February 3, 2026 12:34 PM GMT OVERVIEW Equity analyst Lee.Simpson@morganstanley.com +44 20 7425 3378 Nigel van Putten Equity analyst Nigel.Putten@morganstanley.com +44 20 7425-2803 MORGAN STANLEY TAIWAN LIMITED + Charlie Chan Equity analyst Charlie.Chan@morganstanley.com +886 2 2730-1725 Global Technology Webcast M O R G A N S T A N L E Y R E S E A R C H Global Memory, Semi Cap and AI Supply Chain Outlook Europe MORGAN STANLEY & CO. INTERNATIONAL PLC+ Shawn Kim Equity analyst Shawn.kim@morganstanley.com +44 ...

Group 1 - Broadcom (AVGO.US) is expected to maintain its leading position in the AI server ASIC design partnership field, with a market share projected to reach 60% by 2027 [1] - The shipment volume of AI server ASICs is anticipated to double by 2027, driven by the demand for Google's TPU infrastructure, Amazon's Trainium clusters, and the capacity enhancements from Meta's MTIA and Microsoft's Maia chips [1][2] - By 2028, the shipment volume of AI server ASICs is expected to exceed 15 million units, surpassing the shipment volume of data center GPUs [2] Group 2 - The market for AI server ASICs is diversifying, with Google and Amazon still leading in 2024, but their market shares are projected to decline by 2027, with Google's share dropping from 64% to 52% and Amazon's from 36% to 29% [3] - The top ten AI hyperscale data center operators are expected to deploy over 40 million AI server ASIC chips from 2024 to 2028, supported by large-scale AI infrastructure built on their technology stacks [2][3] - Broadcom and Alchip are projected to capture a significant portion of the ASIC design services market for hyperscale data centers, with shares of 60% and 18% respectively by 2027 [3] Group 3 - Marvell Technology (MRVL.US) is strengthening its end-to-end custom chip product portfolio, benefiting from innovations in custom silicon technology and the acquisition of Celestial AI, which could lead to significant revenue growth [4] - The acquisition of Celestial AI is expected to potentially position Marvell as a leader in the optical scaling connectivity market in the coming years [4]

1月26日，市场研究机构Counterpoint Research在最新研报中预测，非GPU服务器AI芯片——AI ASIC阵 营在近期将经历高速增长，到2027年出货数量将达到2024年的三倍，2028年则有望以1500余万颗的规模 反超GPU。 报告显示，这一爆炸性增长的背后，是对谷歌TPU基础设施的强劲需求、AWS Trainium集群的持续扩 展，以及Meta(MTIA)和微软(Maia)随着其内部芯片产品组合的扩展而带来的产能提升。 就人工智能超大规模数据中心的出货量而言，谷歌预计将在2027年之前保持市场领先地位，这主要得益 于其Gemini生态系统的爆炸式增长。 针对谷歌的统治地位，Counterpoint Research研究员David Wu强调："尽管由于市场规模不断扩大，以及 竞争对手(如博通、Marvell和Alchip等设计公司)纷纷采用自研芯片，预计谷歌的市场份额将在2027年下 降至52%，但其TPU集群仍将是无可争议的行业核心和标杆。这一基准的支撑源于训练和运行下一代 Gemini模型所需的庞大且持续的计算能力，而这又需要谷歌持续、积极地扩充其内部芯片基础设施。" 该机构认为， ...

| [Table_EPS] 盈利预测与估值 | 2023A | 2024A | 2025E | 2026E | 2027E | | --- | --- | --- | --- | --- | --- | | 营业总收入（百万元） | 2,338 | 2,322 | 3,154 | 5,661 | 8,612 | | 同比（%） | (12.73) | (0.69) | 35.84 | 79.49 | 52.11 | | 归母净利润（百万元） | (296.47) | (600.88) | (448.92) | 315.20 | 805.06 | | 同比（%） | (501.64) | (102.68) | 25.29 | 170.21 | 155.41 | | EPS-最新摊薄（元/股） | (0.56) | (1.14) | (0.85) | 0.60 | 1.53 | | P/E（现价&最新摊薄） | (366.46) | (180.81) | (242.01) | 344.67 | 134.95 | 证券研究报告·公司点评报告·半导体 芯原股份（688521） 2025 年业绩预告点评：订单兑现收入高 ...

