Nvidia is underwriting the next leg of the AI buildout with its own balance sheet, committing up to $105 billion to a giant OpenAI campus in Ohio as competitors chip away at its silicon advantage. The move turns the chipmaker into a infrastructure financier, using $48.5 billion of quarterly free cash flow, up eighteen-fold in three years, to guarantee demand for its own products.

The capital moat

Nvidia's technology lead has narrowed. Advanced Micro Devices and Google have closed enough of the performance gap that the company's most durable asset is now its credit capacity. Ram Bala of Santa Clara University's Leavey School of Business describes the posture as paranoid: dominant but terrified of losing ground. Nvidia declined to comment on the characterization.

The Ohio structure

The agreement anchors a 20-year lease between OpenAI and SB Energy, a SoftBank affiliate developing the PORTS-Pike Technology Campus in Pike County. Nvidia is investing $1.5 billion directly in SB Energy and backing roughly four gigawatts of phased development through 2030, covering portions of lease and power costs plus a specified residual-value commitment. The source does not disclose the size of that residual commitment or whether it functions as a put option, a guarantee, or something else.

The balance sheet math

The numbers are staggering by any standard. Marketable equity securities hit $30.2 billion last quarter, more than double the $12.9 billion held a year earlier. In February Nvidia put $30 billion into OpenAI itself, a customer that trains on the Vera Rubin system. In May the board raised the quarterly dividend to 25 cents from one cent, authorized an $80 billion buyback, and pledged to return half of free cash flow to shareholders. The Ohio backstop sits alongside last week's $500 billion financing pact with six Wall Street firms to fund GPU purchases, a facility the company has not sized in terms of its own capital at risk.

Wall Street's view

Cantor analysts dismissed the circular-revenue critique on Monday, reiterating a buy rating and calling the deal a signal that the AI investment cycle will be elongated and durable. They frame the financial engineering as a competitive moat. Jensen Huang, posting on X, argued that frontier labs have extraordinary compute demand but lack the decades-long infrastructure contracts and investment-grade financing capacity to secure AI factories independently. A week earlier he stood on CNBC with six bank CEOs to debut GPUs as a new asset class.

What to watch

The residual-value commitment on the Ohio site is the quiet term that matters. If OpenAI's fortunes turn, Nvidia's exposure crystallizes. The company is effectively writing put options on its own customers' viability, funded by cash flow that has grown faster than any semiconductor cycle in history. Whether that is prudent underwriting or vendor finance at scale will not be clear until the first data center opens in 2028.