Example: AI Infrastructure Spending Drives Semiconductor Rally
Final Outcome Score
Failed
11/5/2024
6 months from November 2024
11/8/2024 - 5/8/2025
9/14/2026
Conclusions
Demo conclusion (staging only). Thesis incorrect. Semiconductor stocks did not reach the 20% gain we expected. Supply constraints eased more slowly than forecast, and none of the tracked data points hit their targets. The lesson: AI spending headlines did not translate into chip-stock gains within a six-month window.
Thesis Details
The Event
Major cloud providers announced $200B+ in combined AI infrastructure spending for 2025 during Q3 2024 earnings calls, signaling massive demand for AI chips and data center equipment.
Our Hypothesis
Semiconductor stocks focused on AI/data center will rally 20%+ over next 6 months as infrastructure buildout accelerates and supply chain constraints ease.
Our Rationale
Cloud providers (AWS, Azure, Google Cloud) are in arms race for AI infrastructure. Nvidia H100/H200 GPUs remain supply constrained but production ramping. AMD MI300 gaining traction as alternative. Memory manufacturers (Micron, SK Hynix) benefit from HBM demand. Historical capex cycles in semiconductors show 18-24 month runs. Taiwan Semi guidance indicates strong AI chip demand through 2025.
Tags
Measurable Data Points (4)
NVDA Stock Price
Price • yahoo finance (NVDA)
Target
Above 168
Current Value
Not yet measured
Last Updated
Never
Semiconductor ETF (SOXX)
Price • yahoo finance (SOXX)
Target
Above 60
Current Value
Not yet measured
Last Updated
Never
Cloud Provider Capex Growth
Percentage • manual
Target
Above 40
Current Value
Not yet measured
Last Updated
Never
HBM Memory Prices
Percentage • manual
Target
Above 20
Current Value
Not yet measured
Last Updated
Never
How we score: Each data point is evaluated as "met", "on track", "off track", or "failed" based on its current value vs. target. The overall confidence score is calculated as the percentage of data points that are "met" or "on track". When a thesis is closed, the final outcome score only counts "met" data points.