J.P. Morgan 电子元件数据手册解读：AI 服务器如何重写 MLCC 需求结构（62 页全图复刻）
逐页复刻 J.P. Morgan《Technology – Electronic Components》（2026-09-11，62 页）并附中文结构化解读：MLCC 需求结构向 AI 服务器迁移、单机架价值量 3.6 倍跃升、技术代际驱动的利润率非线性，以及基板与供给侧证据。
J.P. Morgan 电子元件数据手册解读：AI 服务器如何重写 MLCC 需求结构（62 页全图复刻）
导读：本文逐页复刻 J.P. Morgan 亚太股票研究《Technology – Electronic Components》（2026 年 9 月 11 日，分析师 Akinori Kanemoto / Junya Ayada，62 页）的原始图表，并附本站的中文结构化解读。第一部分是原报告（52 页整页渲染，含全部图表）；第二部分是我基于报告数据独立撰写的解读——MLCC 需求结构如何被 AI 服务器重写、单机架价值量的跃升、技术代际带来的利润率非线性，以及基板与供给侧的交叉验证。
收录边界（重要）：原报告第 2–3 页（16 家公司的评级 / 目标价 / 估值表）与第 55–62 页（标准法务披露页）未收录；本文不含任何个股评级、目标价或投资建议，所有预测数字（CY26E–CY30E）均为 J.P. Morgan 估计而非事实。
第一部分：原报告（J.P. Morgan，2026-09-11）
以下为原报告逐页复刻。已略去：第 2–3 页（评级 / 目标价 / 估值表）与第 55–62 页（标准法务披露页）。其余 52 页全部保留，图表为原始版面整页渲染。
封面
Page 1 · 封面：报告基本信息与分析师
行业总览：股价、盈利预测与估值
这一组是整个板块的仪表盘：覆盖公司的股价表现、J.P. Morgan 的盈利预测与一致预期对比、自由现金流与资本回报指标、汇率敏感性，以及板块相对 TOPIX 的历史估值区间。
Page 4 · Electronic components：股价表现
Page 5 · Earnings forecast summary（盈利预测摘要 I）
Page 6 · Earnings forecast summary（盈利预测摘要 II）
Page 7 · FCF / ROIC / ROE 估计
Page 8 · F/X sensitivity（汇率敏感性）
Page 9 · 板块历史估值与相对 TOPIX 估值（12/24 个月远期 P/E、P/B）
MLCC：市场结构与需求预测
MLCC 是这份手册的重心。这里回答四个问题：需求来自哪些应用、份额如何分布、单台设备贡献多少价值量、以及到 2030 年市场会变成什么形状。
Page 10 · MLCC 收入按应用拆分（年度 / 季度）
Page 11 · MLCC 收入市占率（年度 / 季度）
Page 12 · MLCC 按应用的销售额（手机 / 汽车 / PC 服务器 / 消费）
Page 13 · MLCC 单机价值量：手机 / 电脑 / 汽车
Page 14 · 车用 MLCC：单车价值量、市占率、市场规模
Page 15 · MLCC 销售预测（按应用，至 CY2030E）
MLCC：供给、产能与供需平衡
需求故事的另一面是供给：稼动率、月产能、扩产节奏（以 Murata 工厂建设为样本）、供需平衡模型，以及降价周期对利润的量化冲击。
Page 16 · 四大厂 MLCC 产能稼动率
Page 17 · MLCC 月产能及同比增速
Page 18 · Murata 工厂建设与资本开支明细
Page 19 · MLCC 供需平衡模型（2017–2030E）
Page 20 · 稼动率与降价对营业利润的冲击（Murata / 太阳诱电）
AI 服务器：MLCC 需求的跃迁
本手册的核心章节。从 AI 加速器 TDP 路线图出发，推导单机架 MLCC 用量与容量、新一代产品（侧面成型）的成本与利润率结构，最终落到三大厂的利润预测。
Page 21 · AI 服务器与 GPU/ASIC 出货预测、总算力 TDP（GW）
Page 22 · AI 加速器 TDP 趋势（NVIDIA / Google / AWS / Microsoft / Meta / AMD / Huawei）
Page 23 · AI 服务器单板 MLCC 用量、平均容量、收入与 ASP 对比
Page 24 · AI 服务器中电容的应用位置
Page 25 · MLCC 成本 / 利润率分析（以 Murata 为模型案例，Gen5→Gen7）
Page 26 · MLCC 容量与 ASP 分析：GB300 与 VR200
Page 27 · 什么是侧面成型（Side Forming）MLCC？
Page 28 · MLCC：销售额与营业利润率趋势（Murata / 太阳诱电 / TDK）
封装基板：Ibiden 与 AI 基板
同一枚硬币的封装端：AI GPU / ASIC 的基板面积与层数暴涨，带来的 ABF 载板与 SAP（半加成工艺）需求，以及 Ibiden 的产能如何在 Intel / NVIDIA / AMD 之间再分配。
Page 29 · Ibiden（4062）：分部销售额与营业利润
Page 30 · AI / 通用服务器 GPU / CPU / ASIC 的基板尺寸与层数结构
Page 31 · Ibiden：SAP 需求展望与服务器占比
Page 32 · Ibiden：非 Intel 产能分配、客户与工厂对应、ABF 收入拆分
公司页
手册后半部分是个股数据页，每页给出分部销售额 / 营业利润及关键应用的预测拆分。此处按公司归组呈现。
Niterra（5334）
Page 33 · Niterra（5334）：分部销售额与营业利润
Page 34 · Niterra：汽车业务盈利趋势与产品拆分
Page 35 · Niterra：非汽车业务（工业陶瓷 / 氮化硅）盈利趋势
MinebeaMitsumi（6479）
Page 36 · MinebeaMitsumi（6479）：分部销售额与营业利润
Nidec（6594）
Page 37 · Nidec（6594）：分部销售额与营业利润
Wacom（6727）
Page 38 · Wacom（6727）：分部销售额与营业利润
TDK（6762）
Page 39 · TDK（6762）：分部销售额与营业利润
Page 40 · TDK：分部营业利润预测与传感器 / 电池 / 磁性应用拆分
Alps Alpine（6770）
Page 41 · Alps Alpine（6770）：分部销售额与营业利润
Hirose Electric（6806）
Page 42 · Hirose Electric（6806）：按应用的销售额
Page 43 · Hirose Electric：按应用销售额与通信设备占比
Japan Aviation Electronics（6807）
Page 44 · Japan Aviation Electronics（6807）：分部销售额与营业利润
Rohm（6963）与分立器件
Page 45 · Rohm（6963）：分部销售额与营业利润
Page 46 · 分立器件市场规模与份额（2023 vs 2024）
Page 47 · 服务器电源器件（含 DrMOS）行业展望（ROHM 口径）
Kyocera（6971）
Page 48 · Kyocera（6971）：分部销售额
Page 49 · Kyocera（6971）：分部营业利润
Taiyo Yuden（6976）
Page 50 · Taiyo Yuden（6976）：分部销售额
Murata Mfg.（6981）
Page 51 · Murata Mfg.（6981）：分部销售额与营业利润
Nichicon（6996）
Page 52 · Nichicon（6996）：分部销售额与营业利润
Nippon Chemi-con（6997）
Page 53 · Nippon Chemi-con（6997）：分部销售额与营业利润
NISSHA（7915）
Page 54 · NISSHA（7915）：分部销售额与营业利润
第二部分：结构化解读
以下解读由本站基于上述报告数据独立撰写，观点归属本站，不代表 J.P. Morgan。所有数字均取自原报告图表，CY26E–CY30E 为 J.P. Morgan 预测值，非事实陈述。报告不含个股评级与目标价（该部分未收录，见文首说明）。
一、这份报告是什么：一张被动元件行业的「体检表」
这不是一篇观点文，而是一份图表数据手册（chart pack）：62 页里几乎没有段落式论述，全部是带数据标签的图表与表格。它的价值不在「说了什么」，而在把整条日本电子元件产业链的量化事实摊在一张桌子上——需求按应用拆到 2030 年、供给拆到稼动率和月产能、单机价值量拆到「一个机架用多少颗电容、总共多少微法、值多少钱」。
覆盖链条可以概括为四层：
| 层 | 内容 | 代表页 |
|---|---|---|
| 需求侧 | AI 加速器 TDP 路线图、AI 服务器出货与总算力（GW） | P21–22 |
| 元件侧 | MLCC 用量 / 容量 / ASP、成本与利润率代际模型 | P23–28 |
| 供给侧 | 四大厂稼动率、月产能、工厂建设、供需平衡 | P16–20 |
| 封装侧 | AI 基板尺寸与层数、Ibiden 产能再分配 | P29–32 |
二、核心结论一：MLCC 的需求结构正在被 AI 服务器重写
报告最关键的一张表是 MLCC 销售预测（P15）。把数字抽出来：
| 应用 | CY25（$mn） | CY30E（$mn） | CAGR | 占比 CY25 → CY30E |
|---|---|---|---|---|
| 服务器（含 AI） | 1,200 | 19,681 | 75% | 9% → 56% |
| 汽车 | 4,926 | 7,435 | 8.6% | 39% → 21% |
| 电信 | 3,534 | 4,357 | 4.3% | 28% → 12% |
| PC / 平板 | 1,414 | 1,821 | 5.2% | 11% → 5% |
| 消费 | 978 | 1,059 | 1.6% | 8% → 3% |
| 合计 | 12,752 | 35,147 | 22.5% | — |
三个读法：
- 总量：五年 2.8 倍（$12.8bn → $35.1bn，CAGR 22.5%）。对一个人均资本开支以「亿日元」计、历史上长期个位数增长的被动元件行业，这个斜率本身就是事件。