Group 1 - The capital expenditure (Capex) of the top eight cloud service providers is projected to grow from $145.1 billion in 2021 to $602 billion by 2026, representing an increase of over four times [1][4] - This investment surge is driven not by market speculation but by the fundamental need for computational power, particularly due to the demands of generative artificial intelligence (AI) [1][5] - The current trend is characterized as a "structural transformation" in the semiconductor market rather than a bubble, as the demand for computing resources is fundamentally changing [10][14] Group 2 - The growth in cloud investment is accelerating, with a notable increase following the release of ChatGPT by OpenAI [4][5] - Generative AI requires significantly more computational resources compared to traditional search engines, with processing demands being 10,000 to 100,000 times greater [6][7] - The competition among cloud providers is fierce, as failure to invest in generative AI capabilities could lead to losing market relevance [8][17] Group 3 - The semiconductor market, particularly the data center logic chip sector, is expected to expand significantly, with the GPU market projected to grow from $100 billion to $230 billion and the AI ASIC market from $9 billion to $84 billion by 2030 [19] - The memory market is also anticipated to experience a shift, with DRAM and HBM prices expected to rise due to increased demand from AI applications [22][24] - The transition in TSMC's primary products from N5 to N3 nodes indicates a shift towards advanced technology driven by AI demands, with NVIDIA and Broadcom expected to surpass Apple in chip investments [28][33] Group 4 - The bottleneck in AI semiconductor development is primarily due to the limited capacity of 2.5D packaging technologies like CoWoS, which are essential for integrating high-bandwidth memory with AI chips [37][39] - Once the CoWoS capacity constraints are resolved, it is expected that investment in AI infrastructure will surge, leading to further competition among cloud service providers [39][42] - The ongoing trend signifies a profound and irreversible change in the semiconductor industry, driven by the structural demands of generative AI [42]

Group 1 - The core argument of the article is that the current investment trend in AI and cloud computing is not a bubble but a significant and irreversible shift in the semiconductor market driven by unprecedented computational demands [2][5][21] - The capital expenditure of the top eight cloud service providers is projected to grow from $145.1 billion in 2021 to approximately $602 billion by 2026, representing an increase of over four times [4][10] - The investment surge is primarily driven by the need for computational power required for generative AI, which is fundamentally different from traditional web services [5][10] Group 2 - The article emphasizes that generative AI requires a vastly different computational approach compared to traditional search engines, with computational loads differing by a factor of 10,000 to 100,000 times [12][16] - The growth trajectory of generative AI is expected to remain strong, with projected growth rates of -8.1% in 2023, 19.7% in 2024, and 22.5% in 2025, indicating a robust demand that is unlikely to decline [20][21] - The demand for logic chips in data centers is expected to grow significantly, with the GPU market projected to increase from $100 billion to over $230 billion, and AI ASICs expected to surge from $9 billion to $84 billion by 2030 [30][33] Group 3 - The storage market is anticipated to experience long-term shortages and high prices, with the DRAM market expected to grow from $97 billion to $194 billion, and HBM market reaching $98 billion by 2030 [36][38] - TSMC's revenue is shifting from N5 to N3 process nodes, indicating a transition in profitability driven by AI demands [41][46] - The bottleneck for AI semiconductors lies in CoWoS packaging capacity, and resolving this bottleneck could lead to an acceleration in investment rather than a slowdown [55][59]

