Masayoshi Son Wants to Bury $500 Billion in an Ohio Field — What It Really Means to Build History's Largest AI Campus on a Uranium Enrichment Site

Tracking this project's trajectory across different time horizons reveals dramatically different scenarios.

Over the next one to six months, concrete groundbreaking preparations will proceed. Practical challenges including environmental impact assessments, construction permits, and labor force recruitment are piled high. Since phase one (800 MW) targets completion in early 2028, serious construction needs to begin in the second half of 2026. During this period, opposition from environmental groups and local residents could escalate into legal action, and how that affects the timeline becomes the first key flashpoint to watch. Natural gas price trends are equally important. With the Strait of Hormuz crisis roiling global energy markets, the economics of a natural gas-powered plant may shift from what was assumed during planning.

Between six months and two years out, structural shifts in the AI industry will test this project's viability. JLL's forecast suggests 2027 could mark the inflection point when AI inference workloads overtake training. Since inference is better suited to distributed processing than training, the strategic value of a single mega-campus could face reassessment. On the other hand, new AI paradigms like multimodal AI, Physical AI, and world models will continue demanding massive computing power, and this demand will justify Piketon's existence. During this window, how closely phase one construction progress and actual costs track the original plan becomes the critical indicator for the entire project's credibility.

The long-term outlook spanning two to five years splits into three scenarios. In the bull case, progress toward AGI accelerates beyond expectations, computing demand exceeds projections, and Piketon approaches its full 10 GW buildout. Quantum computing and fusion energy research hubs form around it, positioning it as the heart of American technology infrastructure. In this scenario, Pike County becomes the protagonist of a 21st-century Silicon Valley origin story.

In the base case, phases one and two proceed as planned, but the pace of expansion slows from phase three onward as AI model efficiency improvements and rising energy costs create headwinds. The final scale reaches 40 to 60 percent of the original target, but it still functions as one of the world's largest AI infrastructure sites.

In the bear case, underperformance in AI monetization relative to expectations, surging energy prices, tightening environmental regulations, or shifts in US-Japan political relations combine to effectively freeze the project after phase one completion. Much like SoftBank's Vision Fund experience, a grand promise quietly scales down.

My assessment is that the base case is most realistic. The structural growth of AI demand is certain, but there are simply too many variables for the $500 billion figure to materialize in full. Yet even at half that amount, Piketon becomes an impossible-to-ignore anchor on the global AI infrastructure map. The real value of this project is not the $500 billion number itself — it is that it establishes a benchmark for the scale of bets that nations and corporations need to make in the era of AI infrastructure competition. The EU, China, India, and Saudi Arabia will use Piketon's number as their reference point for future AI infrastructure investments, and that alone makes this announcement a moment worthy of recording in AI industry history.