SoftBank's $33 Billion Power Play: A Bet on Gas to Fuel the AI Boom

Masayoshi Son’s SoftBank Group is making a staggering wager on America’s energy future. The Japanese investment giant has confirmed plans to build a $33 billion natural gas power plant in the United States, an investment of historic scale that underscores a new industrial reality: artificial intelligence runs on electricity, and lots of it.

The project, first reported by TechCrunch, is a direct response to an acute power shortage driven by exploding AI data center demand. Rather than buying power from the grid, SoftBank is building its own dedicated supply. This vertical integration aims to prevent its sprawling AI ambitions, including the massive Stargate initiative with OpenAI and Oracle, from being hamstrung by a lack of reliable juice.

The numbers are staggering. A typical large gas plant costs about $1-2 billion. SoftBank’s budget suggests a facility of unprecedented output, capable of generating power for millions of homes—or, more pertinently, for legions of power-hungry AI servers.

The choice of natural gas is both technical and political. Gas plants provide constant, dispatchable power, a necessity for data centers that cannot tolerate blackouts. The investment also aligns with the current administration’s energy priorities. Following a 2025 announcement with then-incoming President Donald Trump, SoftBank pledged $100 billion in U.S. investment; this plant consumes a third of that commitment, positioning the company favorably for regulatory approvals.

SoftBank’s move is the most aggressive signal yet that the tech and energy sectors are colliding. While Microsoft, Amazon, and Google seek nuclear or renewable deals, Son is opting for direct control. The strategy mirrors Tesla building its own batteries, but at a colossal scale.

Financing the project will involve complex debt structures and likely federal incentives. It will also invite fierce opposition from environmental groups, as a plant this size would generate millions of tons of CO2 annually.

Ultimately, this is more than an energy project. It’s a physical monument to Son’s belief that controlling the power behind AI is as important as developing the algorithms. Its success hinges on executing one of the largest construction projects in U.S. energy history—a formidable task for a company better known for writing checks than building power plants.