Rapid Company Research

Rapid Company Research

Deep value analysis with Munger's models

Installed by
9
CategoryResearch
FromYouMind
宁德时代(300750.SZ)投资备忘录

投资要点摘要

宁德时代是全球动力电池(市占率 39.2%,连续九年第一)与储能电池(市占率 30.4%,连续五年第一)的双料冠军。2025 年营收 4,237 亿元,归母净利 722 亿元(+42%),净利增速是营收增速的 2.5 倍,毛利率 26.3%、净利率 18.1%、ROE 约 25%,均为近五年最高。经营现金流 1,332 亿元,自由现金流约 911 亿元,现金储备近 4,000 亿元。连续三年以净利的 50% 分红,累计分红近千亿。

结论:公司质量 = 好,价格 = 合理偏低估。动作 = 可在 400 元以下分批买入,450 元以上观望或小幅减仓。

关键待验证问题: 固态电池产业化进度对现有液态产线的冲击、海外工厂(匈牙利 100GWh)投产后的实际盈利、储能业务增长持续性。

核心风险: 固态电池技术替代、地缘政治(美国关税、欧盟电池护照)、车企自研电池、行业产能过剩导致的降价压力。


一、商业模式与护城河

1.1 怎么赚钱

宁德时代是一家“卖铲子”的公司——下游车企和储能集成商抢市场,它卖电池。收入来自三个板块(均为 2025 年报数据,口径:年度/中国会计准则):

  • 动力电池系统:营收 3,165 亿元,占比 74.7%,同比 +25%,毛利率 23.84%。卖给全球主流车企(特斯拉、宝马、大众、蔚来、理想等),产品包括麒麟电池、神行超充、M3P、钠离子、凝聚态电池。citation

  • 储能电池系统:营收 624 亿元,占比 14.7%,同比 +9%,毛利率 26.71%(高于动力电池)。储能是利润质量最高的板块。citation

  • 电池材料及回收:营收 219 亿元,占比 5.2%,毛利率 27.27%。打通“矿产-材料-电池-回收”全链路。citation

地区上,国内营收 2,941 亿元(占 69%,毛利率 24%),境外 1,296 亿元(占 31%,毛利率 31.44%,比国内高 7.4 个百分点)。海外是高毛利引擎,但地缘风险也最大。citation

核心增长方程式:净利 = 出货量 × 单价 × 毛利率 − 固定成本。过去三年,出货量持续高增(2025 年 661GWh,+39%),毛利率逆势扩张(23%→24%→26%),规模效应使费用率下降,净利增速远超营收。

来源:SNE Research、公司 2025 年年报 citation

1.2 供给独占性——护城河在变强

逐项判断五大护城河来源:

  • 规模经济:产能 772GWh,产能利用率 96.9%,2026Q1 出货超 200GWh。citation 单位成本比同行低 10-20%,这是最硬的壁垒——对手要追上,需要千亿级资本开支 + 多年客户认证周期。citation

  • 转换成本:电池认证周期 18-24 个月,车企一旦选定供应商,切换成本极高。还与特斯拉、宝马等成立合资公司,“电池+底盘”集成方案进一步锁定客户。citation

  • 技术领先:累计研发投入超 900 亿元,全球专利超 5.4 万项,固态电池相关专利申请量全球第一(超 1,300 项)。citation 神行超充电池 12C(充电 5 分钟续航 520 公里),钠电池已量产上车,-40°C 仍保持 90% 电量。citation

长鑫科技(688825)投资备忘录 — 周期反转视角

标的:长鑫科技集团股份有限公司(CXMT,688825.SH)
视角:周期反转
日期:2026年7月27日(上市首日)
当前市值:约3.66万亿元(午盘54.65元/股,+531%)
发行价:8.66元/股|发行市值约5,792亿元


一、投资要点摘要

一句话结论:长鑫存储是国产 DRAM 唯一量产 IDM,赛道稀缺性无可争议,但上市首日3.66万亿市值对应约30倍2026年峰值利润,已 price-in 了最乐观的周期延长+份额跃升+HBM 突破三重预期。从周期反转视角看,当前价格已处于“巅峰定价”区间——这是周期股最危险的买入时点。

