SoftBank's $40B AI Data Center: What to Expect

Ten gigawatts. Let that number sink in for a moment. The entire state of Ohio currently consumes roughly 14 GW of electricity at peak demand — and SoftBank is reportedly planning a single data center campus that could approach 10 GW on its own. That's not just a data center. That's a new load center on the American grid.

SoftBank Group's plans to develop a massive AI data center in Ohio — carrying a price tag somewhere between $30 billion and $40 billion — represent one of the most consequential infrastructure bets of the decade. Whether you're an energy developer, a transmission planner, a land investor, or a municipal official in the Midwest, this project has implications you need to understand now, before the ground breaks.

The Project at a Glance

SoftBank has been on an aggressive AI infrastructure push since Masayoshi Son made his widely publicized pledge to invest $100 billion in the United States. This Ohio data center is shaping up to be the flagship execution of that commitment.

The numbers are staggering by any measure. A $30–40 billion capital commitment rivals the GDP of several small nations. At full buildout, 10 GW of power demand would make this facility — or campus, really — one of the single largest electricity consumers in the Western Hemisphere. For context, a typical hyperscale data center today might draw 100–500 MW. SoftBank's Ohio project would be 20 to 100 times that scale.

Ohio wasn't chosen arbitrarily. The state sits inside the PJM Interconnection, the largest wholesale electricity market in North America, covering 13 states and serving 65 million people. Access to diverse generation sources, existing transmission infrastructure, and relatively competitive land costs make it a logical anchor point for a project of this magnitude. The region also benefits from proximity to Great Lakes cooling water resources — a genuine operational advantage as AI compute demands push thermal management to its limits.

What 10 GW Actually Does to the Grid

Here's where the story gets complicated, and where most coverage misses the plot.

Ten gigawatts doesn't just show up. It has to be built — generation, transmission, and distribution simultaneously. PJM's interconnection queue is already legendarily backlogged, with hundreds of gigawatts of proposed projects waiting years for approval and grid studies. Adding a 10 GW load of this nature isn't a routine interconnection request; it's a fundamental reshaping of regional power flows.

The transmission buildout alone could dwarf the data center construction cost in terms of complexity and timeline. New high-voltage transmission lines require right-of-way acquisition, multi-jurisdictional permitting, and, in many cases, state regulatory approval processes that can stretch five to ten years. SoftBank will almost certainly need to negotiate bespoke arrangements with both PJM and Ohio's utility providers — potentially including direct power purchase agreements, behind-the-meter generation, or co-located generation assets.

Expect nuclear to be part of the conversation. Ohio is home to the Davis-Besse and Perry nuclear plants, both operated by Energy Harbor (now part of Vistra). AI data centers increasingly covet nuclear's 24/7 carbon-free generation profile, and deals like the Microsoft-Constellation agreement at Three Mile Island have established a template. A project at this scale likely can't be powered by intermittent renewables alone — baseload will be essential.

The local infrastructure implications extend beyond electrons. Water, roads, fiber, substations, and workforce housing all get stressed when a project of this magnitude lands in a region. Ohio communities near the eventual site should be modeling those second-order impacts now.

The Investment Math — and Who Stands to Benefit

A $30–40 billion data center investment doesn't stay contained within a fence line. It radiates outward into land, energy, construction, and supply chain markets across a multi-state region.

For land investors and developers, the opportunity is already in motion. Data center campuses at this scale require thousands of acres, and the land surrounding major facilities typically appreciates sharply as supporting infrastructure — logistics, hospitality, commercial — follows the workforce. Anyone who has watched what happened to real estate markets around Northern Virginia's data center corridor understands the pattern.

The energy development opportunity may be even larger than the data center itself. If SoftBank needs 10 GW of reliable power, somebody has to build it. That means new solar, wind, storage, and possibly nuclear capacity — with long-term offtake contracts that any infrastructure investor would find attractive. Power purchase agreements anchored by a creditworthy counterparty like SoftBank are exactly the kind of stable, long-duration revenue stream that institutional capital has been chasing.

For construction and engineering firms, a project of this scale sustained over years represents a generation-defining contract opportunity. The supply chain implications run deep: transformers, switchgear, cooling systems, fiber optic cable, backup generation — every one of those markets will feel the demand signal.

The ROI case for SoftBank itself is worth scrutinizing. AI compute is expensive to build and operate, but the revenue potential from leasing capacity to AI model trainers and inference workloads at scale is substantial. Masayoshi Son has historically been willing to absorb near-term losses for long-term positioning — his Vision Fund strategy was nothing if not conviction-driven. This project fits that profile exactly.

The Technology Stack Behind the Bet

SoftBank isn't just building warehouses for servers. The technology choices embedded in a project at this scale will reflect — and shape — where AI infrastructure is heading.

Modern AI training clusters are power-dense in ways that challenge conventional data center design. NVIDIA's GB200 NVL72 rack, for instance, draws roughly 120 kW per rack. Legacy data centers were designed around 5–10 kW per rack. That's a tenfold to twentyfold density increase, which means cooling architecture, power distribution, and structural engineering all have to be rebuilt from first principles.

Liquid cooling is no longer optional at this scale — it's the only viable path. Immersion cooling and direct-to-chip liquid systems are becoming standard for high-density AI workloads, and a facility targeting 10 GW will need cooling infrastructure that doesn't exist at commercial scale today. SoftBank's project will effectively have to develop new supply chains and installation methodologies as it builds.

There's also a software and interconnect dimension. At 10 GW of compute, the internal networking fabric of the campus becomes a critical engineering challenge. Moving data between GPU clusters at the speeds required for large-scale model training demands optical interconnects and network architectures that push the state of the art.

From an infrastructure planning perspective, the project also has to be future-proof against chip generations that don't exist yet. Today's H100s will be superseded. The physical plant — power, cooling, connectivity — needs to accommodate hardware that will be two or three generations forward by the time full buildout is complete.

What Comes Next

Projects of this ambition routinely compress timelines in press releases and expand them in execution. The real questions for SoftBank's Ohio project are about sequencing: Can they secure the grid interconnection commitments before breaking ground? Can they lock in the power supply at a cost that makes the economics work? And can they move fast enough to capture the AI infrastructure demand wave before competitors saturate the market?

The answers to those questions will play out over the next 18 to 36 months. Site selection announcements, utility negotiations, and PJM filings will be the leading indicators worth watching closely.

For stakeholders across the infrastructure and energy space, the immediate action is positioning. Land near plausible Ohio site locations is worth evaluating now. Energy developers with Ohio-adjacent generation assets should be mapping their contract opportunities. And anyone in the construction, cooling, or power equipment supply chain should be thinking about capacity before the RFPs land.

A project at 10 GW doesn't announce and then wait — it creates a wake that developers, investors, and communities either ride or get pulled under. The time to engage with what SoftBank is building in Ohio is before the details are public, not after. That's when the real opportunities close.

Explore the InfraSale Marketplace for more insights and opportunities!

[INTERNAL LINK: SoftBank AI Investment]

[INTERNAL LINK: Ohio Energy Market]

[INTERNAL LINK: Data Center Infrastructure Trends]