The Announcement and What It Actually Says
Nine-point-two gigawatts of natural gas. That figure, buried in the technical details of the DOE’s official fact sheet, is more revealing than any number Masayoshi Son said at the podium.
On March 20, 2026, Son and U.S. Secretary of Commerce Howard Lutnick attended a groundbreaking ceremony at the former Portsmouth Gaseous Diffusion Plant, now designated PORTS Technology Campus – in Pike County, Ohio. The Department of Energy announced a public-private partnership to develop a 10-gigawatt AI data center campus on the site. Multiple independent outlets, including an AP wire report, confirmed the announcement. The DOE’s fact sheet is unambiguous: this is a federal agency co-developing AI infrastructure on federal land, not issuing a grant or signing a subsidy agreement.
Son told reporters at the announcement: “The investment starts at $500 billion and could eventually reach $1 trillion.” Per multiple reports of his remarks, he also said: “Instead of many locations, many years, now we are delivering $500 billion in one campus.” Those are aspirational statements from a CEO at a groundbreaking ceremony. They describe his vision for the campus at full buildout, not committed capital.
The committed capital, per the DOE primary source, is $4.2 billion: SB Energy and AEP Ohio have partnered to build that amount in new electrical transmission infrastructure and grid upgrades in Ohio. That’s the figure grounded in a signed partnership with a federal agency. The rest of the financial picture, a first phase reportedly costing $30–40 billion, reportedly $33.3 billion in Japanese investment for the gas generation component, the $500 billion full-campus aspiration, comes from T3 cross-reference reporting and should be read as directional rather than contractual.
What Each Party Brings to the Deal
Understanding the deal structure requires mapping the four principal parties and their distinct contributions.
U.S. Department of Energy. The DOE contributes the asset that makes the scale conceivable: federal land. The PORTS Technology Campus occupies a former uranium enrichment site that the federal government has owned since the plant’s closure in 2001. By designating this land for AI compute development, the DOE is doing something without direct precedent, converting legacy nuclear industrial infrastructure into AI infrastructure. The DOE also brings permitting authority and, by association, the political legitimacy that accelerates the public-private deal structure.
SoftBank Group. The lead private investor. SoftBank’s involvement connects this project to Son’s broader Stargate initiative, a publicly announced $500 billion commitment to AI infrastructure investment in the United States. The Ohio campus, in Son’s framing, is a concentration of that ambition rather than a distribution of it.
SB Energy. SoftBank’s energy subsidiary is responsible for the power generation infrastructure: approximately 9.2 gigawatts of natural gas generation, deployed via a fleet of gas turbines. The Scioto Post confirmed this figure from local coverage of the announcement. SB Energy is, in effect, building a dedicated power plant, or a collection of them, to run a single AI campus.
AEP Ohio. The grid transmission partner. AEP Ohio’s role is the $4.2 billion grid upgrade: building the transmission infrastructure to connect the campus’s enormous power generation capacity to the existing grid and to deliver power reliably at scale. This is the confirmed financial commitment from the DOE fact sheet, and it’s the piece of the project that touches Ohio’s existing energy infrastructure most directly.
U.S. Department of Commerce. Secretary Lutnick’s presence at the groundbreaking signals a trade-deal dimension: the project is framed partly in the context of Japanese investment in the United States, with reportedly $33.3 billion in Japanese capital targeting the natural gas generation component. This is reported but not T1-confirmed.
The Energy Math, and What It Signals
Let’s be precise about scale. Nine-point-two gigawatts of dedicated natural gas generation for a single AI campus is not just a large number. It’s a statement about the energy trajectory of large-scale AI compute.
For context: the total electricity generation capacity of Ohio as a whole is approximately 30 gigawatts, serving the entire state’s residential, commercial, and industrial demand. A single AI campus requiring 9.2 GW of dedicated generation, roughly 30% of Ohio’s entire existing capacity – tells you something about the energy appetite of frontier AI infrastructure at scale.
