PJM Data Center Siting

A developer-perspective NPV model for siting a hyperscale data center across twenty candidate locations in the PJM Interconnection footprint, with explicit treatment of policy risk and carbon pricing.

Motivation

This project extends prior work on AI infrastructure economics and FERC's evolving treatment of large-load interconnection. See the companion AI Infrastructure paper and FERC brief at lambcast.net for the framing this model operationalizes.

(Placeholder: expand once the literature survey is complete.)

What this project produces

• A Python NPV model evaluating twenty PJM candidate sites under a configurable cost stack (energy, capacity, interconnection, property tax, carbon, policy risk).
• Pre-computed sensitivity results across facility size, discount rate, and carbon price.
• A React/Vite frontend, deployed to Vercel, exposing the sensitivity surface interactively.
• A writeup posted to lambcast.net.

Repository structure

docs/         Master document, methods notes, decision log, references
data/         raw/ (gitignored), processed/, reference/
src/pjm_siting/   Python package
scripts/      Runnable data pulls and analysis entry points
notebooks/    Exploration
tests/        Unit tests
frontend/     React/Vite app (deployed to Vercel)
results/      figures/, tables/, precomputed/

How to run

Requires uv and Python 3.12.

uv sync
uv run python scripts/<entry_point>.py

(Placeholder: list actual entry points as scripts are written.)

Key decisions

See docs/decisions.md for the running decision log.

References

See docs/references.bib.

Citation

(Placeholder: add citation block once the writeup is published.)

License

(Placeholder: license to be selected.)