Key Takeaways
- Sector: Artificial Intelligence (AI), Digital Infrastructure.
- Geography: United States.
Analysis
Nvidia is reportedly in advanced discussions to provide a substantial financial backstop for a massive loan aimed at funding OpenAI's ambitious new data center campus. Sources indicate the potential deal could reach an astonishing $250 billion, earmarked for the development of a significant AI infrastructure hub in Ohio.
This colossal project, envisioned to deliver 10 gigawatts of computing power—equivalent to the energy consumption of approximately eight million households—is being developed by SB Energy Inc., a subsidiary of SoftBank Group PBC. SB Energy, which previously secured $500 million from OpenAI and has a track record including constructing a 900-megawatt solar facility for Google LLC, is responsible for both the data center construction and associated energy infrastructure.
The reported $250 billion figure is understood to cover lease and construction expenses for the Ohio facility. Crucially, this sum is separate from the cost of the advanced graphics processing units (GPUs) that will power the campus. Industry analysts estimate that the acquisition of these essential chips could more than double the total project expenditure, potentially pushing the overall investment beyond $500 billion.
The sheer scale of the proposed data center campus suggests it will house multiple facilities, drawing parallels to Meta Platforms Inc.'s Hyperion campus in Louisiana, which is designed to accommodate up to 11 buildings and will offer half the computing capacity of the planned Ohio site. The selection of Nvidia as a potential financial backer strongly suggests that its cutting-edge silicon will be the hardware of choice for OpenAI's next-generation AI computations.
With a project of this magnitude likely requiring several years for completion, OpenAI is expected to leverage not only current but also upcoming GPU architectures. Nvidia is preparing to launch its Rubin Ultra GPU in late 2027, integrated into a 72-accelerator rack system capable of delivering 15 exaflops of performance. This will be followed by the Feynman chip series in 2028, featuring advanced die-stacking technology and custom high-bandwidth memory, designed for unprecedented density and efficiency.
This potential financial commitment from Nvidia could significantly bolster SB Energy's prospects as it reportedly prepares for an initial public offering later this year, with a valuation anticipated to exceed $50 billion. A deal of this magnitude would underscore the critical interdependence between AI developers and semiconductor giants in building the future of artificial intelligence infrastructure.