This matters for three audiences. For infrastructure investors, it confirms that AI compute buildout is inseparable from energy infrastructure investment, the data center and the power plant are the same project. For energy policy professionals, it raises direct questions about grid impact, carbon commitments, and whether gas-powered AI infrastructure creates regulatory exposure for operators. For technology practitioners, it’s a signal about where compute costs are heading: at this scale, energy is not a variable cost. It’s a capital investment of its own.
The natural gas commitment also arrives at a specific policy moment. AI’s electricity consumption has moved from a footnote in technology coverage to an active regulatory and legislative topic. Committing 9.2 GW of new gas generation to a single campus, with federal government co-signature – is a policy statement, whether or not it’s framed as one.
The Federal Precedent
This is the most structurally significant aspect of the announcement, and it received the least attention in initial coverage.
The DOE is not issuing a grant. It’s not providing a tax incentive. It’s not standing at a podium to announce a loan guarantee. The DOE is contributing federal land and institutional partnership to a private AI infrastructure buildout, becoming a structural party to the project. The site’s history – uranium enrichment for the nuclear weapons program, then commercial reactors, then decades of federal ownership after closure, makes it a genuine artifact of American industrial policy. It was always federal. What’s new is what it’s becoming.
The precedent this sets is worth naming. If the DOE model for PORTS Technology Campus proves viable, the federal government has at least a dozen other legacy industrial sites, former weapons plants, decommissioned government facilities, surplus federal land adjacent to power infrastructure, that could be designated for AI compute development on similar terms. That’s not a prediction. It’s the question that the Ohio announcement puts on the table for AI infrastructure policy discussions.
For compliance and regulatory professionals, the governance structure also matters. A project where the federal government is a structural co-developer, not just a passive recipient of permit applications, occupies different regulatory territory than a purely private data center buildout. Environmental review, energy permitting, and grid interconnection processes may move differently when a federal agency is a named party. Whether that accelerates or complicates the permitting timeline for the natural gas infrastructure is an open question.
What Comes Next, and What to Watch
The first phase of the data center is expected to include approximately 800 megawatts of capacity, at a cost of $30–40 billion, with completion targeted for early 2028, according to multiple reports. That’s the near-term milestone.
The longer arc runs to the full 10-gigawatt campus, which Son frames as a multi-year buildout potentially reaching $500 billion in total investment. The credibility of that figure depends entirely on execution: whether the first phase delivers on its 2028 target, whether Japanese capital flows at the reported scale, and whether the regulatory and permitting environment for 9.2 GW of new gas generation stays clear.
Three specific signals are worth tracking. First: DOE permitting for the natural gas infrastructure. This is the component most likely to encounter regulatory friction, particularly if energy regulators scrutinize the environmental implications of dedicated gas generation at this scale. Second: whether AEP Ohio’s $4.2 billion grid upgrade proceeds on a timeline compatible with first-phase completion. The data center and the grid upgrade are interdependent. Third: whether federal agencies beyond DOE begin designating legacy industrial sites for AI compute development, which would confirm that the Ohio model is replicable policy rather than a one-time deal.
TJS Synthesis
The headline number is $500 billion. The real story is $4.2 billion and 9.2 gigawatts.
The committed investment, the figure that has a federal agency’s signature on it, is the grid infrastructure deal. That’s the indicator that this project has moved from announcement to construction reality. The aspirational figure tells you where Son wants to take it. The confirmed figure tells you where it actually is.
What’s structurally new here isn’t the scale of the ambition. It’s the federal co-developer model. The DOE didn’t write a check. It brought land, permitting authority, and institutional legitimacy to a private capital stack. That combination, federal assets plus private capital plus a legacy industrial site, is a template that didn’t exist for AI infrastructure development before March 20, 2026. Whether it becomes a repeatable model or remains a singular arrangement is the question that will define its significance. The Ohio campus is either a one-time deal or the prototype for how the U.S. builds AI infrastructure at scale. The answer won’t be visible in Son’s remarks. It’ll be visible in the permit filings.