Half a Trillion Dollars on a Former Uranium Site: Inside SoftBank's Massive Ohio AI Bet

On March 20, 2026, Commerce Secretary Howard Lutnick stood on 3,700 acres of land in Piketon, Ohio—a site that once produced weapons-grade uranium during the Cold War—and called it "the largest bet on a construction facility ever made in history" [1]. Beside him, SoftBank CEO Masayoshi Son announced that the campus being built on this former Department of Energy site would ultimately channel $500 billion into a single AI data center complex [2].

The scale defies easy comparison. If fully realized, the project would exceed the inflation-adjusted cost of the Interstate Highway System. It would consume more electricity than many mid-sized countries. And it would be built in Pike County, an Appalachian community of 27,000 people where nearly one in five residents lives below the poverty line [3].

But $500 billion is a number Son has used before—and whether this figure represents a firm financial commitment or a headline-grabbing aspiration depends on whom you ask.

The Deal: What Is Actually Committed

The project, branded as the "PORTS Technology Campus," will be built in phases on the site of the former Portsmouth Gaseous Diffusion Plant, which enriched uranium from 1954 until its closure in 2001 [1]. The U.S. Department of Energy retains ownership of the land and is partnering with SoftBank through SB Energy, the conglomerate's energy subsidiary.

The first phase is the most concrete: an 800-megawatt data center costing $30 billion to $40 billion, with construction starting immediately and operations expected by early 2028 [2]. SB Energy says it will invest $10 billion in this initial phase, creating 4,000 construction jobs and 300 to 400 permanent operational positions [4].

The full buildout envisions a 10-gigawatt data center campus—more than half of all current U.S. data center operating capacity—powered by 9.2 gigawatts of new natural gas generation costing $33 billion [5]. SoftBank is also investing $4.2 billion in new transmission lines across southern Ohio in partnership with American Electric Power (AEP) Ohio [4]. At peak construction, the project would employ 35,000 workers; once complete, approximately 2,500 permanent staff [1].

The gap between the $10 billion first-phase commitment and the $500 billion headline is vast. SoftBank has not disclosed binding financial commitments beyond the initial phase, and subsequent phases are conditional on demand for AI computing, power availability, and presumably, continued favorable government treatment.

SoftBank's Track Record: Pledges vs. Reality

Son is no stranger to making large investment announcements alongside U.S. presidents. In December 2016, he appeared at Trump Tower pledging $50 billion and 50,000 American jobs [6]. In December 2024, he was back with a $100 billion pledge and promises of 100,000 jobs [7].

The 2016 commitment provides a useful case study. SoftBank channeled money primarily through its Vision Fund, which eventually invested roughly $75 billion—exceeding the headline number [8]. But whether those investments created 50,000 U.S. jobs is unverifiable. SoftBank has declined to provide an accounting, and because the Vision Fund primarily invested in private companies, public data is unavailable [8]. Nearly half of the fund's capital went into WeWork, which filed for bankruptcy [8]. Many other Vision Fund portfolio companies shed staff during downturns.

As Axios reported when the 2024 pledge was announced, SoftBank was "revisiting an old playbook" [8]. The pattern is consistent: a large number announced alongside a political leader, followed by investments that are difficult to attribute specifically to the pledge.

The $500 billion Ohio figure exists in a similar category. It represents what Son says the campus could channel over its lifetime—not a binding capital commitment backed by identified funding sources. For context, SoftBank's Vision Funds have approximately $166 billion in total assets under management [9]. The company's entire market capitalization fluctuates around $100-200 billion. A $500 billion commitment would require capital mobilization far beyond anything SoftBank has achieved.

Source: FRED / Bureau of Economic Analysis

Data as of Mar 21, 2026CSV

The Power Problem: 10 Gigawatts of Gas

The most technically concrete—and environmentally consequential—element of the project is the power infrastructure. A 10-gigawatt data center requires roughly the output of nine nuclear reactors [5]. SoftBank's plan calls for 9.2 gigawatts of new natural gas combined-cycle generation, with turbines distributed across the southern Ohio region rather than concentrated at a single plant [2].

