🧠 AI Bubble Pressure Score (AIBPS)

A data-driven, multi-pillar early-warning system estimating pressure building inside the modern AI-driven economic cycle.
Combines market signals, credit conditions, hyperscaler + semiconductor capex, infrastructure build-out, enterprise AI adoption, and sentiment intensity into one 0–100 composite.

AIBPS answers a simple question:

“How stretched is the AI ecosystem right now compared to its own historical patterns?”

---

🚀 Live Dashboard

👉 Streamlit App: https://aibps-v0-1.streamlit.app

Updates automatically via GitHub Actions (daily at ~07:00 UTC).

---

📦 Project Structure

aibps-v0-1/
├── app/
│   └── streamlit_app.py               # Interactive dashboard
├── src/aibps/
│   ├── compute.py                     # Composite assembly, normalization, weighting
│   ├── normalize.py                   # Rolling Z, percentile, sigmoid transforms
│   ├── fetch_market.py                # Market data (yfinance)
│   ├── fetch_credit.py                # Credit spreads (FRED)
│   ├── fetch_macro_capex.py           # Hyperscaler, semiconductor, fab, infra CAPEX
│   ├── fetch_infra.py                 # Infrastructure build-out (electricity, cooling, grid)
│   ├── fetch_adoption.py              # Enterprise software, digital labor, cloud adoption
│   ├── fetch_sentiment.py             # News/text sentiment (synthetic + API-ready)
│   └── config.yaml                    # Pillar definitions, weights, normalization settings
├── data/
│   ├── raw/                           # Raw pulls from APIs
│   └── processed/                     # Normalized monthly indicators and composite
├── .github/workflows/
│   └── update-data.yml                # Automated daily refresh
├── docs/
│   ├── METHODS.md                     # How each pillar is built
│   ├── OVERVIEW.md                    # Conceptual framing + economic logic
│   ├── ARCHITECTURE.md                # Dataflow + processing pipeline
│   └── REFERENCES.md                  # Literature & citation support
├── requirements.txt
└── README.md

---

⚙️ How It Works

1. Pillars

AIBPS is composed of six equally weighted pillars (weights configurable):

Pillar	What It Measures	Source Examples
Market	AI-tilted asset returns & valuations	SOXX, QQQ, NVDA basket
Credit	Funding stress & corporate risk appetite	FRED HY OAS, IG OAS
Capex	Hyperscaler + semiconductor + fab investment cycle	FRED PNFI, hyperscaler CSV
Infrastructure	Power, grid build-out, cooling & data-center density	FRED ELECGEN, IPN313
Adoption	Enterprise software, digital labor, productivity	FRED productivity, labor costs
Sentiment	Media hype, keyword fever, attention cycles	NLP-ready sentiment pipeline

Each pillar is normalized using either:

• Rolling Z-score
• Rolling Z-Sigmoid
• Percentile rank
• (configurable in config.yaml)

All subcomponents → normalized → averaged → pillar score → composite.

---

📈 Composite Score (0–100)

AIBPS represents the relative extremity of AI-related conditions:

• 0–25 → 🔵 Cold / Undervalued / Early-cycle
• 25–50 → 🟢 Stable / Neutral
• 50–75 → 🟡 Elevated / Late-cycle
• 75–90 → 🟠 Stretched / Fragile
• 90–100 → 🔴 Bubble conditions historically seen before unwinds

Bands adapt to the selected normalization scheme (default: rolling Z-sigmoid).

---

🔄 Automatic Daily Refresh

A GitHub Actions workflow:

• Pulls fresh data (yfinance + FRED + CSV hyperscaler capex)
• Normalizes using rolling windows
• Recomputes composite & pillars
• Commits new artifacts into /data/processed/
• Streamlit automatically reloads them

You can inspect the workflow at:

.github/workflows/update-data.yml

---

🧪 Local Development

Install dependencies

pip install -r requirements.txt

Run the composite builder manually

python src/aibps/compute.py

Run Streamlit locally

streamlit run app/streamlit_app.py

---

📊 Interpretation Guide

📉 What a Rising AIBPS Means

A rising score typically reflects:

• Rapidly accelerating hyperscaler / semiconductor capex
• Tightening credit conditions
• Surging market valuations
• High media attention or hype intensity
• Infrastructure bottlenecks (power, cooling, grid)
• Low-friction AI adoption in enterprises

📈 What a Falling AIBPS Means

• AI cycles cooling off
• Funding risk improving
• Capex plateauing / deferred
• Sentiment moderation
• Market de-risking

---

📚 Documentation

All full documents stored in /docs/:

• OVERVIEW.md – economic logic, comparisons to dot-com & housing bubbles
• METHODS.md – detailed pillar construction + normalization math
• ARCHITECTURE.md – ETL/dataflow diagrams
• REFERENCES.md – peer-reviewed citations (APA 7)

---

🤝 Contributing

Pull requests welcome!
Please open an issue before adding new pillars, subcomponents, or APIs.

---

📄 License

MIT License — free to fork, modify, and build upon.

---

🙋 Contact

Maintainer: Matt Monnot, PhD
Industrial–Organizational Psychologist | People Analytics | Applied Econometrics
GitHub: https://github.com/mjmonnot