Independent structural monitor
Value Chain Leadership
As of Sep 18, 2026 ET Daily model, weekly publication
SHADOW · READYMethod
Value Chain Leadership
A separate SHADOW monitor of relative price leadership across nine fine AI value-chain groups and five coarse groups. It does not modify Risk Appetite, Regime or Small-Cap Rotation.
- The identification universe freezes the audited 49-company stable core from the five-year study; the wider 58-company dynamic universe is used only for attribution and coverage checks.
- Each group uses daily equal-weight split-adjusted price returns relative to the equal-stock AI-chain composite.
- The 21-session relative log-return score is shifted by one session, so the current close cannot affect the score labelled for that same session.
- A challenger must be clearly top for five consecutive sessions with at least a 0.5 percentage-point score gap before the confirmed state changes.
- The website receives only current aggregate state and at most 52 weekly group observations; company-level attribution remains in the private research package.
Historical taxonomy before its 2026-08-04 review date is retrospective. Prior research did not establish reliable post-confirmation alpha, so signal activation remains WITHHELD.
- Confirmed score
- +5.2%
- Top candidate
- 服务器与系统
- Candidate gap
- +0.6%
- Confirmed score
- +5.2%
- Top candidate
- 云与平台
- Candidate gap
- +1.9%
Confirmed leader and closest challengers
Trailing 21-session relative return, lagged one session · weekly observations · zero is the AI-chain composite
Frozen-universe coverage
The 49-company identification universe and 58-company attribution universe are checked separately.
Missing stable symbols: None
Confirmation discipline
A raw rank change is kept separate from a confirmed state change.
Latest fine-group comparison
Aggregate group evidence only; company-level attribution is withheld from the website
| Rank | Fine group | 21-session excess | Effective N | State role |
|---|---|---|---|---|
| 1 | 服务器与系统 | +8.5% | 5 | Challenger |
| 2 | 逻辑芯片与算力 | +7.8% | 5 | Challenger |
| 3 | 云与平台 | +7.4% | 5 | Challenger |
| 4 | AI软件与数据 | +5.3% | 7 | Confirmed |
| 5 | 网络互连与光通信 | +1.6% | 5 | Challenger |
| 6 | 存储与内存 | -5.1% | 5 | Challenger |
| 7 | EDA封装与测试 | -5.4% | 5 | Challenger |
| 8 | 电力散热与机电 | -7.8% | 5 | Challenger |
| 9 | 晶圆代工与设备 | -11.6% | 7 | Challenger |
Recent confirmed-state changes
Method limits
- Leadership migration is descriptive market structure, not validated alpha or a trading signal.
- The frozen historical taxonomy is current-taxonomy retrospective before its 2026-08-04 review date and therefore contains taxonomy look-ahead bias.
- Returns are split-adjusted price returns and exclude dividends.
- Company classifications are single primary value-chain assignments for this module only; they do not replace the website's overlapping Basket taxonomy.
- A confirmed state is known only after the one-session lag and five-session challenger confirmation rule.