- 结构：到 2030 年，服务器一项将占 MLCC 需求的 56%——它一个人的体量（$19.7bn）比 2025 年整个市场（$12.8bn）还大。汽车仍是稳健的第二大项（CAGR 8.6%），但从主角退居配角。
- 权重转移的含义：手机 / 消费 / 电信这类传统周期驱动合计占比从 47% 降到 20%。这意味着行业未来的估值锚将从「消费电子库存周期」切向「AI 算力资本开支周期」——两者的波动来源、领先指标和估值框架完全不同。
三、核心结论二：单机架价值量的跃升（GB300 → VR200，3.6 倍）
P26 给出了最硬的量化证据——每机架的 MLCC 用量与价值：
| 口径 | 用量（颗） | 总容量（µF） | 平均容量（µF） | 单机架价值（US$） | 单颗 ASP |
|---|---|---|---|---|---|
| 高端智能手机 | 1,200 | 4,100 | 3.42 | — | — |
| 通用服务器 | 3,000 | 22,000 | 7.33 | — | — |
| GB200 机架 | 355,000 | 4,000,000 | 11.27 | — | — |
| GB300 机架 | 355,000 | 5,000,000 | 14.08 | 2,840 | 0.8¢ |
| VR200 机架 | 564,000 | 12,000,000 | 21.28 | 10,155 | 1.8¢ |
| GB300 → VR200 增量 | +209,000 | +7,000,000 | 33.49 | +7,315 | 3.5¢ |
解读链条：
- 量：GB300 → VR200（NVIDIA Rubin 代机架），单机架用量 +59%，但总容量 +140%——平均单颗容量从 14µF 跳到 21µF，增量部分的平均容量高达 33µF。不是「用更多颗」，是「用更高级的颗」。
- 价：增量单颗 ASP 约 3.5¢，是 FY25 AI 服务器 MLCC 平均 ASP（约 $0.01）的 3 倍以上；单机架 MLCC 价值量从 $2,840 提到 $10,155，3.6 倍。
- 为什么是大容量小尺寸：P22 的 TDP 表是需求侧的总开关。NVIDIA 单 GPU 功耗从 A100 的 400W 一路走到 Rubin Ultra 3600W、Feynman Ultra 6000W；AMD MI450X/455X 到 1400–2000W，Meta 自研 MTIA 也到 1700W。功耗越高 → 电流瞬变越剧烈 → 电源完整性越难做 → 越需要在 ASIC 旁边放低等效串联电感（低 ESL）、大容量、小封装的去耦电容。侧面成型（Side Forming，P27）正是为「贴得更近、电感更低」服务的封装工艺。
四、核心结论三：代际跃迁是利润率跃迁，不只是收入跃迁
P25 用 Murata 做模型案例，把「同一颗电容」在不同技术代上的经济性拆开了：
| Gen5（2014–15） | Gen6（2018–20） | Gen7（2023–25） | |
|---|---|---|---|
| 规格 | 1005 / 10µF | 1005 / 22µF | 1005 / 47µF |
| ASP（美分） | 0.8 | 2.3 | 10.0 |
| 每 µF 价格（美分） | 0.080 | 0.102 | 0.213 |
| 介质层粒径（nm） | 150 | 100 | 80 |
| 生瓷片厚度（µm） | 1.2 | 0.8 | 0.5 |
| 叠层数 | 400 | 750 | 1,000 |
| 营业利润率 | 20.0% | 47.4% | 69.2% |
| 增量利润率 | 50% | 67.1% | 80.8% |
这是整份手册里最能解释「为什么被动元件能讲 AI 故事」的一页：
- ASP 十二年 12.5 倍，靠的不是涨价，而是在同样的 1005 封装尺寸里塞进 4.7 倍的容量——介质层从 150nm 磨到 80nm、生瓷片薄到 0.5µm、叠到 1,000 层共烧。这是材料与工艺的护城河，不是资本开支能直接买来的。
- 报告判断（P26）：增量主力规格 1005/47µF 的单颗价格约 10¢（普通版同容量的 12 倍），其增量利润率可能高于 70% 的行业平均，营业利润率可能超过 60%。换言之，AI 服务器带来的增量收入，落在利润表上的杠杆远大于收入表。
- 与 P11 的份额表对照：Murata 长期稳定在 41–45%，且报告明确认为其在侧面成型工艺上「显著领先同业」。份额稳定 + 结构上移，是典型的「量稳价升」利润结构。
五、核心结论四：供给是瓶颈，而瓶颈有结构性偏向
需求侧讲得再好，也必须回答「产能跟不跟得上」。报告给出的供给侧证据：
- 稼动率（P16）：四大厂（Murata、太阳诱电、SEMCO、TDK）历史稼动率区间大致在 60–100% 波动，报告给出 CY26 二季度时点（CoE）的当前读数。
- 产能（P17）：行业月产能以「十亿颗 / 月」计，Murata 单家占据大头。
- 扩产节奏（P18）：以 Murata 为样本的工厂建设明细——新建厂房的楼层数、钢构规格与建筑资本开支逐项列出。这类微观数据是验证「扩产是真是假」最硬的证据。
- 供需平衡（P19）：2017–2030E 的行业供需模型（十亿颗 + 供需差百分比）。
- 降价的利润冲击（P20）：量化了稼动率下行叠加降价对 Murata / 太阳诱电营业利润的季度冲击幅度。
值得注意的结构性矛盾：AI 服务器要的是高端小尺寸高容（1005/47µF 这类 Gen7 产品），与消费类产线并不完全通用。所以即使总量产能富余，高端产能仍可能紧张——这就是 P19 供需模型里「总量」与「高端」可能给出不同结论的原因，也是读这类数据手册时最容易被「合计数」掩盖的地方。
六、核心结论五：基板是同一个故事的封装端镜像
P30 的基板结构表，把 AI 算力对封装的挤压量化到了平方毫米：
| 产品 | 叠层结构 | 基板尺寸（mm） | 面积（mm²） | Build-up 层数 | Build-up 需求（mm²） |
|---|---|---|---|---|---|
| Intel Xeon（基准=100） | 5-2-5 | 77.5 × 56.5 | 4,379 | 10 | 43,788 |
| NVIDIA H100 | 5-2-5 | 58 × 55 | 3,190 | 10 | 31,900 |
| NVIDIA B100/B200 | 8-2-8 | 73 × 81 | 5,913 | 14 | 82,782 |
| NVIDIA Rubin | 10-2-10 | 100 × 100 | 10,000 | 20 | 200,000 |
| AMD MI350 | 8-2-8 | 75.4 × 72 | 5,429 | 16 | 86,861 |
| AWS Trainium3 | 9-2-9 | 73 × 88 | 6,424 | 18 | 115,632 |
| AWS Trainium4 | 11-2-11 | 115 × 120 | 13,800 | 22 | 303,600 |
| Microsoft Maia100 | 8-2-8 | 70 × 70 | 4,900 | 16 | 78,400 |
- Rubin vs Xeon：基板面积 2.3 倍，但 build-up 需求 4.6 倍——因为层数同时从 10 加到 20。基板的增长是「面积 × 层数」的乘积效应。
- Trainium4 vs H100：面积 4.3 倍、build-up 需求 9.5 倍。自研 ASIC 为降本走极端大封装路线，对载板厂商反而是更陡的价值量曲线。
- P31–32 落到 Ibiden：SAP 需求指数、非 Intel 产能如何从 Intel（Cell 5 / Gama）逐步转向 NVIDIA（Cell 2,3,4 → Ono+Cell4）与 AMD（Cell 2,3,4,5），以及 FC-BGA 之外新 2.5D 基板的收入线。
MLCC 与基板在这里是同一个逻辑的两次出现：AI 芯片把功耗和引脚密度同时推高，前者把去耦电容推向高容低感，后者把载板推向大尺寸多层——两者的共性是「单机价值量随代际非线性上升，且供给受工艺与良率约束」。
七、其他值得记录的量化事实
- 市占率稳定性（P11）：Murata 约 41–45%、SEMCO 约 24%、太阳诱电约 11–12%、TDK 在 10–15% 间波动。格局高度集中且多年未变——这是一个「份额换不动、只能换结构」的行业。
- 车用 MLCC（P13–14）：单车 MLCC 价值量持续上行，四大厂（Murata / TDK / 太阳诱电 / SEMCO）合计占据车用市场主导份额；汽车是需求结构中确定性最高、增速中等的一环。
- 离散器件（P46）：2024 年全球分立器件市场约 $35.0bn（同比 -7.6%），Infineon 20.2%、onsemi 9.5%、ST 7.0%、ROHM 3.4%；值得留意的是杭州士兰微（2.6%→3.0%）与 BYD 半导体（1.7%→2.8%）份额在上行——中国厂商在分立器件端的价格压力是结构性的。
- 服务器电源器件（P47）：含 DrMOS 的服务器功率器件行业展望（ROHM 口径），与 MLCC 共享同一个「AI 服务器电源完整性」需求源。
- 汇率敏感性（P8）：日本元件企业收入以美元计价、成本以日元计价，日元汇率对利润的敏感度是这类公司盈利预测的常量变量。
八、怎么用这份手册：三个可跟踪的先行指标
如果要把它变成一个可持续跟踪的框架，报告里最有信号价值的是三组数：
- NVIDIA Rubin / Feynman 的 TDP 与机架渗透节奏（P22）——MLCC 单机架价值量的直接乘数。
- 高端产能（而非总产能）：稼动率 + Murata 工厂建设进度（P16–18）——决定 2027–2028 年高端供需是否再度错配。
- 1005/47µF（侧面成型）的 ASP 与增量利润率（P25–26）——决定这轮需求红利有多少留在利润表。
九、阅读边界与风险提示
- 预测非事实：CY26E–CY30E 全部为 J.P. Morgan 估计，其中部分依赖其台湾科技团队 2026 年 4 月的 AI 服务器出货预测（P21）。AI 资本开支的任何下修都会等比例传导到这里的 2030 年数字。
- 口径注意：MLCC 市场数据为四大厂（Murata、太阳诱电、TDK、SEMCO）口径，非全球全行业；基板数据来自 JPM 对市售产品的实测与调查（P30 注释）。
- 未收录部分：第 2–3 页的评级 / 目标价 / 估值表、第 55–62 页的标准法务披露页未在本页复现；本站亦不提供任何投资建议。