Core Viewpoint - The global semiconductor industry is expected to experience stronger demand, particularly driven by the AI computing infrastructure and a "super cycle" in semiconductor equipment manufacturing, benefiting companies involved in AI chips and DRAM/NAND storage expansion [1][3]. Semiconductor Equipment Sector - KeyBanc Capital Markets highlights that semiconductor equipment manufacturers will be the largest beneficiaries of the AI chip and storage capacity expansion trends [1]. - Citigroup predicts a "Phase 2 bull market" for the semiconductor equipment sector, suggesting a shift from valuation recovery to sustained profit growth, with leading companies like ASML, Lam Research, and Applied Materials being key players [3]. - The semiconductor equipment sector is expected to see significant growth due to the ongoing demand for AI computing and storage solutions, with a focus on advanced manufacturing processes [4][5]. AI Infrastructure Investment - The construction of large-scale AI data centers by tech giants like Microsoft, Google, and Meta is accelerating the expansion of advanced AI chip production and storage capacity [4]. - The global AI infrastructure investment wave is projected to reach $3 trillion to $4 trillion by 2030, indicating that the current phase is just the beginning [5]. - The semiconductor market is expected to grow significantly, with a forecasted value of $772.2 billion in 2025 and $975.5 billion in 2026, driven by strong demand for AI GPUs and storage systems [6][9]. Market Dynamics - The demand for DRAM/NAND storage chips is surging, with prices increasing due to the heightened importance of these products in AI training and inference systems [10]. - TSMC reported a record gross margin exceeding 60% and raised its 2026 revenue growth forecast to nearly 30%, indicating strong demand for AI-related chip manufacturing [10][11]. - The semiconductor investment chain driven by AI demand is expected to lead to increased capital expenditures (capex) from major manufacturers like SK Hynix, Samsung, and Intel [12][13]. Company-Specific Insights - KeyBanc maintains an "overweight" rating on AEI Industries, citing its strong position in the data center sector and potential for revenue growth in semiconductor manufacturing equipment [14]. - Applied Materials is recognized for its diverse product offerings across various semiconductor manufacturing processes, with expectations for significant revenue growth in the coming years [15][16]. - MKS Instruments is positioned to benefit from the ongoing demand for advanced packaging and semiconductor manufacturing technologies, with a focus on maintaining a strong market share in NAND and advanced packaging sectors [18].

Summary of Key Points from the Conference Call Industry Overview - The semiconductor industry is experiencing a shift towards AI-driven demand, with a preference for AI semiconductors over non-AI counterparts. This trend is expected to continue into 2026, with both logic and memory sectors being attractive for investment [6][10][87]. Key Companies Mentioned - **TSMC**: Identified as a top pick in the AI semiconductor space, with expected revenue CAGR of 60% from AI semis between 2024 and 2029 [6][33]. - **SMIC**: Mentioned as a significant player in the semiconductor industry [6]. - **MediaTek, Alchip, GUC, Winbond, Phison, Nanya Tech**: Other notable companies highlighted for their roles in the semiconductor landscape [6]. Core Insights - **Demand Drivers**: - Tech inflation is anticipated to impact demand due to rising costs in wafers, OSAT, and memory, creating margin pressures for chip designers [6]. - AI cannibalization is a concern, as AI technologies may replace certain human jobs, affecting overall demand [6]. - The proliferation of generative AI is expected to drive demand across various verticals, including robotics and AI glasses [6]. - **Market Dynamics**: - The semiconductor supply chain is prioritizing AI semiconductors, leading to shortages in non-AI semiconductors [6]. - The memory sector's stock prices are seen as leading indicators for logic semiconductors, with an attractive industry view on Greater China technology semiconductors [16]. Financial Metrics and Valuation - **TSMC**: - Current share price is 1,760 TWD with a target price of 2,088 TWD, indicating a 19% upside potential [8]. - Expected P/E ratios for 2025, 2026, and 2027 are 26.6, 18.9, and 15.7 respectively, with EPS growth rates of 46%, 40%, and 21% [8]. - **SMIC**: - Current share price is 77.0 HKD with a target price of 80.0 HKD, indicating a 4% upside potential [8]. - Expected P/E ratios are not meaningful (NM) due to negative growth projections [8]. Potential Risks - **Supply Chain Issues**: The semiconductor supply chain is facing challenges, including inventory management and the prioritization of AI semiconductors, which could lead to shortages in non-AI products [6][14]. - **Economic Factors**: Rising costs and inflation in the tech sector may impact overall demand and profitability for semiconductor companies [6]. Additional Insights - **Capex Trends**: Major cloud service providers (CSPs) are expected to increase capital expenditures significantly, with a projected 65% year-over-year increase in Q3 2025 [52]. - **AI Semiconductor Market Size**: The global semiconductor market size is projected to reach $1 trillion by 2030, driven largely by cloud AI [85]. This summary encapsulates the critical insights and data points from the conference call, providing a comprehensive overview of the current state and future outlook of the semiconductor industry, particularly in relation to AI technologies.

