Nvidia agreed on August 17, 2026 to back a massive new data center campus in Ohio, known as the PORTS-Pike Technology Campus, with as much as $105 billion in financing, structured as a credit guarantee rather than a direct cash outlay. SB Energy will build and own the site, which OpenAI will lease for 20 years, and Nvidia is separately investing $1.5 billion directly into SB Energy as part of the arrangement. The initial phase targets 4.25 gigawatts of compute capacity with an option to expand to 8 gigawatts total, and the facility will run exclusively on Nvidia GPUs, potentially around 1.5 million chips once fully built out, which Nvidia estimates could translate into $150 to $200 billion in GPU revenue through 2030 as successive hardware generations get deployed there. The deal follows a now-familiar pattern in AI infrastructure financing: rather than OpenAI or a cloud provider fronting the capital for a new data center outright, the chip supplier itself underwrites the debt, effectively guaranteeing its own future revenue by financing the buildout that will consume its products. That is a meaningfully different risk structure than a typical vendor relationship, and it means Nvidia's balance sheet is now directly exposed to whether OpenAI's compute demand materializes as projected over the next two decades. For anyone tracking AI infrastructure economics, this is a concrete data point on just how capital-intensive the current scaling race has become, and how deeply intertwined chip suppliers, hyperscalers, and model labs now are, since a single company's growth trajectory is increasingly underwritten by the very vendor selling it hardware.