On August 17, Nvidia officially announced a partnership with SB Energy, a subsidiary of SoftBank, to secure land, power, and building (LPS) capacity at the PORTS-Pike technology park in Pike County, Ohio, specifically designated for deploying Nvidia AI computing infrastructure. OpenAI will serve as the tenant at this site, signing a 20-year lease agreement. This project is viewed by the market as a landmark event signaling that the AI infrastructure competition has fully expanded from "chips" to "physical resources." Jensen Huang posted a lengthy statement on X the same day, explicitly stating that "land, power, and buildings are the next critical strategic resources for AI factories."
The project's core structure and scale involve the PORTS-Pike site, located in southern Ohio on former federal land (the old Portsmouth gaseous diffusion plant), which will be developed, owned, and operated by SB Energy. OpenAI will act as a customer utilizing the site's capacity, which will be entirely built on Nvidia's full-stack DSX AI factory platform, encompassing GPUs, CPUs, networking, and infrastructure software.
Key parameters of the deal include an initial lock-in of approximately 4.25 IT-gigawatts of capacity, with Nvidia holding the option to expand by roughly 3.75 IT-gigawatts, bringing the total to about 8 IT-gigawatts. The overall energy plan calls for the construction of at least 10 gigawatts of new power generation capacity and an investment of at least $4.2 billion in regional grid upgrades. Capacity will come online in phases, with deployments expected to begin in 2028. Nvidia will directly invest $1.5 billion to acquire a stake in SB Energy. The project is expected to generate roughly 35,000 construction jobs through 2032 and 2,500 long-term operational roles, with both parties contributing a combined $80 million to a community fund.
Nvidia will provide credit support, primarily guaranteeing the land, power, and building construction for the initial 4.25 gigawatts, a figure that narrows earlier market discussions of a larger-scale backstop and focuses on the first phase. The computing capacity can be resold, reducing dependence on a single customer.
In his statement, Huang emphasized that AI factories are the core infrastructure of the AI era, where computing power converts energy and data into intelligence, and "computing power equals revenue." AI factories require a complete stack of critical resources: advanced chips, packaging, memory, networking, as well as land, power, and buildings. He stressed that Nvidia has already secured semiconductor resources through scale, long-term visibility, and supply chain partnerships, and is now applying the same discipline to lock in dedicated LPS capacity. For most cloud service providers and investment-grade companies, they can address their LPS needs independently. However, frontier AI labs are seeing demand grow far faster than their balance sheets and credit ratings can support, which is why Nvidia has chosen to step in and help them secure physical infrastructure, thereby safeguarding its own chip shipments and long-term revenue.
Huang estimated that the initial 4.25 gigawatts at PORTS-Pike, assuming roughly 1.5 million Nvidia GPUs per system generation, could represent a revenue opportunity of approximately $150 billion to $200 billion. If expanded and combined with other OpenAI commitments, the total opportunity could reach around $600 billion by 2030.
Market analysts interpret this arrangement as a sign that Nvidia is moving beyond simply "selling chips" toward "locking down the entire computing supply chain." Through credit support and equity investment, Nvidia lowers the financing barrier for frontier labs while ensuring the site exclusively uses its own hardware, reinforcing its ecosystem moat. For the market, this alleviates some earlier concerns about "circular financing," given that computing capacity can be resold and risk support is limited and phased, while strengthening the narrative around sustainable AI capital expenditure.
The project also emphasizes covering grid upgrade costs independently without passing them on to local electricity rates, alongside large-scale job creation and community investment, which helps mitigate local resistance to hyperscale data centers. Similar "chips plus LPS" bundling models could potentially be replicated in other regions, further intensifying the competition for power and land resources.