Nvidia Puts $1.5 Billion Into SB Energy, Locking In Exclusive Compute Role at OpenAI's Ohio Data Center
Nvidia announced on August 17, 2026 that it will invest $1.5 billion in SB Energy, a data center developer backed by SoftBank and OpenAI. The deal makes Nvidia the sole supplier of computing hardware at OpenAI's planned Ports-Pike data center near Cincinnati, Ohio, and includes up to $105 billion in credit from Nvidia to help finance construction. In return, Nvidia secures access to up to 8 gigawatts of AI computing capacity at the Ohio campus — roughly the power draw of eight nuclear reactors, or enough to serve millions of simultaneous AI queries TechCrunch Reuters.
SB Energy, whose existing investors include SoftBank and OpenAI, will build a 9.2-gigawatt natural gas power plant on the site. The land belongs to the U.S. Department of Energy and was previously used to enrich uranium for the nuclear weapons program and Navy submarines. The power plant alone is expected to cost $33 billion TechCrunch.
The Ports-Pike facility is designed to scale from an initial 4.25 gigawatts up to 8 gigawatts. It is part of Stargate, a broader initiative from SoftBank and OpenAI that aims to invest $500 billion over four years in new AI infrastructure. Each company committed $19 billion to the joint venture Reuters SoftBank. In September 2025, OpenAI, Oracle, and SoftBank announced five U.S. AI data center sites under the Stargate banner, with over $400 billion planned over three years and a full $500 billion program SoftBank Reuters. The international footprint includes Stargate UAE, launched by SoftBank's Global Tech Alliance to provide 5 gigawatts of AI data center capacity powered by nuclear, solar, and other sources SoftBank.
Before this finalized $1.5 billion commitment, reporting indicated Nvidia was in talks to invest as much as $3 billion in SB Energy and to provide a roughly $250 billion financial backstop for OpenAI to lease the Ohio project The Information WSJ. SoftBank separately committed to investing $3 billion in an Ohio factory for an OpenAI data center Reuters and announced a $22.5 billion investment plan in OpenAI in August 2025 SoftBank. SoftBank funded these AI investments partly by selling its entire stake in Nvidia — $5.8 billion in stock — in November 2025 CNBC.
Nvidia and OpenAI formalized a strategic partnership in September 2025 to deploy 10 gigawatts of Nvidia systems. Under that agreement, Nvidia intends to invest up to $100 billion in OpenAI progressively, releasing funds as each gigawatt comes online Nvidia.
The broader context here is the unusual flow of capital among the parties involved. SoftBank sold its Nvidia shares to fund AI infrastructure ventures like Stargate and OpenAI. Nvidia is now taking an equity stake in SB Energy and extending massive credit to OpenAI, effectively recycling capital through a chain of interdependent entities. Nvidia's willingness to act as both hardware vendor and financier of last resort for its largest customers concentrates operational and financial risk on its own balance sheet.
The decision to anchor a flagship AI data center on 9.2 gigawatts of natural gas generation also enters a shifting cost environment. Building natural gas power plants has become 66% more expensive in the last two years, according to BloombergNEF TechCrunch. Forecasts suggest the combination of new gas power plants and export markets could triple natural gas prices in some parts of the United States TechCrunch. Committing tens of billions of dollars to gas-fired generation locks operators into fuel price trajectories that may diverge from expectations.
The Ohio site adds its own layer of complexity. Building a large AI data center and a massive gas power plant on former Department of Energy nuclear enrichment land adds regulatory and environmental oversight to an already intricate development timeline. Site remediation, security clearances, and federal land-use negotiations are distinct variables from standard greenfield construction.
Over the long arc, the demand for inference capacity (running trained AI models to answer user queries) and model training at this scale is real. The industry has weathered infrastructure buildouts before, during the cloud transitions of the 2010s and the mobile network expansion before that. What differs now is the vertical integration — a single small group of companies owns the computing hardware, the facilities, the power generation, and the AI models themselves. The Ports-Pike project will be an early test of whether that tightly coupled model can deliver on its capacity targets.
