power grid

AI data centers use electricity for GPU compute, cooling, and grid conversion losses, not just chips. See real training vs. inference power figures.

On September 1, 1859, British astronomer Richard Carrington watched a white-light flare erupt from the sun's surface — the first solar flare ever recorded by a human being. Within 17 hours, a coronal mass ejection carrying several billion tons of magnetized plasma slammed into Earth's magnetosphere at roughly 2,400 km/s (5.4 million mph), triggering the largest geomagnetic storm in recorded history. Telegraph operators received electric shocks from their equipment. Aurora lit the skies over Cuba, Hawaii, and Colombia. If the same event happened today, a 2008 National Academies report estimated the damage at nearly $2 trillion in the United States alone.

As electric vehicles proliferate globally, the power grid faces an unprecedented challenge: peak charging demand from 100 million EVs will add 200–400 GW of simultaneous load equivalent to 50–100 large coal plants coming online at once. Grid operators must manage demand response, deploy smart charging, enable vehicle-to-grid (V2G) technology, and upgrade distribution infrastructure from local substations to generation capacity. The real numbers show both feasibility and urgency — nations planning 50% EV adoption by 2035 are already implementing $100+ billion grid modernization programs, and the grid's ability to absorb uncontrolled charging without brownouts hinges on sophisticated load balancing within the next five years.

From Edison's Pearl Street DC station in 1882 to the AC war of currents, Insull's monopoly model, and today's smart grid, the electrical power grid shaped every hour of modern life.

How electricity works: voltage, current, resistance, AC vs DC, power generation, the grid, and why birds don't get electrocuted.

Power grids are already too complex for humans to optimize in real time. AI could run them perfectly — until something goes terribly wrong.