A stock-investment thesis across global AI infra markets
1,500+ analyst-hours of research compressed into 67 pages. What 38 AI infra CEOs/CFOs said across 6 countries — ~74,000 pages of real earnings calls distilled into a fact-checked, falsifiable, position-sizable thesis on the entire AI compute stack. Every claim cited to a dated management quote.
Concrete, fact-checked numbers from real management disclosures — each linked to a verbatim quote in the report:
This is not a chat-bot summary. It's the output of a multi-agent pipeline reading real public filings:
Earnings call recordings from 38 AI infra companies, 2024-06 to 2026-03, across US, China HK/A-share, Korea, Taiwan, Japan, Netherlands.
Structured digests of 55 Annual / Interim Report PDFs — ~13,000 pages — surfacing supply-side signals invisible to English IR.
Dated corporate events from A-share & HK public filings — product launches, contracts, lawsuits, joint ventures.
Verifiability: Every numerical claim and quote traces to a specific dated public source (company IR pages, SEC EDGAR, HKEx, Cninfo). Forward-looking statements explicitly flagged. Falsification criteria stated upfront.
Real pages from the report — page 1 (cover & corpus stats), page 5 (macro thesis · Section 1), page 8 (supply-chain coverage matrix):