Nvidia’s total exposure to OpenAI now sits at approximately $600 billion, a figure that transforms the chipmaker from a hardware vendor into the de facto central bank of the artificial intelligence era. This massive financial footprint comprises a reported $250 billion financing guarantee for a 10-gigawatt data center in Piketon, Ohio, alongside $350 billion in proposed chip financing and an existing $30 billion investment. By underwriting the operational existence of a single, unrated, and unprofitable customer, Nvidia has moved beyond a standard vendor-client relationship into a profound, and potentially precarious, integration of balance sheets.
The core mechanism driving this arrangement is credit substitution. OpenAI, which faces projected cumulative losses of $44 billion between 2023 and 2028 and a potential $74 billion operating loss in 2028 alone, lacks the traditional creditworthiness required to secure project financing for a facility of this magnitude. Nvidia is effectively filling this void, replacing OpenAI’s missing financial credibility with its own. This allows the Piketon project—situated at a former uranium enrichment site—to proceed despite the underlying customer’s lack of profitability. It creates a circular flow of capital where Nvidia provides the financial backing to build the very infrastructure designed to consume its own silicon.
This infrastructure project is deeply entangled with state power, shifting from private market dynamics into the realm of industrial policy. The 10-gigawatt capacity at the Piketon site is subject to power allocation controlled by Commerce Secretary Howard Lutnick, who holds the authority to determine which entities gain access to the grid. Furthermore, the energy requirements are being addressed through a $33 billion natural-gas power project funded by Japan under a specific U.S.-Japan trade deal. Under this agreement, Japan recovers its outlay first, after which the U.S. share of proceeds rises to 90 percent. This structure highlights how the physical constraints of AI—specifically electricity—have necessitated a fusion of corporate ambition and geopolitical trade maneuvering.
The scale of the Piketon proposal is unprecedented, yet it mirrors the logic of the $410 million RSI/AWS deal, where 63 percent of the total capital was allocated to compute. Both represent the compute landlord thesis: the entity that controls the silicon and the infrastructure dictates the terms of the digital economy. While the RSI/AWS deal operates at a smaller magnitude, the underlying strategy remains identical. By financing the physical environment—the data center itself—Nvidia secures a captive market for its hardware, effectively locking in demand for years to come.
For OpenAI, this proposed deal marks a critical shift from renting capacity from Microsoft, Amazon, and Oracle to owning its infrastructure. By moving toward a direct lease model, OpenAI attempts to reduce its dependence on the cloud hyperscalers that currently host its models. However, this independence is arguably illusory if the infrastructure is built entirely on the back of Nvidia’s financial guarantees. The developer, SB Energy, a subsidiary of SoftBank, is tasked with the construction, but the financial risk remains concentrated within the Nvidia-OpenAI nexus. The $40 million Community Benefits Agreement and the $4.2 billion transmission build-out with AEP Ohio underscore the massive physical footprint required to sustain this model.
These figures and arrangements remain in the proposal stage and are not finalized. As reported by the Wall Street Journal, these discussions involve people familiar with the matter, and the entire structure could fall apart. The complexity of coordinating a $500 billion data center project, a $33 billion power plant, and the regulatory hurdles of federal power allocation creates significant execution risk. Nvidia, which held $62.6 billion in cash and equivalents at the close of fiscal year 2026, is attempting to leverage its market position to bridge a massive gap in the AI ecosystem, but the volatility of OpenAI’s financial outlook remains a persistent variable.
The Piketon proposal exposes the fragility of the current AI boom. If the industry’s leading model developer cannot sustain its operations without a $600 billion backstop from its primary hardware supplier, the entire economic model rests on a precarious foundation. Nvidia is betting that by becoming the landlord of the physical infrastructure, it can force the market to mature into profitability. Whether this structural gamble succeeds or collapses under the weight of its own capital intensity will determine the viability of the current compute-heavy economic model.