OpenAI Secures 20-Year Ohio Data Center Lease Backed by NVIDIA's $105 Billion Guarantee
Key Takeaways
- •OpenAI signed a 20-year lease for the PORTS Technology Campus in Ohio, which is expected to begin operations in 2028 and will be built and operated by SB Energy.
- •NVIDIA disclosed in an SEC filing that its financial commitment for the arrangement is capped at $105 billion, and it may be required to pay deficit amounts if OpenAI fails to meet lease conditions.
- •NVIDIA secured an initial 4.25 IT-GW of capacity at the campus and holds an option for the remaining 3.75 IT-GW, which would raise its total to eight IT-GW.
- •Jensen Huang said OpenAI remains responsible for paying the lease and that unused computing power could be resold to cloud providers, businesses, AI labs, and startups.
- •The campus will run on NVIDIA's full-stack DSX AI factory platform, which includes GPUs, CPUs, and networking technology.
OpenAI has signed a 20-year lease for a data center in Ohio that will begin operations in 2028, with NVIDIA backing the project through financial commitments capped at $105 billion. SB Energy will build and operate the planned campus.
The agreement covers the PORTS Technology Campus in Ohio, which will serve as a large artificial intelligence data center. OpenAI will use the computing capacity available at the facility. According to NVIDIA, the project will feature its full-stack DSX AI factory platform, a system that includes GPUs, CPUs, and networking technology. These components will support the facility's AI computing infrastructure.
The scale and long duration of the lease reflect how major AI developers are securing access to computing power well before sites come online, as data center capacity has become a central input for training and running large models. For OpenAI, a campus of this size adds another long-term source of infrastructure without changing the fact that the facility is not expected to start operating until 2028.
Why NVIDIA Is Backing OpenAI's Ohio Campus With $105B
Under the arrangement, the chipmaker secured an initial 4.25 IT-GW of capacity and holds the option to take the remaining 3.75 IT-GW at the campus. Exercising that option would raise its total capacity to eight IT-GW.
NVIDIA disclosed the financial commitment in a filing with the US Securities and Exchange Commission. The total payment amount for the initial agreement is limited to $105 billion, and the document details the mechanism behind the lease guarantee.
Under that mechanism, NVIDIA may have to pay any deficit amount if OpenAI fails to meet the lease conditions. The payment would be based on the guaranteed minimum amount after recoveries, which could come from replacement leases or asset sales.
The arrangement comes amid rising demand for artificial intelligence infrastructure. NVIDIA pointed out that the use of artificial intelligence is moving from a technology to an infrastructure, underscoring why long-term capacity deals are becoming part of how the sector plans for future compute needs.
How NVIDIA Can Resell OpenAI's Unused Capacity
In a post on X, NVIDIA CEO Jensen Huang addressed concerns that the agreement involved circular financing. He said OpenAI would remain responsible for paying the lease, while NVIDIA's role would be to support the project through its financial guarantee.
Huang said the computing power could be sold off if the tenant failed to move forward. He also noted that NVIDIA compute is common and fits the needs of different customers, with possible purchasers including cloud service providers, businesses, AI labs, and startups.
— Jensen Huang (@JensenHuang) August 17, 2026
The deal comes at a time when speculation about an OpenAI initial public offering remains rife. Polymarket showed a 20% chance of the firm making a listing in 2026, while the AI developer has yet to reveal a timetable for its IPO.
Huang referred to the PORTS-Pike deal as the next phase of NVIDIA's development. He pointed out that the firm began with chipsets before branching into other technological capabilities such as systems, networking, CUDA, and full-fledged AI factories.
What NVIDIA Will Provide at OpenAI's Ohio Campus
The CEO described NVIDIA as an all-stack artificial intelligence infrastructure platform, with technologies spanning hardware, software, connectivity, and computing. The OpenAI project will bring all of these technologies together in one vast campus in Ohio.
NVIDIA has released its Nemotron 3.5 Lightning AI model, which is open-source, and Huang has expressed support for open artificial intelligence models. The release adds to the company's broader focus on AI hardware and software.