- 来源说明：本 PDF 为第三方整理流传版本（带第三方水印），非 J.P. Morgan 官方分发渠道；数据以原报告为准，本站仅作研究性解读。
展开：原报告文字层（图表数据标签，共 52 页，供检索与核对数字）
Page 1 · 封面
Technology – Electronic Components Technology – Electronic Components Akinori Kanemoto AC +813-6736-8628 akinori.kanemoto@jpmorgan.com JPMorgan Securities Japan Co., Ltd. Technology – Consumer/Industrial/Precision Junya Ayada +813-6736-8631 junya.ayada@jpmorgan.com JPMorgan Securities Japan Co., Ltd. Asia Pacific Equity Research 11 September 2026 See the end pages of this presentation for analyst certification and important disclosures, including non-US analyst disclosures.
Page 4 · 股价表现
Electronic components Share price performance Note: Share price as of 9/9/2026 Source: Bloomberg Finance L.P., J.P. Morgan
Page 5 · 盈利预测摘要 I
Earnings forecast summary Note: BBG consensus as of 9/9/2026. Source: Company data, Bloomberg Finance L.P. consensus, J.P. Morgan estimates
Page 6 · 盈利预测摘要 II
Earnings forecast summary Note: BBG consensus as of 9/9/2026. Source: Company data, Bloomberg Finance L.P. consensus, J.P. Morgan estimates
Page 7 · FCF/ROIC/ROE
Electronic components FCF/ROIC/ROE estimates Note: J.P. Morgan estimates are all as of July-Sep 2026 (Nidec is as of Dec 2025). Source: Company data, J.P. Morgan estimates
Page 8 · 汇率敏感性
F/X sensitivity Source: Company data
Page 9 · 板块历史估值
Electronic components sector Historical valuations and relative valuation to TOPIX (Bloomberg consensus basis) 12-month forward P/E and relative P/E to TOPIX 12-month forward P/B and relative P/B to TOPIX 24-month forward P/E and relative P/E to TOPIX 24-month forward P/B and relative P/B to TOPIX Note: BBG consensus and share price as of 9/9/2026 Source: Bloomberg Finance L.P. consensus, J.P. Morgan estimates
Page 10 · MLCC 收入按应用
MLCC revenue by application $mn (yearly, quarterly) Note: Company data includes Murata, Taiyo Yuden, TDK, and SEMCO. 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q CY10 CY11 CY12 CY13 CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25CY26 $mn Automotive Computer Telecom Customer Industrial & Others 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q CY10 CY11 CY12 CY13 CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25CY26 Automotive Computer Telecom Customer Industrial & Others
Page 11 · MLCC 市占率
MLCC revenue market share (yearly, quarterly) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Murata Taiyo Yuden TDK Kyocera SEMCO Yageo 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 18,000 $mn Murata Taiyo Yuden TDK Kyocera SEMCO Yageo 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25 CY26 $mn Murata Taiyo Yuden TDK kyocera SEMCO Yageo 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q 3Q 1Q CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25 CY26 Murata Taiyo Yuden TDK Kyocera SEMCO Yageo M/S CY20 CY21 CY22 CY23 CY24 CY25 CY26 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q Murata 41.2% 40.8% 39.7% 41.8% 40.5% 41.5% 40.4% 41.9% 40.8% 42.8% 42.8% 44.3% 41.7% 40.9% 43.4% 44.9% 43.6% 43.8% 42.5% 44.1% 41.7% 45.2% 45.1% 45.8% 44.6% 45.0% Taiyo Yuden 13.0% 12.1% 13.3% 12.9% 12.1% 11.6% 11.7% 12.4% 12.3% 11.6% 12.4% 12.1% 12.4% 11.4% 11.4% 12.3% 12.1% 12.0% 11.6% 12.1% 12.3% 12.5% 12.5% 11.8% 11.8% 10.9% TDK 11.2% 10.1% 9.8% 10.4% 10.6% 10.3% 9.9% 10.6% 10.6% 12.2% 13.8% 14.9% 15.1% 15.1% 13.8% 13.5% 14.0% 13.2% 11.5% 11.9% 11.8% 12.4% 11.8% 12.7% 12.8% 12.7% SEMCO 24.8% 26.6% 25.4% 23.6% 25.5% 24.6% 25.7% 23.2% 24.7% 21.3% 18.0% 17.5% 18.3% 20.2% 19.4% 17.4% 18.3% 19.4% 23.5% 21.4% 22.8% 23.8% 24.9% 23.6% 24.2% 24.6% Others 9.8% 10.4% 11.8% 11.2% 11.3% 11.9% 12.3% 11.9% 11.6% 12.2% 13.0% 11.2% 12.5% 12.4% 12.0% 12.0% 12.0% 11.5% 10.9% 10.6% 11.3% 6.1% 5.8% 6.1% 6.6% 6.7%
Page 12 · MLCC 按应用销售额
MLCC Sales by application Note: Company data includes Murata, Taiyo Yuden, TDK, and SEMCO. For Smartphone For Automotive For PC/Servers For Consumer and Others 0 200 400 600 800 1,000 1,200 1,400 1,600 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q CY10CY11CY12CY13CY14CY15CY16CY17CY18CY19CY20CY21CY22CY23CY24CY25CY26 $mn 0 200 400 600 800 1,000 1,200 1,400 1,600 1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q3Q1Q CY10CY11CY12CY13CY14CY15CY16CY17CY18CY19CY20CY21CY22CY23CY24CY25CY26 $mn 0 200 400 600 800 1,000 1,200 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q CY10CY11CY12CY13CY14CY15CY16CY17CY18CY19CY20CY21CY22CY23CY24CY25CY26 $mn 0 100 200 300 400 500 600 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q CY10CY11CY12CY13CY14CY15CY16CY17CY18CY19CY20CY21CY22CY23CY24CY25CY26 $mn
Page 13 · MLCC 单机价值量
