Nvidia Is Becoming the ‘Central Bank of AI,’ As It Weighs $250 Billion OpenAI Data-Center Backstop

Listen to the audio version of this article (generated by AI).

Powerhouse chipmaker Nvidia (NVDA) is in talks to backstop $250 billion for OpenAI in the construction of the largest artificial intelligence (“AI”) data center ever announced. Nvidia’s financial guarantee would help the ChatGPT company lease a 10-gigawatt data center being developed in southern Ohio by SoftBank (SFTBY), according to the Wall Street Journal.

Nvidia’s backing would help further grease the skids for AI investment. This move, along with a separate $350 billion financing arrangement being discussed with OpenAI to buy chips for the facility, shows how the company is acting like the “central bank of AI,” increasingly supporting its customers so they can keep buying Nvidia’s products.

The first phase of the new data center wouldn’t open until 2028, starting with just 800 megawatts of power, only 8% of the total expected capacity of the center.

Nvidia’s guarantee means that the data-center developer, a subsidiary of SoftBank, would be able to obtain debt financing on more attractive terms. Nvidia is the world’s largest company by market capitalization, so its participation would make a deal more feasible.

In contrast, the new data center’s proposed tenant, OpenAI, recorded a $20.9 billion operating loss in 2025, and it doesn’t have an investment-grade credit rating, hurting its ability to invest.

The magnitude of this project shows that the AI build-out is not just about the supply of chips – it’s a reminder that attractive financing and plentiful access to power can be huge advantages.

Of course, Nvidia benefits from the transaction, too. Nvidia’s support fuels further demand for its graphics processing units that would populate these data centers. Factoring in the Nvidia chips needed for this massive project, the total cost of the AI data center could surpass $500 billion.

Nvidia’s support would include guaranteeing various financing entities to gain lenders’ trust that the project’s financing is secure. The $250 billion guarantee would backstop the lease and the debt needed to build the facility, though not the Nvidia chips to be used inside it.

Nvidia is also working on a $350 billion deal to finance Nvidia chips for OpenAI, per the Journal. Such “circular financing” – in which a company finances the purchase of its own products – could create much greater risk for Nvidia, especially given OpenAI’s poor finances.

Nvidia’s moves show the AI industry is entering an expanded phase of the build-out. To keep its own sales growing, it appears Nvidia must now help finance its own customers. But these moves make Nvidia even more reliant on the money-burning OpenAI to maintain its own sales growth, and they signal that the AI bubble now requires even more creative financing to sustain itself.

OpenAI Continues to Burn Through Cash

Nvidia’s potential backstop of this deal showcases one of the key weak spots at the center of the AI industry: the poor financial shape of AI modelers and the strong possibility that they’re unable to turn a profit. They need regular infusions of capital so that they don’t run out of money.

For example, OpenAI’s 2025 operating loss of $20.9 billion and an estimated $7 billion loss in the first quarter of 2026 show that it’s burning more money even as revenues skyrocket.

Understandably, a prospective landlord or investor in a data center project would be hesitant about a tenant with those kinds of losses and the potential for even higher losses in later years.

Given the economics of AI, it may simply be impossible for OpenAI to produce profits reliably. That means OpenAI would need to continue to raise money from investors or other large companies, such as with its $122 billion funding round in March. The round was funded mostly by large companies: Amazon (AMZN), Nvidia, and SoftBank put up $110 billion of the total.

But OpenAI is so tightly connected with the entire AI industry that its weak points have become the industry’s concerns, too. OpenAI has committed to $1.4 trillion in spending in the coming years, according to Barron’s, and it’s money that other AI players such as CoreWeave (CRWV) are counting on receiving as revenue. If OpenAI can’t pay, it would spiral across the AI industry.

So, Nvidia’s backstop of this deal means it could be taking on more direct exposure to OpenAI, depending on how its guarantees are structured. In any case, even the world’s largest company needs the AI party to continue, even if things look increasingly unsteady.

Given the huge levels of debt being used to finance the AI build-out, even a modest slowdown in growth could quickly burst the AI bubble. Now, low-cost Chinese AI is threatening the industry.

Chinese AI Threatens American AI Strategy

The emergence of top-tier Chinese AI models threatens the debt-fueled “compute-maxxing” strategy that America’s AI strategy is based on. While Chinese models were previously notable for their low cost, their quality has improved significantly, even as they underprice American rivals. Users are increasingly pivoting to low-cost models, to OpenAI’s detriment.

Big corporate users are moving to low-cost models when it makes sense, such as for completing basic tasks. Meanwhile, users will stick with high-cost models for advanced tasks. They may even use high-cost AI to plan tasks, while turning to low-cost models when it comes time to execute those tasks.

The cost difference is enormous in some cases. American AI models may charge five to 10 times more than Chinese models and perhaps up to 150 times more for high-end models.

This price differential has affected usage significantly, it seems. Some 80% of startups using open-source software are now using Chinese AI models, according to Andreessen Horowitz.

The emergence of high-quality, cost-leading Chinese AI models threatens revenue growth at OpenAI and Anthropic. It also threatens a key premise of the American AI “compute-maxxing” strategy of investing heavily in AI hardware, often with significant debt.

So, if OpenAI can’t keep revenue growing at the breakneck pace of the recent past, it will be tougher to meet its spending commitments, including rent on an Nvidia-backed AI data center.

This latest Nvidia deal, if it goes through, brings the company even closer to the likely center of the AI bubble: OpenAI. Nvidia’s creative financing would likely increase its own risk substantially, as it tries to keep the AI bubble afloat by bankrolling the money pit at its very center.

Regards,

James Royal, PhD

Editor’s Note: A highly secure site in West Texas now houses an emerging potential $10 trillion technology backed by Elon Musk and Sam Altman. This breakthrough could completely replace our need for foreign oil – and send one small group of stocks soaring in the process. Click here to learn how you can invest in Elon’s next $10 trillion move.