I strongly believe in the transformative power of artificial intelligence (AI) and the incredible advancements it brings to our world. But I also believe the energy conversation around AI is being handled badly, and the consequences will land on ordinary people. Data centers do not sleep.
They run 24 hours a day, 365 days a year. They do not follow the human rhythm of morning ramp-up, lunchtime lull, and evening shutdown. They need a flat, constant load.
The International Energy Agency (IEA) reports that electricity demand from data centers soared 17 percent in 2025, well outpacing global electricity demand growth of 3 percent. The IEA's Electricity 2026 report projects global power demand to grow 3.6 percent annually through 2030, a sharp acceleration from the 2.8 percent rate of the past decade. That demand has to be met.
Solar and wind are essential to a cleaner energy system, but their variable output means that supplying data centres around the clock requires storage, grid flexibility or additional sources of firm power. The IEA estimates that meeting data centres' reliability requirements with on-site gas-fired generation can require 30-70 percent more capacity than their electricity demand. Gas turbines create noise, emissions, and local pushback.
If a data center connects to a constrained grid, the additional generation and infrastructure investment it requires can raise electricity costs for other consumers. There are two sources that can meet flat, 24/7 demand. I am a geoscientist and an engineer.
I have spent more than 20 years across subsurface development, defense systems, and offshore operations. I have looked at the nuclear timeline. My assessment is that micro-nuclear remains years away from commercial deployment at the scale data centers need.
Extract — continue reading at the source.