三情景估值速览

情景核心假设2026E 归母净利中周期正常化利润合理市值区间vs 当前市值
乐观AI 结构性短缺至2028+,份额升至15%,HBM3量产~1,300亿~700亿1.0–1.7万亿严重高估
中性周期2027见顶,2028温和回落,份额10–12%~1,200亿~450亿4,500–6,000亿高估
悲观AI capex 骤降+新增产能集中释放+设备制裁~1,000亿~150亿1,200–2,400亿极度高估

当前市值(~3.66万亿)已超出乐观情景上沿,市场在用“成长股估值法”给一只“周期股”定价——这是周期投资中最经典的陷阱。

关键风险:①DRAM 强周期性,2023年亏损163亿的镜像可能在下一轮下行重演;②30x 峰值 PE 隐含“这次不一样”的叙事,但 DRAM 历史上每次“超级周期”最终都回归均值;③EUV 禁令锁死先进制程上限;④HBM 良率仅35–50%,AI 高端市场可望不可即;⑤首日涨幅531%,战略配售锁定期结束后抛压巨大。

关键待验证问题:本轮 DRAM 涨价是结构性(AI 驱动,持续2–3年)还是周期性(2027见顶回落)?长鑫2026年 Q1净利率49%是否可持续?HBM3能否在2026年底如期量产且良率达标?


二、详细正文

模块1:商业模式与护城河

Description

Editor's Recommendation

Recommended by

淡苍

Why we love this skill

Deeply dissects companies from a buy-side research perspective, integrating Munger's mental models with multi-dimensional data verification, and provides a professional investment memorandum including moat analysis and three-scenario valuations.

Enter a company name or stock ticker. Using Charlie Munger's multiple mental models and a value investing perspective, this SKILL conducts deep investment research and analysis on stocks. It automatically searches the internet for in-depth information, enforcing a 'search first, then judge' approach, cross-referencing dual sources, and annotating data definitions and sources. It performs deep analysis according to a seven-dimensional value investing framework, provides three-scenario valuations and a 2×2 quality/price decision matrix, and outputs an actionable 'Value Investing Deep Analysis' report.

Related Skills

View all
Research

Slow Teacher's Keyword Method

Use the keyword learning method to quickly get started in any field: output a table of 20 core keywords (with one-sentence explanations, application scenarios, and best practices), hand-drawn comic-style SVG logic relationship diagrams, simulate a domain expert answering 5 key questions, recommend 3–5 professional books, and assemble them into a well-formatted report; enter 'Interpret [Book Title]' to switch to the seven-part in-depth book interpretation mode.

P
4500
Research

Signal Room: Interview Synthesis

YouMind already transcribes your calls, interviews and podcasts. Signal Room is what happens next. Drop in one transcript or twenty and get back a research synthesis an actual analyst would sign: coded themes, verbatim evidence with timestamps, the places people disagree, and a ranked answer to the decision you are trying to make. The method is real qualitative practice, not summarisation: • Open coding that works quote-first — no quote, no code — with codes named in the participant's own words rather than analyst jargon • Every code tagged Behaviour, Belief or Wish, because "I would definitely pay for that" is not the same class of evidence as "I paid for that last month" • Themes stated as falsifiable sentences, with strength counted in participants rather than quotes, and disconfirming evidence hunted for on purpose • A tension map showing where your participants genuinely split and what predicts which side they fall on • An opportunity backlog written as "when [situation], [who] wants [outcome] because [reason]", each rated Strong, Suggestive or Anecdotal • A direct answer to your decision question, with a stated confidence level and what would change it • The three questions this round could not answer, and who to interview next Guardrails that matter: it never invents or polishes a quote, it refuses to report percentages on fewer than twelve participants, it pseudonymises participants by default, and it will tell you to your face when n=1 means you have a hypothesis rather than a finding. For product managers, UX and market researchers, journalists, consultants, founders doing customer discovery, and anyone sitting on hours of recordings and no findings.

M
6100
Research

AI Paper Master

Helps product managers, founders, and app developers understand AI papers through historical causal links, turning those insights into product judgments, technical boundaries, engineering intuition, and opportunity analysis.

J
150

Find your next favorite skill

Explore more curated AI skills for research, creation, and everyday work.

Explore all skills