Meta has ordered 10 gas-fired power plants for its Hyperion AI campus in northeastern Louisiana, more than tripling its initial generation plan. The move supports a $10 billion data center investment announced in December 2024, spanning 2,250 acres. This is one of the clearest signals yet that hyperscalers are abandoning grid dependency in favor of behind-the-meter, dispatchable generation to meet AI training and inference loads.

The Players

Meta is the buyer and developer. The specific turbine supplier and EPC contractor have not been disclosed, but the scale suggests involvement from GE Vernova, Siemens Energy, or similar Tier 1 OEMs. The campus is located in rural Louisiana, likely targeting low land costs, favorable tax treatment, and proximity to gas infrastructure.

The Numbers

  • $10 billion total campus investment
  • 2,250 acres
  • 10 gas-fired power plants (capacity not disclosed, but likely 50–100 MW each based on typical hyperscaler co-location sizing)
  • Original plan called for ~3 plants; this is a 3x+ increase

Assuming 75 MW per plant, this could represent 750 MW of behind-the-meter generation—enough to power a top-tier AI training facility without touching the local utility’s transmission system.

So What?

This is the new normal. Hyperscalers are no longer waiting for utilities to build out transmission or navigate interconnection queues. They’re buying dispatchable generation directly, and gas is the only technology that can deliver at scale in the 18–24 month window AI infrastructure demands.

For power developers: if you’re not in conversations with Meta, Google, Microsoft, Amazon, or Oracle about co-located gas generation, you’re missing the biggest deployment cycle in a decade. These deals are being done off-market, often with preferred developer relationships.

For utilities: this is a warning. Hyperscale load growth was supposed to be your next decade of capex justification. Instead, it’s moving behind the meter. You’ll see the gas burn, but not the revenue.

For gas turbine OEMs and EPCs: this is your moment. Frame agreements with hyperscalers are worth chasing aggressively.

Source: MSN / Fortune, March 28, 2026