Core Insights - The global semiconductor industry is expected to experience a stronger demand in the coming year, driven by the AI computing infrastructure wave and a "super cycle" in memory chips, benefiting semiconductor equipment manufacturers significantly [1][2] - Major investment firms like Citigroup and KeyBanc Capital Markets predict a "Phase 2 bull market" for semiconductor equipment, with a focus on leading companies such as ASML, Lam Research, and Applied Materials [1][2] Semiconductor Industry Outlook - The semiconductor equipment sector is identified as a major beneficiary of the surging demand for AI chips and DRAM/NAND storage chips, with expectations of a robust growth trajectory leading into 2026 [1][2] - The global semiconductor market is projected to grow by 22.5% in 2025, reaching a total value of $772.2 billion, and further expanding to $975.5 billion in 2026, indicating a year-on-year increase of 26% [5] AI Infrastructure Investment - The AI infrastructure investment wave is still in its early stages, with estimates suggesting a total investment of $3 trillion to $4 trillion by 2030, driven by the demand for AI computing hardware [3][4] - Companies like TSMC are experiencing significant growth, with a projected revenue increase of nearly 30% in 2026, largely due to the demand for AI-related chips and advanced packaging technologies [9][10] Key Players and Strategies - KeyBanc has raised target prices for semiconductor equipment companies, including AEI Industries, Applied Materials, and MKS Instruments, reflecting a bullish outlook on their growth potential [12][14][16] - Applied Materials is expected to benefit from its diverse product offerings and strong position in advanced packaging and DRAM markets, with a target price increase from $285 to $380 [14] - MKS Instruments is anticipated to see accelerated revenue growth due to its strong cash flow and leading position in power products for NAND etching tools [16][17]

Core Insights - Cerebras Systems Inc. is in discussions for a new funding round of approximately $1 billion to enhance its AI chip capabilities and compete with Nvidia, which currently holds a 90% market share in the AI chip sector [1][4] - The company's valuation is set to reach $22 billion, reflecting a significant increase of 170% from its previous valuation of $8.1 billion in September [2][4] - Cerebras aims to challenge Nvidia's dominance by leveraging its unique wafer-scale engine architecture, which reportedly offers superior performance and efficiency in AI inference tasks compared to Nvidia's GPU systems [3][5] Funding and Valuation - Cerebras Systems is seeking $1 billion in new financing, which would elevate its valuation to $22 billion, a substantial increase from $8.1 billion in September [1][2] - The funding is intended to support the company's long-term competition with Nvidia and to facilitate its planned IPO [1][4] Competitive Landscape - Cerebras Systems is recognized as one of the strongest competitors to Nvidia in the AI chip market, particularly in the rapidly growing AI inference segment [3] - The company utilizes a distinct wafer-scale engine architecture that enhances performance and memory bandwidth, providing a competitive edge over traditional GPU clusters [3][5] - Recent market dynamics indicate a growing interest in AI chips, with Nvidia's acquisition of Groq and its licensing agreement further intensifying competition in the sector [2][10] Technological Advantages - Cerebras' latest CS3 system, featuring the WSE3 chip, reportedly outperforms Nvidia's Blackwell architecture by approximately 21 times in specific large language model inference tasks [5] - The wafer-scale architecture allows for higher performance density and energy efficiency, particularly in large-scale inference scenarios [3][5] - While Cerebras excels in specific inference tasks, Nvidia maintains advantages in general computing tasks and compatibility with its CUDA ecosystem [5] Market Trends - The demand for AI inference capabilities is rapidly increasing, with projections indicating that the need for such technology is doubling every six months [9] - Companies are increasingly seeking cost-effective AI ASIC accelerators for cloud-based solutions, driven by the rising costs associated with AI inference [8][9] - The competitive landscape is evolving, with companies like Google also enhancing their AI capabilities through advancements in their TPU technology, further challenging Nvidia's market position [9][10]