The Maze: Nvidia is discussing a $3 billion investment in SB Energy, the SoftBank-backed developer tied to OpenAI's proposed data-center campus in Ohio. The move would go beyond selling chips or supporting project debt. Nvidia would own part of the company assembling the land, power, construction and financing around future computing demand. The supplier is not waiting for the AI factory to appear. It is helping finance the foundation—and positioning itself to collect when the machines arrive.
The fresh money buys ownership, not just reassurance. Nvidia has reportedly discussed investing about half of the $3 billion when the Ohio project agreement is signed and the other half alongside SB Energy's planned initial public offering. SB Energy could seek at least $5 billion in that offering, potentially as soon as September, but neither timing nor terms are final. The equity proposal is separate from talks over roughly $100 billion of credit support. A guarantee lowers lender risk; an equity stake gives Nvidia direct upside in the developer. That distinction is the new event.
SB Energy is the physical operating layer between models and machines. The company develops, finances, builds and operates power and data-center campuses. OpenAI and SoftBank invested $500 million each in January, while OpenAI selected SB Energy to build and operate a 1.2 GW campus in Texas. The partnership combines OpenAI's facility design with SB Energy's construction, energy and financing capabilities. An Nvidia stake would add the chip supplier to a structure already linking the tenant, developer and capital provider.
Ohio shows why balance sheets now matter as much as processors. OpenAI has been negotiating a long-term lease for a proposed SB Energy campus at the U.S. Department of Energy's Portsmouth site in Pike County. The plan targets 10 GW at full scale, with an approximately 800 MW first phase expected in 2028. Those are ambitions, not operating capacity. Yet the financing mechanism is straightforward: a long lease supports project borrowing; Nvidia credit support can make that borrowing cheaper; Nvidia equity strengthens the developer constructing the capacity.
The chip company is manufacturing demand around the chip. Nvidia and OpenAI previously outlined at least 10 GW of Nvidia systems, with Nvidia intending to invest progressively as capacity is deployed. Nvidia has also organized potential institutional-capital pools for customer infrastructure and may support selected projects' residual value. The strategy can reduce the balance-sheet disadvantage of developers competing with Amazon, Alphabet and Microsoft. It can also steer more campuses toward Nvidia hardware rather than rivals' accelerators or custom chips.
Vertical alignment comes with circular risk. Nvidia earns when customers buy its systems, may guarantee financing that lets them build, and could own the developer receiving the capital. If demand holds, the loop expands Nvidia's moat across hardware, financing and capacity. If utilization disappoints, the loop pulls project, credit and customer-concentration risk closer to Nvidia shareholders. A supplier helping create demand is strategy. A supplier carrying too much of that demand is exposure wearing a strategy badge.
Why it matters: Ecommerce and retail operators do not buy gigawatts, but they increasingly rent intelligence built on them: product discovery, service agents, merchandising, media and automation. Whoever finances the compute stack shapes its cost, availability and platform power. Nvidia's proposed SB Energy stake shows that the AI contest is moving below the model layer into ownership of power, construction and credit. Watch for definitive investment documents, the Ohio lease, final support terms and SB Energy's IPO filing before treating any reported figure as committed.
Sources: The Information | Reuters | SB Energy | Data Center Dynamics | NVIDIA
