The Story You’re Not Hearing About AI Data Centers | Ayșe Coskun | TED

Our Take

Ayșe Coskun's TED talk, flagged for us by a reader, makes a genuinely interesting technical argument: data centers could shift and shape their compute loads to act like grid-balancing batteries, potentially easing the very strain they're creating. It's a real engineering insight, and credit to her and TED for putting the grid physics in front of a general audience instead of just more AI hype.

But here's the gap between a stage in San Francisco and a substation in rural Ohio or Georgia: 'could become' flexible assets is a research direction, not a permitting condition. Right now, the interconnection queues, tax abatements, and water withdrawal permits being signed this year assume data centers as steady, maximum-draw customers — not cooperative grid partners. Utilities are socializing transmission upgrade costs onto residential ratepayers today, banking on demand-response software that doesn't exist at scale yet. The idea that AI can fix the grid problem it's causing is appealing precisely because it lets everyone defer the harder question of who pays for buildout happening right now.

We're not against the technology — flexible load is a good thing if it materializes and utilities are required to use it. We are against treating a hopeful TED talk as a reason to wave through today's deals with no strings attached. Check our facility map to see what's being built near you under current terms, and hit take action if you want to push your utility commission to demand flexibility clauses now, not someday.