The term sheet, if it closes, would be unlike anything in corporate finance: NVIDIA negotiating to back OpenAI’s $500 billion Ohio data center project with $250 billion to $322 billion in financing guarantees, according to The Wall Street Journal. A chip maker guaranteeing the debt of its largest customer at that scale has no precedent. The guarantees would make NVIDIA’s balance sheet the collateral behind the most expensive construction program in the history of the technology industry.
The Ohio project is part of the broader Stargate initiative that OpenAI and its partners announced with White House backing, and the $500 billion figure matches the program’s publicly stated scale. NVIDIA’s role would be to stand behind the financing, in effect betting its own cash — which exceeds $100 billion — that the data centers get built and that the chips inside them come from NVIDIA. The company has not commented beyond the Journal’s reporting, and people familiar with the negotiations say the terms remain in flux.
The guarantee structure reflects a fundamental shift in how the AI supply chain is financed. Until recently, the biggest customers borrowed against their own balance sheets to buy chips. Now the chip maker itself is becoming a financier, because the scale of demand — and the speed at which it is arriving — exceeds what customers can fund on their own. Jensen Huang has said repeatedly that the chip industry needs to grow tenfold in a decade to support 100 billion AI agents; a guarantee of this size is the corporate equivalent of putting that conviction on paper.
The same day brought a second move: NVIDIA announced a $1 billion investment in Naver, the Korean internet giant, for a 4.5 percent stake, alongside a plan with Brookfield to build a 200-megawatt AI factory, targeted for operations in 2028. Naver’s shares jumped 10 percent in a single session, a market verdict on what NVIDIA’s endorsement means for a company that has spent years building its own AI stack against foreign competition.
The Naver deal is strategic on two levels. It gives NVIDIA a partner in the Korean market, where Naver runs the dominant search and messaging platforms and controls a trove of Korean-language data that global AI companies have struggled to access. And it extends NVIDIA’s pattern of building regional AI infrastructure with local champions rather than competing with them — the same approach the company has used in India, Japan and the Middle East. The 200-megawatt factory, co-owned with Brookfield, gives the partnership a physical base for serving Korean and regional customers.
The third piece, also reported by the Journal, is a long-term strategic partnership with Safe Superintelligence Inc., the company founded by Ilya Sutskever after his departure from OpenAI, valued at $32 billion. SSI gets early access to NVIDIA’s Vera Rubin platform, the next-generation chip line designed for the largest training runs. For SSI, the deal means computing capacity that money alone could not buy. For NVIDIA, it means a seat at the table of one of the most ambitious — and most secretive — efforts in the field.
Sutskever’s SSI has described its goal in almost religious terms: building an AI system that is not just powerful but safe, with safety engineered from first principles rather than patched on later. The company has raised capital at escalating valuations while revealing almost nothing about its technical approach. NVIDIA’s willingness to give it early hardware access suggests the chip maker is betting that SSI’s approach — whatever it is — will need enormous amounts of compute, and that NVIDIA will be the supplier.
Taken together, the three moves describe a company behaving less like a chip vendor and more like the central bank of the AI economy. NVIDIA is financing customers it believes in, taking equity stakes in regional partners, and reserving its best hardware for the labs it thinks will define the frontier. Each move carries risk — a guarantee structure that size could, in a downturn, become a liability rather than an asset — but the company’s calculus is simple: the AI buildout is happening, NVIDIA is the bottleneck, and the bottleneck gets to set the terms.
The three transactions, announced within days of each other, will each draw scrutiny — the guarantee from rating agencies, the Naver stake from Korean regulators, the SSI partnership from competitors watching for allocation advantages. NVIDIA’s response to all three questions is expected to be the same: the moves are investments in demand for the one product the company controls.
Huang’s tenfold-growth line frames the ambition. If the world is really heading toward 100 billion AI agents, the current chip supply is a rounding error, and the winners will be whoever locks in capacity, capital and customers now. The $250 billion question — literally — is whether the guarantees create an obligation NVIDIA can carry, or an exposure that will come back to haunt it. For now, the company is acting like a firm that believes the buildout is the surest bet it has ever seen, and it is pricing its own balance sheet accordingly.
