OpenAI has signed a record-breaking data center lease in Ohio, with Nvidia providing financial backing worth up to $105 billion, according to The Decoder. The figure alone puts this deal in a different weight class than most corporate real estate transactions — it is closer to the GDP of a mid-sized country than a typical lease agreement.

The deal underscores a pattern that has defined OpenAI's 2025-2026 growth: it does not build compute capacity so much as it assembles it from a coalition of partners, each with its own reason to keep the lights on. Nvidia's involvement here is notable not because the chipmaker is unfamiliar with financing arrangements — it has backed several AI infrastructure deals over the past two years — but because of the scale at which it is now willing to underwrite a single customer's real estate footprint.

For a company still operating at a significant cash burn relative to revenue, tying a lease this large to a single chip supplier's balance sheet is a structural choice, not a footnote.

This lease does not stand alone. It adds to a run of multi-hundred-billion-dollar infrastructure commitments OpenAI has made over the past year, including its Stargate joint venture with Oracle and SoftBank, and separate compute deals with AMD and Broadcom. The Ohio agreement is notable less for being unprecedented in kind and more for how concentrated the financial risk now is — a single chip vendor underwriting a single site at nine figures.

Why Nvidia is underwriting real estate, not just chips

Nvidia's core business is selling GPUs, not guaranteeing data center leases. Its willingness to back a deal of this size signals how directly the company's own revenue now depends on OpenAI's ability to keep expanding. Every additional campus OpenAI commits to is, in effect, forward demand for Nvidia hardware — the more compute OpenAI leases, the more GPUs it will eventually need to fill that space.

This is the mechanism critics have started calling circular financing: a chip supplier helps fund the infrastructure that its own products will occupy, which in turn generates the revenue that justifies the chip supplier's valuation. It is not unique to this deal — Nvidia has taken similar positions in other AI infrastructure buildouts — but a $105 billion backstop on one lease is large enough that it stops being a side arrangement and becomes a load-bearing part of OpenAI's expansion strategy.

What it means for the compute market

For teams building on top of OpenAI's models, this deal doesn't change API pricing today, but it does say something about the runway OpenAI is building for itself:

The practical read for developers: this is not a reason to expect near-term price cuts on GPT API access, but it is a reason to expect OpenAI's capacity ceiling to keep rising, which matters if your product roadmap depends on higher-volume or higher-context usage becoming affordable.

AiiN's takeaway

The headline number — $105 billion — is designed to be eye-catching, and it is. But the more useful signal is what the financing structure reveals: OpenAI's growth is now underwritten as much by its hardware suppliers as by its own revenue or outside investors. That is a bet on continued demand growth that, in our estimation, only makes sense if OpenAI's usage curves keep climbing at close to their current pace. If they don't, a $105 billion backstop tied to a single Ohio campus becomes a very expensive test of how patient Nvidia is willing to be with its most important customer.