Meta's AI Data Center Ambitions Require 10 New Natural Gas Plants — Enough to Power South Dakota

Meta's Hyperion AI data center, the company's next-generation compute facility, will require electricity from 10 new natural gas power plants to meet its energy demands — a figure that TechCrunch contextualizes bluntly: that's enough power generation capacity to supply the entire state of South Dakota.

The Scale Is the Story

AI data center power consumption has become one of the defining infrastructure challenges of the current technology cycle, but individual announcements can be hard to contextualize. The Hyperion figure cuts through the abstraction: 10 new natural gas generation plants is not incremental grid consumption. It is new generation infrastructure purpose-built for a single corporate customer's AI compute needs.

For comparison, a typical new combined-cycle natural gas plant generates roughly 500-800 megawatts. Ten plants at that scale implies Hyperion's total power draw could approach 5-8 gigawatts at full utilization — a figure that would make it one of the most power-intensive data center facilities on the planet, likely exceeding the combined consumption of several existing hyperscale campuses.

Why Natural Gas, Not Renewables?

The natural gas choice reflects a practical constraint: renewable energy — solar and wind — cannot provide the 24/7 baseload power that AI training and inference clusters require. Data centers running continuous GPU workloads cannot accommodate the intermittency of renewable generation without massive battery storage infrastructure that doesn't yet exist at this scale. Natural gas peakers and combined-cycle plants can spin up and down on demand, making them the default choice when power reliability and density are the primary constraints.

Meta, like other hyperscalers, has made public commitments to renewable energy matching and net-zero carbon goals. The Hyperion natural gas plants will need to be squared with those commitments — likely through renewable energy credits, carbon offsets, or long-term power purchase agreements with renewable generators elsewhere on the grid.

Broader Industry Implications

Hyperion is not an outlier. Microsoft, Google, Amazon, and Oracle are all announcing data center expansions at similar scales, and all are grappling with the same power density problem. The AI buildout is becoming one of the primary drivers of new electricity generation investment in the United States — and that investment is flowing toward natural gas, not renewable energy, in the near term.

For energy markets, grid operators, and AI policy makers, the Hyperion announcement is a preview of the infrastructure commitments the AI industry is making — commitments that will take years to build and decades to depreciate.