The turbines have already been sourced, with the first delivery expected within a year and full deployment by decade's end [2]. This makes the power buildout the most credible near-term component of the project.

But it also makes this one of the largest new sources of fossil fuel emissions in the country. Energy analysts estimate the gas plants could emit approximately 16 to 19 million metric tons of CO2 annually, depending on capacity factor assumptions [10]. Ohio's total energy-related CO2 emissions are roughly 200 million metric tons per year, meaning this single project could add 8-10% to the state's carbon footprint.

Son addressed energy costs at the groundbreaking, pledging directly: "I will commit right here, right now, that we will protect the electricity bill" [1]—an implicit acknowledgment that residents fear the data center's enormous power appetite could raise rates for existing customers.

SoftBank has made no public commitments to renewable energy for this facility. Environmental organizations have argued that $33 billion could fund renewable generation and battery storage sufficient to meet AI computing needs without the carbon footprint [10]. The Trump administration, however, has framed fossil fuel power for AI as a feature, not a bug, of its energy policy.

Ohio's Incentive Gamble

Ohio has become one of the most aggressive states in courting data centers, and the tax incentives on offer are substantial. The state provides sales tax exemptions of up to 100% for qualifying data center equipment and construction materials [11]. Local governments routinely offer property tax abatements lasting 15 to 30 years [11]. State job creation tax credits provide additional benefits.

The track record of these incentives is contested. Data centers in Ohio received $2.5 billion in state and local tax incentives between 2017 and 2024 [12]. A report from Policy Matters Ohio found that 13 approved data center tax agreements covered nearly $5.1 billion in investment but generated just 356 jobs—a cost-per-job ratio that critics call indefensible [12]. The data center sales tax exemption alone costs Ohio more than $140 million per year in forgone revenue [13].

For the SoftBank project specifically, Pike County commissioners have begun negotiating a payment in lieu of taxes (PILOT) agreement with SB Energy [1]. Because the federal government retains ownership of the land, the standard property tax framework does not apply, creating additional complexity. SoftBank has pledged $40 million to local schools and medical infrastructure—a gesture, but a small fraction of the project's stated value [1].

Supporters argue that the economic multiplier effects—housing demand, supply chain spending, and secondary employment—justify the incentive costs. State Senator Shane Wilkin (R-Hillsboro) described the project as transformative for a region that has hemorrhaged jobs since the uranium plant closed [4].

Source: FRED / Bureau of Labor Statistics

Data as of Mar 21, 2026CSV

Pike County: Appalachian Hope and Skepticism

The choice of Pike County adds a layer of complexity to the project's narrative. The county's 27,080 residents have a median household income of $49,552—roughly 70% of the national median [3]. The largest employers are in healthcare, manufacturing, and retail [3]. The county sits in the Appalachian region, where decades of economic extraction followed by abandonment have bred both desperation for investment and wariness of outside promises.

DOE Assistant Secretary Tim Walsh captured the local sentiment: "Over the last 20 years, they've lost so many jobs, and now they see this as a way to really grow the future for themselves and their families" [4].

The workforce math is challenging. The 300-400 operational jobs in Phase 1 and 2,500 at full buildout are significant for a county this size, but data center operations require specialized skills—network engineering, cooling systems management, cybersecurity—that the local workforce largely does not possess. Construction jobs, while more numerous, are temporary and often filled by workers who travel to sites rather than relocate permanently. What percentage of the 35,000 peak construction workers would be Pike County residents versus imported labor is an open question that SoftBank has not addressed.

Why Ohio? Geography, Politics, and Trade

The site selection for the Stargate project—the broader $500 billion AI infrastructure initiative of which this campus is a part—drew more than 300 proposals from 30 states [14]. Ohio secured two Stargate facilities: this Pike County campus and a separate site in Lordstown, where SoftBank has already broken ground on a data center that is expected to be operational in 2026 [14].

