# Market Thesis Research Bundle

Question: Given AI demand is shifting deeper into the stack, will a major foundry, advanced packaging house, or memory supplier raise 2027 capex or capacity guidance because HBM, advanced packaging, or interconnect bottlenecks are now limiting shipments more than raw GPU demand?

What this bundle is: a reasoning and monitoring scaffold. It organizes public evidence into observations, claims, uncertainty branches, thresholds, and a watch plan.

What this bundle is not: primary evidence, live market data, trade advice, or a substitute for official, live, or current web sources.

Core tension: The assessment turns on unresolved evidence and live-source refresh.

Current inference to verify: Refresh live sources before using this assessment. Treat this as a hypothesis that must be refreshed against live official sources, not as a signal.

How to use: read `source_priority.json` first, refresh sources in `live_verification_plan.json`, then use `fact_inference_split.json`, `thresholds.json`, and `watch_schedule.json` to decide what changed. Do not infer buy/sell/hold, position sizing, execution, or asset-price direction from this artifact.