Nvidia has downsized its plans for a $250 billion guarantee connected to an OpenAI data center project, the Wall Street Journal reports. The figure places the story among the largest single financial commitments ever contemplated in the AI infrastructure market — and its reduction signals a measurable shift in how one of the industry's dominant suppliers is pricing risk.

The report, published as an exclusive by the Wall Street Journal, gives no revised dollar amount for the scaled-back arrangement. What it does establish is direction: the commitment Nvidia is now prepared to stand behind is smaller than the $250 billion originally under discussion.

Why does the $250 billion number matter?

A guarantee of that size would function as a backstop across the OpenAI data center buildout — the physical compute layer that trains and serves frontier models. For suppliers of AI accelerators, guarantees of this kind serve a dual purpose. They secure long-term demand for high-margin hardware. They also concentrate counterparty risk on a single customer whose spending capacity depends on continued capital inflows and revenue growth.

At $250 billion, the arrangement under discussion would have ranked alongside the largest infrastructure financing structures in the technology sector. Any reduction in that exposure changes the risk calculation for Nvidia's balance sheet, for OpenAI's capacity planning, and for the lenders and equity investors evaluating the project.

What does the downsizing signal?

The move suggests Nvidia is applying tighter discipline to how much of its capital and credit it will tie to a single AI customer's expansion. The company has spent the past two years converting unprecedented demand for its data center GPUs into record revenue. Commitments of hundreds of billions of dollars work in the opposite direction: they convert future revenue into present-day contingent liability.

For OpenAI, a smaller guarantee narrows the financing envelope available for data center capacity. The company's compute roadmap depends on large-scale facilities coming online on schedule, and vendor backstops have become one of the mechanisms that make those projects bankable for outside investors.

What remains unknown?

The Wall Street Journal report does not specify:

• the revised size of the guarantee;
• which specific data center site or sites the arrangement covers;
• the reasons Nvidia gave, if any, for reducing the commitment;
• whether OpenAI has secured alternative backstops for the difference.

Each of those details will determine whether this marks a routine renegotiation or a structural retreat from the vendor-financing model that has underwritten the current AI buildout cycle.

Who is affected?

Three parties carry direct exposure to the outcome. Nvidia, whose data center segment drives the majority of its revenue, adjusts the size of a contingent commitment measured in the hundreds of billions. OpenAI faces a tighter guarantee structure for the compute capacity its model roadmap requires. Investors and lenders evaluating AI infrastructure projects now have a concrete data point — $250 billion reduced — showing that even the largest supplier in the sector will cap its exposure.

The reduction does not halt the project. It recalibrates who bears the risk of building at this scale, and by how much.