MLCC $ contents for smartphones, computers, autos Note: Company data includes Murata, Taiyo Yuden, TDK, and SEMCO. Source: Company data, J.P. Morgan MLCC $ contents per handset MLCC $ contents per computer MLCC $ contents per car -40% -30% -20% -10% 0% 10% 20% 30% 40% 50% 60% 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q CY11CY12CY13CY14CY15CY16CY17CY18CY19CY20CY21CY22CY23CY24CY25CY26 MLCC contents per handset (lhs,$) yoy % chg. (rhs) -40% -20% 0% 20% 40% 60% 80% 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q CY11CY12CY13CY14CY15CY16CY17CY18CY19CY20CY21CY22CY23CY24CY25CY26 MLCC contents per car (lhs,$) yoy % chg. (rhs) -100% -50% 0% 50% 100% 150% 200% 250% 300% 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q 3Q 2Q 1Q 4Q CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25CY26 MLCC $ contents per box (LHS, $) yoy % chg.(RHS)
Page 14 · 车用 MLCC
Automotive MLCC: $ contents per car, market share, market size Note: Combined Total for Murata Mfg., TDK, Taiyo Yuden, and SEMCO. Source: J.P. Morgan, Company data. Automotive MLCC revenues for major four suppliers $ contents per car of MLCC for major four suppliers Automotive MLCC market share for four major makers
Page 15 · MLCC 销售预测至 CY30E
MLCC sales forecast by application Source: Company data, J.P. Morgan estimates 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 CY20 CY21 CY22 CY23 CY24 CY25 CY26E CY27E CY28E CY29E CY30E $ mn Consumer PC/Tablet PCs Telecom Automotive Others Servers 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Consumer PC/Tablet Telecom Automotive Others Server AI Server $ mn CY10 CY11 CY12 CY13 CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25 CY26E CY27E CY28E CY29E CY30E Consumer 1,811 1,577 1,221 909 823 756 872 921 1,184 1,013 1,112 1,317 1,002 783 764 959 978 998 1,018 1,038 1,059 YoY % chg. -12.9% -22.6% -25.6% -9.4% -8.2% 15.3% 5.6% 28.5% -14.4% 9.8% 18.4% -23.9% -21.9% -2.3% 25.4% 2.0% 2.0% 2.0% 2.0% 2.0% PC/Tablet 1,577 1,467 1,450 985 803 679 848 949 1,379 1,443 1,629 1,898 1,617 1,017 1,147 1,414 1,424 1,515 1,619 1,717 1,821 YoY % chg. -6.9% -1.2% -32.1% -18.5% -15.5% 24.9% 11.9% 45.3% 4.6% 12.9% 16.5% -14.8% -37.1% 12.8% 23.2% 0.7% 6.4% 6.9% 6.1% 6.1% Server 390 421 430 438 451 465 511 614 859 902 1,082 1,515 837 598 816 1,200 2,694 5,641 10,122 14,839 19,681 YoY % chg. 8.0% 2.0% 2.0% 3.0% 3.0% 10.0% 20.0% 40.0% 5.0% 20.0% 40.0% -44.7% -28.6% 36.4% 47.1% 124.5% 109.4% 79.4% 46.6% 32.6% Telecom 1,790 2,510 2,994 2,820 2,737 3,013 3,098 3,238 4,114 4,036 4,552 4,997 3,761 3,088 3,131 3,534 3,577 3,758 3,948 4,148 4,357 YoY % chg. 40.2% 19.3% -5.8% -2.9% 10.1% 2.8% 4.5% 27.1% -1.9% 12.8% 9.8% -24.7% -17.9% 1.4% 12.9% 1.2% 5.1% 5.1% 5.1% 5.1% Automotive 740 864 1,013 1,080 1,175 1,332 1,684 1,957 2,559 3,036 3,046 4,119 4,252 4,409 4,516 4,926 5,317 5,812 6,288 6,840 7,435 YoY % chg. 16.8% 17.2% 6.6% 8.8% 13.4% 26.4% 16.2% 30.7% 18.7% 0.3% 35.2% 3.2% 3.7% 2.4% 9.1% 7.9% 9.3% 8.2% 8.8% 8.7% Others 627 545 391 302 365 411 431 745 1,065 888 964 1,229 926 784 797 719 733 748 763 778 794 YoY % chg. -13.0% -28.4% -22.7% 21.0% 12.5% 4.9% 73.0% 42.9% -16.7% 8.6% 27.6% -24.7% -15.4% 1.6% -9.7% 10.0% 2.0% 2.0% 2.0% 2.0% Total 6,934 7,384 7,498 6,534 6,354 6,656 7,444 8,424 11,160 11,318 12,385 15,076 12,395 10,678 11,171 12,752 14,722 18,471 23,757 29,360 35,147 yoy % chg. 6.5% 1.5% -12.9% -2.7% 4.8% 11.8% 13.2% 32.5% 1.4% 9.4% 21.7% -17.8% -13.9% 4.6% 14.2% 15.5% 25.5% 28.6% 23.6% 19.7% Consumer 26% 21% 16% 14% 13% 11% 12% 11% 11% 9% 9% 9% 8% 7% 7% 8% 7% 5% 4% 4% 3% PC/Tablet 23% 20% 19% 15% 13% 10% 11% 11% 12% 13% 13% 13% 13% 10% 10% 11% 10% 8% 7% 6% 5% Server 6% 6% 6% 7% 7% 7% 7% 7% 8% 8% 9% 10% 7% 6% 7% 9% 18% 31% 43% 51% 56% Telecom 26% 34% 40% 43% 43% 45% 42% 38% 37% 36% 37% 33% 30% 29% 28% 28% 24% 20% 17% 14% 12% Automotive 11% 12% 14% 17% 18% 20% 23% 23% 23% 27% 25% 27% 34% 41% 40% 39% 36% 31% 26% 23% 21% Others 9% 7% 5% 5% 6% 6% 6% 9% 10% 8% 8% 8% 7% 7% 7% 6% 5% 4% 3% 3% 2% AI Server 1% 3% 5% 13% 25% 38% 46% 52%
Page 16 · 四大厂稼动率
MLCC utilization for major four makers 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 110.0% 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q CoE CY13 CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25 CY26 Simple average UT Source: Company data (Murata Mfg., Taiyo Yuden, SEMCO, TDK), J.P. Morgan
Page 17 · MLCC 月产能
MLCC monthly capacity and yoy % chg. Source: Company data (Murata Mfg., Taiyo Yuden, SEMCO, TDK and Yageo), J.P. Morgan estimates 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 (bn units/month) Murata Taiyo Yuden TDK SEMCO Yageo yoy % chg.
