Nvidia invests $1.5 billion in SoftBank's data centre unit for OpenAI project in Ohio

Nvidia announced on 17 August 2026 that it is investing $1.5 billion in SB Energy, the SoftBank Group subsidiary that develops data centres and energy infrastructure. The investment is tied to a large-scale data centre project in Pike County, Ohio, with OpenAI as the sole tenant. Through the deal, Nvidia secures its chips as the source of computing capacity for that project.

This is one of the largest individual investments Nvidia has made in an external party to date. The Ohio project forms part of the broader Stargate roll-out across the American Midwest, a joint initiative by OpenAI, SoftBank and others to build large-scale AI infrastructure. The total financing architecture surrounding the Ohio initiative is estimated at around $500 billion, a figure that encompasses guarantees for lease obligations and hardware purchases.

Former uranium plant as AI campus

The project site is the PORTS-Pike Technology Campus near Piketon, Ohio, a former uranium enrichment facility that was operated by the US Department of Energy. SB Energy is developing the site to become one of the largest AI data centre campuses in the world. OpenAI has signed a twenty-year lease agreement for the campus.

The initial phase provides for 4.25 IT-gigawatts of infrastructure, with an option to expand to a total of 8 IT-gigawatts. Capacity will come online in phases from 2028. In addition to Ohio, SB Energy is also developing sites in Georgia (3.2 gigawatts) and Lordstown, Ohio, while OpenAI already has a 1.2-gigawatt data centre under construction in Milam County, Texas, as the first phase of its next-generation AI infrastructure.

Billions for power and grid

A data centre of this scale requires enormous amounts of energy. SB Energy and SoftBank plan to build at least 10 gigawatts of new power generation and are investing $4.2 billion in Ohio's grid infrastructure. They are doing so in collaboration with AEP Ohio, which is carrying out regional network upgrades.

At the PORTS-Pike site itself, a natural gas power plant with a capacity of 9.2 gigawatts will be built, with estimated construction costs of $33 billion. Japan has also committed $33.3 billion in natural gas financing to enable that same 9.2 gigawatts of generation capacity. The project further includes an $80 million community fund for the Piketon region.

In January 2026, OpenAI and SoftBank had already jointly invested $1 billion in SB Energy, split equally, earmarked specifically for data centre infrastructure. Nvidia's $1.5 billion investment comes on top of that and underscores the importance that the chipmaker and its customers attach to securing capacity at an early stage.

Nvidia's credit role extends beyond the investment

In addition to the direct $1.5 billion investment, Nvidia has committed to providing credit support for land, power and the construction of the building envelope at the PORTS-Pike campus. According to available information, this involves a maximum credit guarantee of $105 billion for project development. Earlier reports cited a guarantee of $250 billion, which was subsequently revised down to less than $120 billion; the final architecture stands at a maximum of $105 billion.

The total financing envelope of around $500 billion consists of multiple layers. These include, alongside the credit guarantees, financing for OpenAI's lease obligations and amounts for the purchase of GPUs and related hardware. The precise breakdown and ultimate size will only become clear as the project progresses and capacity comes online.

For Nvidia, the structure is strategic: by entering early as both investor and guarantor, the company anchors its chips as the default computing resource for one of the largest AI infrastructure projects currently being built in the United States.

What this means for the European AI scene

The scale of the Ohio project highlights the gap between AI infrastructure investment in the United States and what is currently available in Europe. While individual European countries and EU programmes are working on plans for sovereign AI infrastructure, the Ohio project involves a single campus with potential capacity of 8 IT-gigawatts, backed by guarantees and investments running into the hundreds of billions of dollars.

For Dutch and European investors, founders and policymakers, this raises the question of what position European parties will occupy in the global distribution of AI computing capacity. The campuses now being built in the US will be operational within two to three years and will lock in demand for chips, energy contracts and AI services for years to come. Those in Europe who fail to make comparable commitments will later be buying into a market already shaped by others.