Why AI uses so much power
AI work runs on dense racks of specialized chips that draw much more power than standard servers and give off more heat. A single AI campus can ask for as much power as a large town or a city.
What it means for the grid
Grid operators have raised their demand forecasts, and large load requests have grown fast. Not every request becomes a project: developers often ask about several sites for one planned campus, so the totals overstate what will be built. Forecasts differ widely. For current figures, use the reports of the grid operator, such as ERCOT.
What it means for siting
With many projects chasing the same grid, the sites that win are the ones that can get power soonest. That puts time to power, connection points and on-site generation at the centre of the search.
Questions people ask
How much power does an AI data center use?
It varies widely with the design and the size of the campus. Voltane does not publish a typical figure. Ask the developer for the requested megawatts and check grid operator reports for totals.
Will all the announced AI data centers be built?
Unlikely. Many requests are early, duplicated or speculative. Rules such as Texas SB 6 are meant to filter for serious projects.
Related
More glossary
- What is time to power?
- What is powered land?
- What is a powered shell?
- What is large load interconnection?
- What is a point of interconnection?
- What is an interconnection queue?
- What is queue position?
- What is ERCOT?
Need a site, or have one?
Tell us what you need and Voltane finds, reports on and prices candidate sites in ERCOT. Or bring a site you already have for a path-to-power report (by request for now).