Ohio offers several advantages. The state sits at the intersection of major fiber-optic networks running between the East Coast and Midwest. Electricity costs are below the national average. The former DOE site provides a massive land footprint with existing high-voltage transmission infrastructure—a legacy of the decades when the plant consumed roughly 2,000 megawatts to enrich uranium [1].

But the selection cannot be separated from geopolitics. The project is part of a $550 billion U.S.-Japan trade agreement, itself a product of tariff negotiations between the Trump administration and Tokyo [1][4]. Commerce Secretary Lutnick and Energy Secretary Chris Wright both attended the groundbreaking, framing the project as a pillar of American AI competitiveness against China [15].

The Stargate Project, formally announced by President Trump on January 21, 2025, positions massive domestic AI infrastructure as a national security imperative [15]. The stated logic: if AI models will determine economic and military advantage in the coming decades, the physical infrastructure to train and run those models must be located on American soil, powered by American energy, and controlled by American-aligned companies.

The China Question and the Case for Scale

The national security framing provides the strongest argument for public support of private AI infrastructure at this scale. China has invested heavily in state-backed AI infrastructure, and the race to build the physical capacity for frontier AI models is intensifying [16]. The Stargate network—spanning sites in Texas, New Mexico, and Ohio with nearly 7 gigawatts of planned capacity and over $400 billion in projected investment—represents the most ambitious private-sector response [14].

Critics of this framing, however, note that the security argument can justify almost any subsidy level. DeepSeek, the Chinese AI lab, demonstrated in early 2025 that competitive AI models can be built with significantly fewer resources than assumed, disrupting the premise that raw compute scale is the only path to AI leadership [16]. If efficiency gains continue, the demand projections underlying a 10-gigawatt campus could prove overstated.

Alternative models were available. Direct federal investment in AI computing infrastructure—a public option for AI, in effect—was proposed by some policy researchers but gained no traction in the current administration [16]. Other private-sector partnerships, including with domestic companies like Microsoft or Google, could have achieved similar scale with potentially different terms. The administration chose SoftBank, an entity with a complicated financial history and a CEO whose relationship with the sitting president dates back a decade.

The Scale in Context

To understand what $500 billion means: the entire U.S. data center market generates roughly $115-135 billion in annual revenue, depending on which research firm you ask [17]. Alphabet, Amazon, Microsoft, and Meta combined plan to invest about $350-400 billion in data centers globally across 2025 and 2026 [17]. SoftBank's headline figure for a single campus exceeds what the four largest tech companies on Earth plan to spend everywhere in two years.

The largest infrastructure projects in American history offer some scale comparison. The Interstate Highway System cost roughly $600 billion in inflation-adjusted dollars—spread over 35 years and 48,000 miles [18]. The Apollo program cost approximately $260 billion adjusted. Even if the $500 billion figure is aspirational, the first phase alone—$30-40 billion—would rank among the largest single-site construction projects in U.S. history.

Whether the AI computing market will support this level of investment remains the central uncertainty. If demand for AI inference and training continues its exponential growth trajectory, the capacity could be absorbed. If AI efficiency improvements reduce compute requirements faster than demand grows—a scenario that DeepSeek's success makes plausible—enormous overcapacity could result.

What Comes Next

The first turbines should arrive at the site within 12 months. The initial 800-megawatt data center is projected to go operational in early 2028. Beyond that, each subsequent phase depends on market conditions, SoftBank's ability to raise capital, and the continued political will to support the project through multiple election cycles.

Pike County residents will see construction crews first—thousands of them. Whether they will see the long-term economic transformation that federal and state officials have promised depends on factors largely outside their control: the trajectory of AI computing demand, the durability of political agreements between Washington and Tokyo, and whether Masayoshi Son's latest mega-pledge follows a different pattern than his previous ones.

The site that once enriched uranium for nuclear warheads is being repositioned as the foundation of America's AI future. The ambition is unmistakable. The execution remains to be seen.