A live research project that tracks and analyzes real-time AI adoption disclosure across the S&P 500.
Independent research, direct to your inbox.
Live data tracking and analysis. Deep research that cuts through consensus. Evidence-backed insights.
Across 301 S&P 500 earnings calls, NVIDIA — a chip vendor — is the most-named AI partner at 22. OpenAI (21) is tied with Anthropic (21) for second. Google sits at 19.
Everyone says AI is transformative.
What are companies actually telling investors under oath, and how specific are they?
AI commentary on earnings calls is dense, narrative-heavy, and hard to compare across companies. This project runs every S&P 500 earnings transcript through a structured extraction pipeline, producing a comparable dataset of every AI claim management makes — speaker, role, section, specificity score (1–5), and the verbatim quote.
The dataset answers questions the headlines can't: who is quantifying real AI revenue (vs. forward-looking metrics), what analysts ask that management refuses to answer, and how AI exposure varies sector to sector. Updated as new calls land.
Three buckets of AI exposure.
Across 301 companies, the AI conversation splits cleanly along the dominant type of exposure: who's supplying the buildout, who's using AI internally, and who's selling AI as part of their product. Each bucket maps to a distinct set of sectors.
Suppliers to the buildout
Selling power, real estate, fuel, and physical materials to the AI buildout. Not building AI — landlording it.
- Real Estate
- Energy
- Utilities
- Materials
Internal productivity
Using AI inside the company for productivity. No AI products to sell. Largely banks, telecom, and packaged-goods companies.
- Communication Services
- Consumer Staples
- Financials
AI in products
AI inside the products being sold to customers — software, medtech, industrial automation. The cohort with the most quantified revenue stories.
- Information Technology
- Health Care
- Industrials
What 301 calls reveal so far.
Each finding is sourced to the underlying transcript. Click into the full report (forthcoming) for verbatim quotes, methodology, and the full per-company breakout.
Of 23 companies disclosing AI revenue, only Broadcom reports it as a GAAP segment — $8.4B, +106% YoY.
NVIDIA named as AI partner by 22 S&P 500 companies — alone at the top. OpenAI (21) tied with Anthropic for second.
Disclosure gaps — analyst questions management would not quantify.
MSCI's CFO: AI is so embedded across the product line that breakouts are meaningless.
Auditable. Cross-validated. Versioned.
- Single-pass extraction via Claude Sonnet 4.6. Every claim must be backed by a verbatim quote that exists in the source transcript.
- Cross-model validation on a 10% random sample using Claude Opus 4.6. Aggregate agreement: 80% on substantive judgments.
- Schema versioned with semver. Old extractions are never deleted; methodology changes are logged in the public changelog.
- Quality gates run on every extraction — quote integrity, scope coherence, and schema compliance. Flags surface for manual review.
Get the next finding before it makes the headlines.
The chart above updates as new earnings calls land. Each significant finding ships to subscribers first — sourced, sized, and on the record.