SoftBank planning massive $500 billion data center in Ohio

SoftBank plans an AI-focused data center campus in Ohio that CEO Masayoshi Son said could channel $500 billion into a single campus at a former U.S. Energy Department uranium enrichment site. The complex is designed to draw up to 10 GW of power; the first phase is expected to provide ~800 MW, cost $30–40 billion and be completed in early 2028. U.S. Commerce Secretary Howard Lutnick and Energy Secretary Chris Wright unveiled the project alongside Son.

Analysis

This project will shift winners away from pure-play co-location landlords toward firms that build and harden the underlying grid and on‑site power infrastructure. Expect outsized profits for suppliers of high-voltage switchgear, transformers, custom gensets and large-scale cooling systems because interconnection queues and long lead times mean procurement orders will be placed well ahead of campus occupancy, locking in multi-year revenue streams for manufacturers and EPCs.

A second-order effect is regional power economics: sustained, concentrated demand materially alters capacity and ancillary markets in the host grid, improving margins for merchant generators and storage providers while creating political pressure for rate-base recovery by local regulated utilities. That transfer from retail/wholesale power to infrastructure owners is durable once transmission upgrades are permitted and financed, but it’s concentrated geographically and therefore tradable.

Key risks are timing, regulatory friction and financing cadence — each can turn an immediate industrial win into a multi-year slugfest. If interconnection approvals, environmental reviews or local labor bottlenecks push schedules out, equipment prices and labor escalation could erode contractor margins and force scope changes; conversely, clear approvals would compress project risk and catalyze a multi-quarter sourcing spike.

The consensus playbook is to buy AI chip names; that’s sensible but crowded and largely priced. A higher expected return is available in mid-cap industrials and regulated utilities that will monetize incremental load through long-term contracts or rate-base treatments, and in engineering firms that win the ‘first mover’ contracts to design the campus and grid tie‑ins.

Key Decisions for Investors

• Long AEP (AEP) — 12–24 month horizon. Size 1–2% position to capture regulated rate-base growth and contract approvals; upside 15–25% if interconnection/rate filings proceed, downside ~10% if permitting stalls.
• Long Jacobs Engineering (J) or Fluor (FLR) — 18 month horizon. Target 2–3% position in the winner(s) of EPC scope; risk/reward roughly 3:1 as early contract awards re-rate margins, but bid‑win risk and working‑capital exposure are real.
• Long Eaton (ETN) via 9–12 month call spread (buy calls, sell higher strike) to play switchgear/PDUs and power-management demand. Limited premium outlay mitigates execution risk from schedule slips; payoff if equipment orders accelerate.
• Pair trade: long NRG Energy (NRG) or Vistra (VST) / short Digital Realty (DLR) — 12–24 months. Capture higher merchant power and capacity prices vs. pressure on third‑party leasing in the local market; thesis pays if the campus self‑hosts significant compute, fails if REITs become preferred partners.
• Convex/option hedge: small long-dated NVDA (NVDA) call position as optionality on persistent GPU demand — keep allocation tiny versus industrial exposure because semiconductor upside is consensus‑priced and execution risk is regulatory and supply‑chain driven.