← All news·2026-06-22·4 min read

Microsoft’s 2GW AI Campus Has a Gas Power Plant

Microsoft is developing a 2GW AI datacenter campus in Pecos, TX that’s so power hungry it’s building its own gas turbine power plant powered by Chevron equipment instead of drawing power from the public grid.

microsoftai-infrastructuredata-centerenergy

Microsoft is developing a ~2GW AI datacenter campus in Pecos, TX that’s so power hungry it’s building its own gas turbine power plant powered by Chevron equipment instead of drawing power from the public grid. This is one of Microsoft’s largest capacity expansions ever and a harbinger of how AI infrastructure needs are impacting energy planning.

This project will be completed over a 5–7 year timeframe, with operations starting in approximately 2028. During periods of peak development, this project will support over 6,000 construction jobs, with hundreds of long term roles when the facility becomes operational. Total cost is undisclosed but ‘multibillion dollar’.

Why Microsoft stopped waiting for the grid

This project comes in the wake of dozens of cancelled US datacenter projects in 2026 due to community resistance to increased electricity costs and water usage. In order to preemptively address these concerns, Microsoft wrote an open letter to the people of Pecos prior to beginning construction. They promise the datacenter won’t increase electricity costs for Pecos residents, will utilize a closed loop cooling system that they claim uses ‘only a fraction’ of the water consumed by an average fast food restaurant each year, and will actually put more water back into the local ecosystem than it pulls out.

This private power generation solution directly addresses the issue of grid capacity limitations. The ERCOT power grid serves the greatest amount of electricity in the US and has frequently experienced capacity shortages. By generating power on site, these concerns are eliminated. Other companies are following suit as AI training and inference workloads skyrocket, forcing tech firms to think about power generation just like they do server deployment, designing and deploying their own solutions at scale rather than relying on external providers.

ℹWhat I'd actually do

For anyone building AI products, this is worth watching for one practical reason: massive compute buildouts like this typically translate into lower costs for running AI two to three years after capacity comes online. The enormous bet on AI infrastructure today is, in effect, a bet that AI gets cheaper for everyone building with it tomorrow. That's relevant if you're pricing a product or deciding which AI features are viable right now.

Source: the-decoder.com

Free course

Stop reading about AI — start building with it

The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.

Start free →
EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Founder & CEO, Dad4Kids