Page 18 · Murata 工厂建设
Murata Mfg. factory production Announced Products Location Start of construction Completion Building Total Capex for building (¥mn) Construction Size Total floor area (m^2) Building area (m^2) Total area (m^2) 2014/09 MLCC Fukui 2014/12 2015/09 For Production 10,000 Steel frame above 5 stories 22,620 3,600 - 2015/04 R&D Nagaoka 2015/05 2016/02 Training Facilities NA Steel frame above 2/3 stories Training 3,386 Lodging 953 Training 2,006 Lodging 357 - 2015/05 MLCC Izumo 2015/04 2015/12 For Production Steel frame above 4 stories 5,334 1,331 - 2015/05 MLCC Izumo 2015/04 2016/10 For Production Steel frame above 6 stories 22,436 3,504 - 2015/10 NA Kanazawa 2015/11 2016/09 For Production 12,000 Steel frame above 7 stories 34,709 5,694 - 2015/12 Tooling Sabae 2016/01 2017/01 For Production 2,700 Steel frame above 4 stories 8,181 1,500 - 2016/05 Modules Komoro 2016/08 2017/05 For Production NA Steel frame above 5 stories 10,624 2,537 - 2016/7 MLCC Toyama 2016/10 2017/06 For Production NA Steel frame above 6 stories 28,500 4,443 - 2016/10 MLCC Fukui 2017/02 2017/10 For Material Production 10,000 Steel frame above 7 stories 18,119 3,522 - 2017/03 Inductor Tome 2016/05 2017/03 For Production 1,900 Steel frame above 3 stories 5,750 NA - 2017/03 HQ Nagaokakyo 2016/03 2017/03 NA NA Steel frame above 2/3 stories 9,525 NA - 2017/06 NA Komoro 2016/08 2017/06 For Production NA Steel frame above 5 stories 10,624 2,537 - 2017/06 Metrocirc Toyama 2016/10 2017/06 For Production 17,000 Steel frame above 6 stories 29,299 5,906 - 2017/07 MLCC Yokaichi 2016/08 2017/07 For Material NA Steel frame above 7 stories 12,574 1,985 - 2017/10 Metrocirc Wakura 2018/02 2019/01 For Production 3,500 Steel frame above 3 stories 11,310 3,310 - 2017/10 Metrocirc Kanawaza/Nomi 2017/10 2017/10 Factory acquisition NA NA NA NA 121,760 2017/11 MLCC Fukui 2017/02 2017/11 For Production 10,000 Steel frame above 10 stories 18,119 3,500 - 2017/11 NA Okayama 2017/12 2018/11 For Production 10,000 Steel frame above 6 stories 28,384 4,908 - 2017/11 R&D Yokohama 2018/05 2020/10 Innovation center 40,000 Steel frame above 18 / below 2 stories / 1 tower building 65,335 7,415 7,415 2017/12 NA Okayama 2017/04 2017/12 For Logistic 1,600 Steel frame above 5 stories 4,893 1,317 - 2018/03 MLCC Philippine 2017/02 2018/03 For Production 5,000 RC above 2 stories 73,117 32,278 - 2018/05 Metrocirc Ise 2018/06 2019/04 For Production 4,000 Steel frame above 6 stories 13,887 2,779 - 2018/06 MLCC Fukui 2018/06 2019/12 For Production 29,000 Steel frame above 6 stories Production: 54,144 Energy : 4,295 Production: 9,383 Energy : 1,800 - 2018/06 MLCC Izumo 2017/05 2018/06 Logistic, Welfare 3,800 Steel frame above 3/6 stories L 7,319 / W 4,640 L 1,470 / F 1,914 - 2018/08 Battery Wuxi - 2018/08 Logistic, Welfare 16,000 Production steel frame above 5 stories Welfare steel frame above 6 stories Production: 88,178 Welfare: 7,308 22,508 - 2018/08 MEMS sensor Finland 2018/08 2019/12 For Production 5,000 Above 5 stories 16,000 NA - 2018/09 MLCC Izumo 2018/10 2019/11 For Production 40,000 Production: Steel frame above 4 stories Energy: Steel frame above 3 stories 41,125 10,210 - 2018/10 MLCC Okayama 2018/10 2019/10 For Production, Material 10,000 Steel frame above 5 / below 1 stories 25,845 5,841 - 2018/10 NA Malaysia 2018/10 2020/01 For Production 4,000 Reinforced concrete, 3 floors above ground / 1 tower building 23,007 7,699 - 2018/11 MLCC Wuxi 2018/11 2019/12 For Production 14,000 Production: Reinforeced concrete, 4 stories Office: Reinforced concrete. 5 stories Energy: Reinforced concrete, 3 stories Poduction: 94,925 Office: 15,708 Energy: 16,595 Poduction: 22,155 Office: 15,708 Energy: 16,595 - 2019/12 EMI Filter/Antenna Coil Thailand 2018/12 2019/10 For Production EMI Filter Plant: 1,500 Antenna Coil Plant: 1,400 EMI Filter Plant: Reinforced concrete, 2 stories Antenna Coil Plant: Reinforeced concrete, 2 stories EMI Filter Plant: 9,993 Antenna Coil Plant: 9,995 EMI Filter Plant: 5,166 Antenna Coil Plant: 5,138 - 2019/07 Electrode Materials Shiga (Yasu) 2019/07 2021/02 For Material Production 14,000 Steel frame above 7 stories 23,052 4,289 - 2019/08 MLCC Shimane (Izumo) 2019/08 2020/07 For Production 4,990 Steel frame above 4 stories 9,328 2,812 - 2019/12 Ceramic material Okayama 2019/12 2020/12 For Production 16,000 Production: Steel frame above 7 stories Office: Steel frame above 3 stories Energy: Steel frame above 2 stories Production: 33,262 Office: 3,812 Energy: 1,706 Production: 5,306 Office: 1,361 Energy: 873 - 2021/11 RF module Nagano 2021/11 2022/12 For Production 5,770 Steel frame above 5 stories 13,460 3,300 - 2021/11 MLCC Thailand 2021/07 2023/03 For Production 12,000 Reinforced concrete, 2 floors above ground 80,950 35,088 - 2022/01 R&D (Plating technology) Fukui (Sabae) 2022/02 2023/08 For R&D 6,400 Steel frame above 6 / below 1 stories 11,322 1,797 - 2022/03 MLCC Shimane (Izumo) 2022/03 2023/04 For Production 12,000 Steel frame above 4 stories 22,120 6,314 - 2022/03 NA Shenzhen 2020/12 2022/03 Welfare NA Reinforced concrete above 5 stories 9,446 2,019 - 2022/05 NA Yokaichi 2022/05 2023/05 Welfare 3,550 Steel frame, above 3 stories (4 stories in some areas) 6,584 2,400 - 2022/06 Inductor DaNang 2022/06 2023/08 For Production 4,320 RC above 4 stories 22,637 5,486 - 2022/11 MLCC Wuxi 2022/11 2024/04 For Production 44,500 Production: RC above 3 stories Wareghouse: Steel frame above 4 stories EC: RC, 1 basement level, above 4 stories Production: 27,140 Warehouse: 11,630 EC: 12,519 Production: 6,557 Warehouse: 2,797 EC: 2,409 - 2023/08 MLCC Philippines 2023/08 2025/09 For Production 11,200 RC above 2 stories 77,981 32,454 - 2024/5 MLCC Izumo 2024/3 2026/3 For Production 47,000 NA 69,676 10,742 2025/05 Inductor Ho Chi Minh 2025/05 2026 1H For Production 3,000 RC above 3 stories 10,576 3,500 - 2026/02 MLCC Fukui (Echizen) 2023/11 2026/02 For R&D 35,000 Steel frame above 5 stories 41,709 - 54,450 2026/05 Thermistor Yokaichi 2026/05 2026/08 For Production 16,900 Steel frame, 5 stories 18,010 2,951 - 2026/05 Chip inductors noise suppression filters Miyagi (Tome) 2026/05 2027/12 For Production 9,700 Steel frame, 4 stories 13,391 3,902 - 18,000 Source: Company data
Page 19 · 供需平衡模型
MLCC supply-demand model - 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 -10% -5% 0% 5% 10% 15% 20% 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026E 2027E 2028E 2029E 2030E bn units MLCC industry procurement (RHS) MLCC industry volume (supply, RHS) MLCC industry S-D (%, LHS) Source: Company data, J.P. Morgan estimates
Page 20 · 降价对 OP 的冲击
MLCC capacity utilization and impact on operating profit from price declines Note: Impact on operating profit from price declines for Murata Mfg. and Taiyo Yuden Source: Utilization company data from Murata Mfg., Taiyo Yuden, SEMCO, TDK Murata Mfg. – Impact on operating profit from price declines and MLCC industry capacity utilization Taiyo Yuden – Impact on operating profit from price declines and MLCC industry capacity utilization Murata Mfg. – Impact on operating profit from price declines and % of price decline Taiyo Yuden – Impact on operating profit from price declines and % of price decline 60% 70% 80% 90% 100% -15,000 -10,000 -5,000 0 5,000 10,000 15,000 20,000 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q CY13 CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25CY26 OP impact by price decline (QoQ, ¥bn, LHS) MLCC Industry Utilization 60.0% 70.0% 80.0% 90.0% 100.0% -4,000 -3,000 -2,000 -1,000 0 1,000 2,000 3,000 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q 4Q 2Q CY13 CY14 CY15 CY16 CY17 CY18 CY19 CY20 CY21 CY22 CY23 CY24 CY25 CY26 OP impact by price decline (QoQ, ¥bn, LHS) MLCC Industry Utilization -16.0% -14.0% -12.0% -10.0% -8.0% -6.0% -4.0% -2.0% 0.0% -140,000 -120,000 -100,000 -80,000 -60,000 -40,000 -20,000 0 JPY mn Price decline (LHS) % of price decline (RHS) -25.0% -20.0% -15.0% -10.0% -5.0% 0.0% 5.0% -30,000 -25,000 -20,000 -15,000 -10,000 -5,000 0 5,000 JPY mn Price Deline (LHS) % of price decline (RHS)
