From Megawatts to Homes: Why AI Data-Center Grid Queues Matter (Part A)
Interconnection to the energy grid is causing significant delays for AI data center buildouts
Summary
A single 150-megawatt (MW) artificial-intelligence (AI) data center running at a 90 percent capacity factor can consume roughly 1.18 terawatt-hours per year—enough electricity for about 112,000 U.S. homes.
In the PJM Interconnection (a regional transmission operator covering 13 states), large data-center projects now wait an average of five years (~1,825 days) to get grid-connection approval.
This post shows how to verify those numbers using ChatGPT (GPT-4.5), Python, and Microsoft Power BI.
1 | Challenge
Meta’s planned 200 MW AI campus in Ohio will draw more electricity than the entire city of Dayton so Meta has decided to build its own on-site gas plant to skip the grid queue.
2 | Datasets
Source 1: PJM Interconnection Queue (regional transmission operator)
Fields: QUEUE_DATE, ISA_EXECUTION_DATE, STATE_PROVINCE, MAX_SUMMER_CAPACITY_MW, PROJECT_NAME
Link: Use pdf attachment at the bottom of the post (interconnection_sample_queue.pdf)
Source 2: Energy Information Administration (EIA) Form-860
Fields: Nameplate MW, plant status, power-usage effectiveness (PUE)
Link: https://www.eia.gov/electricity/data/eia860/
Source 3: EIA Residential Electricity Use
Fields: Average household consumption (~10 500 kilowatt-hours per year)
Link: https://www.eia.gov/energyexplained/use-of-energy/electricity-use-in-homes.php
3 | Goal
Use GPT-4.5 to identify which U.S. states have the longest grid-connection queues for data-center projects over 50 MW, then convert project size into an easy “homes powered” metric.
4 | Foundational Math Box
Annual MWh = MW × 24 hours × 365 days × Capacity Factor
Homes Powered = Annual MWh ÷ 10.5 MWh per home
5 | Roadmap to Part B
Prompt GPT-4.5 → receive Python code.
Run code → calculate wait times and homes-powered equivalents.
Visualize in Power BI → choropleth map plus narrative.
Discuss business and policy impacts.
Ready to see the code? Jump to Part B that will be released on July 22, 2025