outpacing overall capital expenditure growth [24].

Group 1 - The China Securities Digital Economy Theme Index decreased by 0.71% as of December 15, 2025, with key stocks like Zhongwei Company leading gains, while Nasta and others faced declines [1] - The capital expenditure of the four major overseas Cloud Service Providers (CSPs) reached $97.9 billion in Q3 2025, reflecting a quarter-on-quarter increase of 10%, indicating a continued upward trend in investment [1] - Domestic CSPs are still in a catch-up phase compared to their overseas counterparts, although leading firms like ByteDance are approaching the scale of Google in terms of token call volume and business size [1] Group 2 - Domestic GPU companies have made significant technological breakthroughs, achieving single-chip support for AI computing and graphics rendering, although they still lag behind international leaders in technology accumulation and ecosystem development [2] - The digital economy ETF closely tracks the China Securities Digital Economy Theme Index, selecting companies with high digital infrastructure and application levels to reflect the overall performance of the digital economy theme [2] - As of November 28, 2025, the top ten weighted stocks in the China Securities Digital Economy Theme Index accounted for 54.6% of the index, with companies like Dongfang Wealth and Zhongxin International among the leaders [2][3]

Group 1 - The A-share market saw a collective rise in the three major indices, with the Shanghai Composite Index increasing by 0.17%, driven by strong performances in the electronics, communications, and defense sectors, while the comprehensive and retail sectors lagged behind [1] - The chip technology stocks performed well, with the chip ETF (159995.SZ) rising by 1.72%, and notable individual stocks such as Tuojing Technology up by 9.37%, Longxin Zhongke up by 6.42%, Beijing Junzheng up by 5.63%, and Haowei Group up by 4.48% [1] Group 2 - The global demand for AI inference is rapidly increasing, leading overseas Cloud Service Providers (CSPs) to significantly increase their capital expenditures on computing infrastructure, with a total capital expenditure of $97.9 billion in Q3 2025, reflecting a quarter-on-quarter increase of 10% [3] - Domestic CSPs are still in a catch-up phase compared to their overseas counterparts, but leading domestic firms like ByteDance are approaching the scale of Google in terms of token call volume and business size [3] - Domestic advanced process expansion is steadily progressing, and the acceleration of self-controllable advancements in the industry chain is expected to significantly enhance the supply capacity of the domestic computing power industry, allowing domestic computing power manufacturers to benefit from the rising demand for AI inference and training [3]

Investment Rating - The report maintains a "stronger than the market" rating for the electronic industry, indicating an expected relative performance that exceeds the benchmark index by over 10% [2][32][34]. Core Views - The report emphasizes that AI inference is driving innovation, with a focus on demand-driven Opex-related sectors, particularly in storage, power, ASIC, and ultra-node technologies [2][26][27]. - The electronic industry has shown resilience, with the Shenwan Electronic Index rising by 1.09% this week and 42.15% year-to-date, ranking 3rd among 31 sectors [10][11]. Summary by Sections Market Review - The overall market saw an increase, with the Shanghai Composite Index rising by 0.37% and the Shenzhen Component Index by 1.26%. The Shenwan Electronic Index's performance was ranked 13th among 31 sectors this week [10][11]. Weekly Insights - The report highlights the semiconductor equipment market, noting that China remains the top region for semiconductor equipment sales, with a total of $145.6 billion in Q3 2025 [21][22]. - It also discusses the anticipated growth in storage capacity, particularly with new products from Yangtze Memory Technologies and Changxin Memory Technologies, which are expected to drive expansion in the domestic storage industry [26]. Focus Areas - **Storage**: The report suggests a significant opportunity in the domestic storage industry, driven by increased demand for SSDs and HBM products [26]. - **Power**: It highlights the importance of new technologies in both the generation and consumption sides of the power industry, recommending specific companies for investment [27]. - **ASIC**: The report anticipates an increase in ASIC market share, focusing on key domestic and international CSP manufacturers [27]. - **Ultra-node**: It predicts growth in high-speed interconnects, cabinet manufacturing, liquid cooling, and PCB demand, with specific companies identified for potential investment [27][28]. Related Research - The report references several related studies that support the ongoing trends in AI computing and storage capabilities, indicating a strong outlook for the semiconductor and electronic sectors [4][21].

Core Viewpoint - Google's TPU is set to disrupt Nvidia's dominance in the AI chip market, with significant production increases and cost advantages for inference tasks [2][4][79]. Group 1: TPU Production and Market Strategy - Morgan Stanley predicts that Google's TPU production will surge to 5 million units by 2027 and 7 million by 2028, a substantial increase from previous estimates of 3 million and 3.2 million units, representing a 67% and 120% upward adjustment respectively [2]. - Google aims to sell TPUs to third-party data centers, complementing its Google Cloud Platform (GCP) business, while still utilizing most TPUs for its own AI training and cloud services [2][3]. Group 2: Comparison with Nvidia's GPU - Nvidia has historically dominated the AI chip market, controlling over 80% of it by 2023, but faces challenges as the market shifts from training to inference, where Google's TPU offers superior efficiency and cost advantages [8][12]. - By 2030, inference is expected to consume 75% of AI computing resources, creating a market worth $255 billion, growing at a CAGR of 19.2% [8][52]. Group 3: Cost and Efficiency Advantages of TPU - Google's TPU is designed for inference, providing a cost per hour of $1.38 compared to Nvidia's H100 at over $2.50, making TPU 45% cheaper [20]. - TPU's performance in inference tasks is four times better per dollar spent compared to Nvidia's offerings, and it consumes 60-65% less power [20][22]. Group 4: Industry Trends and Client Migration - Major AI companies are transitioning from Nvidia GPUs to Google's TPUs to reduce costs significantly; for instance, Midjourney reported a 65% reduction in costs after switching to TPU [34]. - Anthropic has committed to a deal for up to 1 million TPUs, highlighting the growing trend of companies seeking cost-effective solutions for AI workloads [35]. Group 5: Future Implications for Nvidia - Nvidia's profit margins, currently between 70-80%, may face pressure as Google captures even a small portion of the inference workload, potentially leading to over $6 billion in annual profit loss for Nvidia [22][59]. - The shift towards TPUs indicates a broader trend where companies are diversifying their AI infrastructure, reducing reliance on Nvidia's products [67].