Page 21 · AI 服务器出货与 TDP(GW)
- 5 10 15 20 25 30 2022 2023 2024 2025 2026E 2027E 2028E mn units High end datacenter GPU volumes AI GPU peripheral cards ASIC chip volume Shipment forecasts for AI servers and ASIC/GPU for AI servers, total TDP GW and average TDP forecast Note: Our Taiwan technology team‘s forecasts as of April 2026. Source: Company reports, Gartner®, J.P. Morgan estimates AI server shipment forecasts GPU/ASIC shipment forecasts for AI server - 5.0 10.0 15.0 20.0 25.0 2022 2023 2024 2025 2026E 2027E 2028E mn units AI Servers Non AI Servers 0% 20% 40% 60% 80% 100% 120% 140% 160% 180% - 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 2023 2024 2025 2026E 2027E Total TDP (GW) NVDIA AMD Google AWS META Microsoft yoy % chg. (lhs) 0% 10% 20% 30% 40% 50% 60% - 200.00 400.00 600.00 800.00 1,000.00 1,200.00 1,400.00 1,600.00 1,800.00 2023 2024 2025 2026E 2027E TDP (W) Average TDP (W, LHS) yoy % chg. (RHS) AI server total TDP shipment (GW) GPU/ASIC average TDP
Page 22 · AI 加速器 TDP 趋势
AI accelerator TDP trend Nvidia GPU Generation Ampere Product A100 H100 H200 B100 B200 B300 Rubin Rubin Ultra Feynman Feynman Ultra GPU TDP W 400 700 700 700 1000 1400 2300 3600 4400 6000 % chg. vs prior gen. - 75% 0% 0% 43% 40% 64% 57% 22% 36% Google TPU Product TPU v1 TPU v2 TPU v3 TPU v4 TPU v5e TPU v5p TPU v6 Trillium TPU v7e Ironwood TPU v7p Ironwood ASIC TDP W 75 280 220 170-175 120-200 250-300 120-200 AWS Product Trainium1 Trainium2 Trainium3 Trainium4 ASIC TDP W 300 500 500-1000 600-1200 Microsoft Microsoft Maia100 Maia200 Maia300 ASIC TDP W 700 750 750 META Product MTIA 100 MTIA 200 MTIA 300 MTIA 400 MTIA 450 MTIA 500 ASIC TDP W 25 90 800 1200 1400 1700 AMD Product MI300X/325X MI350X/355X MI430X/440X/455X GPU TDP W 750/1000 1000/1400 1400/2000 Huawei Product Ascend 910B Ascend 910C Ascend 950PR ASIC TDP W 400 600 600 Hopper Blackwell Rubin Feynman 600-700 Source: Company data
Page 23 · AI 服务器 MLCC 用量/ASP
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Average IT average Server Average Automotive Average MLCC ASP (US cent) MLCC units per box, MLCC average capacitance, server MLCC revenue, MLCC ASP comparison by application MLCC capacitance per unit base board basis for AI server Source: Company data, J.P. Morgan estimates as of June 2026 FY25 Sever MLCC revenue image 0.0 5.0 10.0 15.0 20.0 25.0 High end Smartphone NBPC Server H100 GB200 GB300 Vera Rubin Average MLCC Capacitance (uF) - 50,000 100,000 150,000 200,000 250,000 300,000 350,000 400,000 High end Smartphone NBPC Server H100 GB200 GB300 Vera Rubin Average MLCC Capacitance (uF) MLCC average capacity per unit 0 50,000 100,000 150,000 200,000 250,000 Total Server Ai Server General Server JPY mn Murata SEMCO Taiyo Yuden MLCC ASP by application (2025)
Page 24 · AI 服务器电容应用位置
Applications for capacitors in AI servers Source: Company data, J.P. Morgan
Page 25 · MLCC 成本/利润率 Gen5-7
MLCC cost/margin analysis (model case for Murata Manufacturing) Gen5 Gen6 Gen7 Launched 2014-2015 2018-2020 2023-2025 Standard High-end New Generation 1005/10uF 1005/22uF 1005/47uF ASP (cent) 0.8 2.3 10.0 Price per uF (cent) 0.080 0.102 0.213 BT diameter (nm) 150.0 100.0 80.0 Thickness Green sheet (um) 1.20 0.80 0.50 Layer count 400.0 750.0 1,000.0 Variable 0.40 0.74 1.92 Assuming 50% yield for New Generation Fixed cost 0.24 0.44 1.15 New generation Fixed cost up 50-100% due to lower yield and and additional step Operating profits 0.16 1.07 6.92 Variable cost 50.0% 32.9% 19.2% Fixed cost 30.0% 19.7% 11.5% OPM 20.0% 47.4% 69.2% Incremental Margin 50.0% 67.1% 80.8% Source: J.P. Morgan calculations
Page 26 · GB300 与 VR200 分析
MLCC capacitance, ASP analysis for GB300 & VR200
•
In FY25, the average ASP for AI server MLCCs is around US$0.01.
•
Comparing GB300 versus VR200 on a per-rack basis, unit count increases by roughly 210k pieces and total capacitance by 7,000,000 µF, with
average capacitance rising to ~33 µF. The incremental demand is driven mainly by specs such as 1005/22µF and 1005/47µF—and the 1005/47µF
in particular carries a very high unit price. Based on the implied average capacitance, we estimate the incremental mix is roughly 50% 22µF / 50%
47µF, which suggests the ASP on the incremental units could exceed US$0.06.
•
We believe the incremental margin on 1005/47µF could be even higher than the ~70% average incremental margin, and operating margins could
potentially exceed 60%. Murata is meaningfully ahead of peers in side-forming MLCCs, and we think it is best positioned to capture the full benefit
of this demand ramp.
Volume
Capacitance uF
Average capacitance (uF)
US$
US cent Per unit
High End Smartphone
1,200
4,100
3.42
General Server
3,000
22,000
7.33
GB200 (Rack)
355,000
4,000,000
11.27
GB300 (Rack)
355,000
5,000,000
14.08
2,840
0.8
VR200 (Rack）
564,000
12,000,000
21.28
10,155
1.8
VR-GB300
209,000
7,000,000
33.49
7,315
3.5
Type
Size (mm)
uF
Pricer per 1k units
ASP (US cent)
Price per uF
General Normal
0603
2.2uF
2.0
0.20
0.091
General Normal
1005
10uF
8.0
0.80
0.080
General Normal
1608
22uF
16.0
1.60
0.073
General Normal
2012
47uF
40.0
4.00
0.085
High-end Normal
0603
4.7uF
5.0
0.50
0.106
High-end Normal
1005
22uF
22.5
2.25
0.102
High-end Normal
1608
47uF
55.0
5.50
0.117
High-end Normal
2012
100uF
90.0
9.00
0.090
New gen Side Forming/Side Edge
0603
10uF
16.0
1.60
0.160
New gen Side Forming/Side Edge
1005
47uF
100.0
10.00
0.213
New gen Side Forming/Side Edge
1608
100uF
220.0
22.00
0.220Page 27 · 侧面成型 MLCC 定义
④Cutting+（Side forming process） What is side forming MLCC? Source: Company data, J.P. Morgan Side forming Normal MLCC MLCC production process ⑤Firing Side-surface edging / laser trimming Side-sheet press bonding （BaTiO3） Electrode Normal MLCC Side forming type Electrode Electrode L L + 2α （Side view） （Top view）
Page 28 · MLCC 销售与 OPM 趋势
MLCC: Sales and OP margin trend Note: All J.P. Morgan estimates are as of Sep 2026. Source: Company data, J.P. Morgan estimates MLCC revenue forecast MLCC OP forecast MLCC OPM forecast - 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 3,500,000 JPY mn Murata Mfg Taiyo Yuden TDK - 200,000 400,000 600,000 800,000 1,000,000 1,200,000 1,400,000 1,600,000 1,800,000 2,000,000 JPY mn Murata Mfg Taiyo Yuden TDK 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% Murata Mfg Taiyo Yuden TDK
Page 29 · Ibiden 分部
Ibiden (4062) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026. Source: Company data, J.P. Morgan estimates
Page 30 · AI 基板尺寸与层数
Substrate size and layer structure for AI/Gen server GPUs/CPUs/ASICs
Note: This is based on measurements and investigations we conducted on the products currently on the market.
