投资研究 · 更新于 2026-09-20Investment Research · Updated 2026-09-20

杂志级美股投研报告,Magazine-quality US stock research,
一行命令即可生成。generated in one command.

每张卡片显示三档公允价 — Bear / Base / Bull,完全基于最新季度盈利计算。没有买卖评级,没有主观意见。看完下面的公式,自己判断。Each card shows three fair-value scenarios — Bear, Base, Bull — computed from the latest quarterly earnings. No buy/sell ratings, no opinions. Read the formula below, then judge yourself.

32 份报告 · 覆盖 10 个热门赛道 · 点击任一卡片查看完整分析32 reports across 10 hot sectors · Click any card to view the full analysis

三档价格怎么算出来的How the numbers are computed§ 00

100% 透明,每只票同一套公式。100% transparent. Same formula for every stock.

第 1 步 · 取锚Step 1 · Anchor

latest_quarterly_EPS × 4

取最新一季的摊薄 EPS,乘以 4,得到 run-rate 年度化 EPS — 当前盈利能力的最新鲜数据,不是 12 个月前的旧账。Take the most recent quarter's diluted EPS and multiply by 4. This is the run-rate annualized EPS — the freshest signal of current earnings power.

第 2 步 · 乘倍数Step 2 · Multiply

× target_P/E_multiple

对 Bear / Base / Bull 三种情景分别用三个不同的 P/E 倍数。倍数取决于股票是周期股还是增长股(见第 3 步)。Apply three multiples for Bear / Base / Bull scenarios. The multiple depends on whether the stock is classified as cyclical or secular growth (see Step 3).

	BEAR	BASE	BULL
周期股Cyclical	8×	14×	20×
增长股Secular	18×	25×	35×

第 3 步 · 分类Step 3 · Classify

cyclical / secular?

用盈利数据判定,不看行业名字。NVDA 和 MU 表面都是「半导体」,但实际走势完全不同。We classify from income data, not industry name. NVDA and MU are both 'Semiconductors' on paper but behave entirely differently.

5 年内有亏损年,或峰值利润年不在最近 2 年 → 周期股;否则 → 增长股。Cyclical if any negative-income year in 5y history, OR peak income year is not in the latest 2 years.

最终公式Final formula

Bear / Base / Bull = latest_Q_EPS × 4 × P/E_multiple

每份单独报告里还有另外 3 种 DCF 方法(历史 FCF / 前瞻 FCF 调整 / 前瞻纯 EPS)做交叉验证。主页只显示 Q-annualized,因为它反映的是最新 run-rate。Three more DCF lenses (Historical FCF, Forward FCF-adjusted, Forward Pure EPS) appear inside each individual report for cross-check. Q-annualized is the headline because it reflects the most current run-rate. 查看完整方法论 →Read full methodology →

AI 算力AI Compute§ 01

光模块Optical Modules§ 02

存储芯片Memory & Storage§ 03

AI 基建 & 电力AI Infrastructure§ 04

科技巨头Mega-cap Tech§ 05

加密 & 稳定币Crypto & Stablecoin§ 06

制药 · GLP-1Pharma · GLP-1§ 07

电动车Electric Vehicles§ 08

价值 · 综合体Value · Conglomerate§ 09

其他覆盖Other Coverage§ 10

估值方法The methodology

传统 DCF 模板用历史 FCF 增长率,对周期股和加速增长股都会系统性失真。本报告默认主估值锚是 季度年度化(Q-annualized) — 最新季 EPS × 4 × 周期感知 P/E 倍数 — 再配合另外三种 DCF 镜头交叉验证。Most stock-DCF templates use historical FCF growth, which systematically misprices cyclicals and accelerating growers. The default primary anchor here is Q-annualized — latest quarterly EPS × 4 × cycle-aware P/E multiple — paired with three other DCF lenses for cross-check.

周期股从盈利数据判断,不从行业名称。市场情绪由价格动量代理生成,并对基本面评级做缓冲调整。Cyclicality is detected from income data, not industry name. Sentiment is computed from price-action proxies and buffered against the fundamental verdict.

查看完整方法论 →Read methodology →

自己跑起来Run it yourself

开源项目。生成这些报告的同一套流水线已打包为 Claude Code skill。克隆仓库,配置 FMP API key,10 秒内分析任意美股票代码。This is open-source. The same pipeline that generated these reports is packaged as a Claude Code skill. Clone the repo, set your FMP API key, and analyze any US ticker in 10 seconds.

./run.sh NVDA

GitHub 仓库 ↗View on GitHub ↗

杂志级美股投研报告,Magazine-quality US stock research,一行命令即可生成。generated in one command.