The AI Capex Race Index

## What this measures

The AI Capex Race Index is the historical percentile of three lenses on whether the AI capex commitment is real or rhetorical: average AI mentions per call across the cohort, net capex guidance direction (% raised - % lowered), and average leader confidence. High readings mean companies are talking about AI a lot AND committing real capital AND speaking with conviction. Low readings mean the discussion is rhetorical, capital is being held back, or management is hedging.

## Why it matters

AI mentions alone are noise — every call mentions AI now. Capex direction is the financial truth — money flowing into infrastructure is real commitment. Confidence captures whether management actually believes what they are saying or hedging the bet. A high reading requires all three: high discussion + real capital + firm conviction.

## How to read it

When AI Capex Race is in the ELEVATED+ band, the build-out is structurally real. Stewards positioning portfolios around AI infrastructure have a coincident read. When the index falls while AI Mania is still high, the gap is the warning — narrative is outrunning the commitment.

## What it does not say

The index does not measure whether the capex pays off. The 1999 capex cycle in telecom infrastructure was also high on all three dimensions. The index tells you the commitment is real today — not whether it will earn returns. Read against the Mania Index: when AI Capex Race is high AND Mania is high, the build-out is being capitalized into stock prices simultaneously.

## History note

History recomputed 2026-09-21 on the rebuilt five-year Discourse corpus under the canonical discourse.* definition: readings before 2026-09-19 are weekly as-of recomputes that replace the history originally computed on the pre-rebuild corpus.

## Source

Computed daily by Palanor from 3 component signals in the lattice. Each component's weight, transform and current contribution is shown on the index's methodology page.