Source: J.P. Morgan estimates
Substrate size by product (Xeon LGA4677=100)
SAP size by product (Xeon LGA4677=100)
Compay
Product
Layer count
L (mm)
W (mm)
Substrate Size
(mm2)
Build up layer
demand
(mm2)
Build-up layer
count
Intel
Xeon
LGA4677
5-2-5
77.50
56.50
4,378.75
43,788
10
Xeon Max
EMIB
5-2-5
NA
NA
5,700.00
57,000
10
Gaudi
NA
79.00
71.00
5,700.00
57,000
10
AMD
EPYQ
(Genoa, Bergamo)
5-2-5
72.00
75.40
5,428.80
54,288
10
MI350
8-2-8
75.40
72.00
5,428.80
86,861
16
CPU
9-2-9
78.00
78.00
6,084.00
109,512
18
Tranium3
9-2-9
73.00
88.00
6,424.00
115,632
18
Tranium4
11-2-11
115.00
120.00
13,800.00
303,600
22
Huawei
Ascend9100
NA
55.60
52.90
2,941.24
-
Microsoft
Maia100
8-2-8
70.00
70.00
4,900.00
78,400
16
H100
5-2-5
58.00
55.00
3,190.00
31,900
10
H200
6-2-6
58.00
55.00
3,190.00
38,280
12
B100
8-2-8
73.00
81.00
5,913.00
82,782
14
B200
8-2-8
73.00
81.00
5,913.00
82,782
14
Rubin
10-2-10
100.00
100.00
10,000.00
200,000
20
AWS
Nvidia GPUPage 31 · Ibiden SAP 需求
0
50
100
150
200
250
300
CY2024
CY2026 CoE
CY2027 CoE
Ono
Gama
Ogaki Central
Ogaki
Philippines
Malaysia
Aoyanagi
Cell8
Cell6
Cell4/5
Cell2/3
--
--
--
Launch
2025/FQ2
2026 -
(negotiating)
2007/2020
1998
2000
Items produced
AI Server
Networking
Next Gen
Leading PKG
AI Server
General Server
PC
AI Server
PC
PC
Core Process
Core process
Server Ratio
-
-
55% -> 70%
75% ->85%
0%
-
-
Server Ratio is based on sales (FY24 1H -> FY24 2H）
Factory
Ibiden: SAP demand outlook
Note: All CoE are as of May 2026.
Source: Company data, J.P. Morgan
Ibiden: Index of load on server substrates (converted to SAP,
CY24=100)
Ibiden: SAP demand index for general purpose servers, AI servers
and PC (CY22=100)
Ibiden: FC-BGA factory and server production ratioPage 32 · Ibiden 产能分配
Current Future Nvidia Cell 2,3,4 Ono + Cell4 AMD Cell2,3,4 Cell2,3,4,5 Others Cel 2-3 Cel 2-3 Intel Cell 5 Gama Ibiden: Non-Intel capacity allocation and factory by customer, revenues by customer Note: All J.P. Morgan estimates are as of March 2026. Source: Company data, J.P. Morgan estimates Ibiden: SAP capacity (non-Intel capacity 24/9=100) Ibiden: ABF substrate revenue by customer Ibiden: Factory by customer 0 0 00 0 200 2 0 00 0 00 0 00 2 2 26 6 2 2 2 0 ama ell hilippines ell y 2 Non ntel apacity Ibiden: ABF substrate OP by customer - 100,000 200,000 300,000 400,000 500,000 600,000 700,000 800,000 900,000 1,000,000 FY3/2021 FY3/2022 FY3/2023 FY3/2024 FY3/2025 FY3/2026 FY3/2027E FY3/2028E FY3/2029E FY3/2030E FY3/2031E JPY mn Intel Intel EMIB-T Nvidia AMD Others - 50,000 100,000 150,000 200,000 250,000 300,000 350,000 400,000 FY3/2021 FY3/2022 FY3/2023 FY3/2024 FY3/2025 FY3/2026 FY3/2027E FY3/2028E FY3/2029E FY3/2030E FY3/2031E JPY mn FC-BGA substrate New 2.5D substrate
Page 33 · Niterra 分部
Niterra (5334) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026. Source: Company data, J.P. Morgan estimates
Page 34 · Niterra 汽车
Automotive business earnings trend Automotive business sales by product OP margin trend by product Automotive business OP by product Note: All J.P. Morgan estimates are as of July 2026. Source: Company data, J.P. Morgan estimates 0 100,000 200,000 300,000 400,000 500,000 600,000 700,000 800,000 JPY mn Plugs Sensors 0 20000 40000 60000 80000 100000 120000 140000 160000 180000 JPY mn Plugs Sensors 0% 5% 10% 15% 20% 25% 30% 35% JPY mn Total OPM Plugs OPM Sensors OPM
Page 35 · Niterra 非汽车
0 50,000 100,000 150,000 200,000 250,000 JPY mn Cutting Tools Industrial Ceramics/Nitride Silicon SPE Medical Ceramic Package Fuel Cells Others Non Automotive - Sales (JPY mn) FY3/2020 FY3/2021 FY3/2022 FY3/2023 FY3/2024 FY3/2025 FY3/2026 FY3/2027E FY3/2028E FY3/2029E Cutting Tools 12,700 9,800 11,835 12,783 600 0 0 0 0 0 Industrial Ceramics/Nitride Silicon 5,800 6,200 6,507 7,419 7,000 1,600 23,000 31,000 36,000 39,600 SPE 17,811 22,915 29,448 40,638 35,200 44,300 52,000 68,975 86,219 103,463 Medical 20,671 24,805 25,579 29,415 38,700 40,000 37,500 37,510 35,960 35,960 Ceramic Package 16,501 15,330 20,055 18,534 13,310 15,211 17,000 18,600 18,972 19,351 Fuel Cells 0 0 0 0 0 400 500 2,000 4,000 6,000 Others 3,658 9,884 10,534 9,790 14,321 12,588 13,702 12,000 12,000 12,000 Total Sales 77,141 88,934 103,958 118,579 109,131 114,099 143,702 170,085 193,151 216,374 Non Automotive - Operating profits (JPY mn) FY3/2020 FY3/2021 FY3/2022 FY3/2023 FY3/2024 FY3/2025 FY3/2026 FY3/2027E FY3/2028E FY3/2029E Cutting Tools 300 -400 980 800 1,500 300 0 0 0 0 Industrial Ceramics/Nitride Silicon -900 -300 -900 0 -600 -1,700 -3,600 -5,500 -3,200 -1,760 SPE -506 2,549 6,450 15,300 7,100 6,100 11,000 21,190 28,088 34,985 Medical -427 -1,184 8,550 -2,600 -5,000 -7,300 -9,800 -2,000 -1,200 -400 Ceramic Package 35 -890 -397 -2,495 -3,591 -3,100 -1,500 -800 0 0 Fuel Cells 0 0 0 0 0 -300 -200 -200 200 400 Others -3,061 -8,976 -13,513 -15,046 -13,063 -5,196 6,956 0 0 0 Total OP -4,559 -9,201 1,170 -4,041 -13,654 -11,196 2,856 12,690 23,888 33,225 -30,000 -20,000 -10,000 0 10,000 20,000 30,000 40,000 JPY mn Cutting Tools Industrial Ceramics/Nitride Silicon SPE Medical Ceramic Package Fuel Cells Others Total OP Niterra: Earnings trend for non-automotive Note: All J.P. Morgan estimates are as of July 2026. Source: Company data, J.P. Morgan estimates Non-automotive business sales Non-automotive business OP
Page 36 · MinebeaMitsumi
MinebeaMitsumi (6479) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026.
Page 37 · Nidec
Nidec (6594) Sales and OP by segment Note: All J.P. Morgan estimates are as of Dec 2025.
Page 38 · Wacom
Wacom (6727) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026.
Page 39 · TDK 分部
TDK (6762) Sales and OP by segment Note: All J.P. Morgan estimates are as of Sep 2026.
Page 40 · TDK 分部预测
TDK: OP forecast by segment, sensor, battery, magnetic application sales breakdown OP forecast by segment Sensor business – Sales forecast by application Battery business – Sales forecast by application Note: All J.P. Morgan estimates are as of Sep 2026. Source: Company data, J.P. Morgan estimates Magnetic business – Sales forecast by application -200,000 -100,000 0 100,000 200,000 300,000 400,000 500,000 600,000 700,000 FY15 FY16 FY17 FY18 FY19 FY20 FY21 FY22 FY23 FY24 FY25 FY26EFY27EFY28E JPY mn Passive Components Sensor Applied Products Magnetic Applicaton Products Energy Application Products Film Application Other (-FY3/18) Other (FY3/19-) Elimination Total Operaing Profit - 200,000 400,000 600,000 800,000 1,000,000 1,200,000 1,400,000 1,600,000 1,800,000 2,000,000 FY15 FY16 FY17 FY18 FY19 FY20 FY21 FY22 FY23 FY24 FY25 FY26EFY27EFY28E JPY mn Small cells - Smartphonbe Small cells - NBPC/Tablet Small cells - Others Mid cells - 50,000 100,000 150,000 200,000 250,000 300,000 FY15 FY16 FY17 FY18 FY19 FY20 FY21 FY22 FY23 FY24 FY25 FY26EFY27EFY28E JPY mn Tempreture and Pressure Sensor Magnetic Sensor MEMS products - 100,000 200,000 300,000 400,000 500,000 600,000 FY15 FY16 FY17 FY18 FY19 FY20 FY21 FY22 FY23 FY24 FY25 FY26EFY27EFY28E JPY mn HDD head HDD suspension Magnet Others
Page 41 · Alps Alpine
Alps Alpine (6770) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026.
Page 42 · Hirose 按应用
Hirose Electric (6806) Sales by application Note: All J.P. Morgan estimates are as of Sep 2026.
Page 43 · Hirose 明细
Hirose Electric: Sales by application and sales breakdown for telecom equipment Note: All J.P. Morgan estimates are as of Sep 2026. Source: Company data, J.P. Morgan estimates Sales forecast by application Sales forecast for telecom equipment 0 50,000 100,000 150,000 200,000 250,000 300,000 350,000 FY3/2016 FY3/2017 FY3/2018 FY3/2019 FY3/2020 FY3/2021 FY3/2022 FY3/2023 FY3/2024 FY3/2025 FY3/2026 FY3/2027E FY3/2028E FY3/2029E JPY mn Connector -- Industrial Connector -- Automotive Connector -- Smartphone Connector -- Consumer Mobile Connector -- Others Others 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 FY3/2025 FY3/2026 FY3/2027E FY3/2028E FY3/2029E JPY mn Non AI AI Server
Page 44 · JAE
Japan Aviation Electronics (6807) Sales and OP by segment Note: All J.P. Morgan estimates are as of Sep 2026.
Page 45 · Rohm 分部
Rohm (6963) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026.
Page 46 · 分立器件份额
2023
2024
2023
2024
Discrete Total
Sales (mn$) Sales (mn$)
Share
Share
Infineon
7,855
7,060
20.7%
20.2%
Onsemi
3,824
3,335
10.1%
9.5%
STMicroelectronics
3,092
2,465
8.2%
7.0%
Nexperia
1,761
1,673
4.6%
4.8%
Mitsubishi Electric
1,638
1,598
4.3%
4.6%
Fuji Electric
1,386
1,245
3.7%
3.6%
ROHM
1,322
1,175
3.5%
3.4%
Vishay
1,302
1,125
3.4%
3.2%
Hangzhou Silan Microelectonics
982
1,056
2.6%
3.0%
BYD
657
980
1.7%
2.8%
Toshiba
1,113
935
2.9%
2.7%
Others
12,956
12,383
34.2%
35.3%
Total
37,888
35,030
100.0%
100.0%
Rohm + Toshiba
2,435
2,110
6.4%
6.0%
Rohm + Toshiba + Mitsubishi Elec.
4,073
3,588
10.8%
10.2%
2023
2024
2023
2024
Small Signal Discretes
Sales (mn$) Sales (mn$)
Share
Share
Nexperia
637
687
16.1%
19.0%
Onsemi
654
502
16.6%
13.9%
ROHM
406
353
10.3%
9.8%
Diodes
336
260
8.5%
7.2%
Vishay
274
237
6.9%
6.6%
STMicroelectronics
228
221
5.8%
6.1%
Infineon
193
178
4.9%
4.9%
China Resources Microelectronics
180
176
4.6%
4.9%
Semtech
167
167
4.2%
4.6%
Toshiba
202
162
5.1%
4.5%
Others
673
673
17.0%
18.6%
Total
3,950
3,616
100.0%
100.0%
Rohm + Toshiba
608
515
15.4%
14.2%
Rohm + Toshiba + Mitsubishi Elec.
693
629
17.5%
17.4%
2023
2024
2023
2024
Power Discretes
Sales (mn$) Sales (mn$)
Share
Share
Infineon
7,399
6,617
24.6%
23.9%
Onsemi
3,166
2,829
10.5%
10.2%
STMicroelectronics
2,798
2,216
9.3%
8.0%
Mitsubishi Electric
1,553
1,484
5.2%
5.4%
Fuji Electric
1,386
1,245
4.6%
4.5%
Hangzhou Silan Microelectonics
967
1,040
3.2%
3.8%
BYD
657
980
2.2%
3.5%
Vishay
1,028
888
3.4%
3.2%
Nexperia
956
831
3.2%
3.0%
ROHM
913
819
3.0%
3.0%
Toshiba
905
769
3.0%
2.8%
Others
622
655
2.1%
2.4%
Total
30,102
27,722
100.0%
100.0%
Rohm + Toshiba
1,818
1,588
6.0%
5.7%
Rohm + Toshiba + Mitsubishi Elec.
3,371
3,072
11.2%
11.1%
0
500
1,000
1,500
2,000
2,500
3,000
3,500
$mn
Rohm
Rohm+Toshia
Rohm+Toshiba+Mitsubishi Elec.
0
500
1,000
1,500
2,000
2,500
3,000
Automotive
Wired
Communication
Wireless
Communication
Cosumer
Electronics
Computer
Storage
Industrial, others
$mn
Rohm
Rohm + Toshiba
Rohm+Toshiba+Mitsubishi Elec.
Discrete semiconductor market share, Rohm/Toshiba/Mitsubishi Electric (power
semi) revenue by application and product
Source: Omdia, Gartner® and J.P. Morgan
Discrete semiconductor market share
Rohm/Toshiba/revenue by product and application
FY3/27 earnings level
•
Rohm sales ¥560bn/OP ¥63bn
•
Toshiba (semiconductor) sales ¥420-440bn/OP ¥20-30bn
2 company total sales roughly ¥1trn/OP ¥90bn in FY3/27
•
Mitsubishi Elec (power semi) sales ¥215bn/OP ¥25bn
3 company total sales ¥1.22trn/OP ¥115bnPage 47 · 服务器电源器件
Server Power Device (incl. DrMOS) Industry outlook
Source: Company data (ROHM estimates)Page 48 · Kyocera 销售额
Kyocera (6971) Sales by segment Note: All J.P. Morgan estimates are as of July 2026.
Page 49 · Kyocera OP
Kyocera (6971) OP by segment Note: All J.P. Morgan estimates are as of July 2026. Source: Company data, J.P. Morgan estimates
Page 50 · 太阳诱电
Taiyo Yuden (6976) Sales by segment
Page 51 · Murata 分部
Murata Mfg. (6981) Sales and OP by segment Note: All J.P. Morgan estimates are as of Sep 2026.
Page 52 · Nichicon
Nichicon (6996) Sales and OP by segment Note: All J.P. Morgan estimates are as of Sep 2026.
Page 53 · Nippon Chemi-con
Nippon Chemi-con (6997) Sales and OP by segment Note: All J.P. Morgan estimates are as of Sep 2026. Source: Company data, J.P. Morgan estimates
Page 54 · NISSHA
NISSHA (7915) Sales and OP by segment Note: All J.P. Morgan estimates are as of July 2026.
本文由作者按照 CC BY 4.0 进行授权