The AI Buildout Is Becoming a Power-and-Capital Consortium

The headlines said “8 GW.” The real story was everything required to make 8 GW possible.

This week, AI infrastructure stopped looking like a data-center construction boom and started looking more like a national industrial coordination exercise.

OpenAI, NVIDIA, SB Energy, AEP Ohio, the U.S. Department of Energy, investors, regulators, and Pike County are all now attached to one planned campus in southern Ohio. That alone tells you where the market is headed.

The future of AI is not being negotiated solely in chip labs or cloud contracts.

It is being negotiated across transmission corridors, project-finance models, utility tariffs, brownfield sites, community-benefits agreements, and the very uncomfortable question of who gets curtailed when the grid is short.

Welcome to This Week in Infrastructure.

The Big Story: 8 GW Is Not a Data Center Announcement

OpenAI has agreed to secure approximately 8 GW of IT capacity at SB Energy’s PORTS-Pike Technology Campus in Pike County, Ohio. NVIDIA will be the exclusive AI-compute infrastructure provider, invest $1.5 billion in SB Energy, and provide credit support associated with the project’s land, power, and shell buildout. SB Energy will build, own, and operate the campus under a 20-year lease to OpenAI. OpenAI | SB Energy

That is not one company “building a data center.”

That is a consortium assembling an AI campus at infrastructure scale.

The Pieces Include:

• OpenAI: long-term demand for AI capacity

• NVIDIA: compute infrastructure and capital support

• SB Energy: campus development, land, power, and shell

• AEP Ohio: transmission and grid delivery

• U.S. government: site and regional coordination

• Pike County: workforce, local approval, and long-term community participation

The companies say the broader plan includes at least 10 GW of new generation, plus at least $4.2 billion in regional grid infrastructure. AEP Ohio says the developer will fund that transmission investment in an effort to avoid shifting its cost to Ohio ratepayers; AEP expects power to begin flowing in 2029. AEP Ohio

That ratepayer language is doing a lot of work.

For years, the AI infrastructure conversation has been: Can we build enough compute?

Now it is: Can we build enough compute without making everyone else pay for the wires?

IOI Take: A gigawatt-scale AI campus is not a real-estate product. It is a regional infrastructure compact.

Power: PJM Is Quietly Writing the New Rules

While Ohio made the biggest headline, PJM delivered the more important operating signal.

The grid operator, which serves 67 million people across a territory from Washington, D.C., to Chicago - has proposed an emergency procedure that would ask utilities to reduce or transfer demand from data centers and other large power users before shutting off traditional customers. PJM does not currently have the authority to force curtailment and would need cooperation from state governments. Reuters

Translation: the large-load contract is changing.

A few years ago, a data center’s power plan largely focused on getting an interconnection agreement. Increasingly, major users may need to prove how they will behave when the system is stressed.

That could mean demand response. Flexible load. Onsite generation. Storage. Contractual curtailment. Or a combination of all four.

Golden Signal: Watch for utilities and regulators that begin treating flexible consumption as part of the price of fast interconnection.

The first question in site selection is no longer just, “How much power can you give me?”

It is becoming, “How much flexibility will you require in return?”

Capital: NVIDIA Is No Longer Sitting Outside the Deal

NVIDIA’s Ohio investment is part of a much larger shift: the AI supply chain is increasingly financing the infrastructure required to consume its products.

Last week, NVIDIA announced a partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR aimed at creating compute-financing platforms and attracting more than $500 billion in third-party capital for AI infrastructure. Reuters reported that NVIDIA could backstop up to $125 billion of those deals. Reuters

The logic is straightforward.

A GPU supplier can only sell into demand that has somewhere to live. That “somewhere” needs land, grid access, buildings, cooling, capital, and customers willing to sign long enough contracts to support financing.

In other words: chips do not create deployable capacity by themselves.

IOI Take: The AI capital stack is becoming more integrated because the physical buildout is too large, too slow, and too strategically important to leave to disconnected counterparties.

That does not eliminate risk. It concentrates it.

Investors will be watching the durability of the tenant commitment, the cost of power, the financing terms, the useful life of rapidly evolving compute, and whether project-scale infrastructure can be built before technology moves to its next generation.

Land: Brownfield Is Back

PORTS-Pike is being developed at the former Portsmouth Gaseous Diffusion Plant site, on a campus spanning private and federal land. SB Energy

The location matters.

Former industrial sites increasingly offer something ordinary greenfield acreage cannot: scale, existing infrastructure context, a history of heavy industrial use, and a clearer path toward multi-party coordination.

Ohio is not alone. This month, Brookfield and NextEra outlined plans to transform part of the former Paducah Gaseous Diffusion Plant in Kentucky into a planned 1.2-GW AI data-center campus with dedicated onsite power generation. Data Center Knowledge

The New Land Hierarchy Looks Like This:

Land → Industrial Land → Data Center Land → Powered Land

Every step toward real infrastructure readiness changes the site’s strategic value.

There is plenty of land in America.

There is considerably less land where the grid, generation, permitting, transmission, fiber, workforce, and community path can all line up at the same time.

Sustainability: The Power Plan Is Getting More Complicated, Not Less

The AI buildout is not resolving into one preferred energy strategy.

It is becoming an all-of-the-above market.

In Singapore, Equinix signed its fourth renewable-energy power-purchase agreement in two years, this time with Flo Energy Singapore. Reuters

At the same time, the Ohio project is being developed around major new generation and transmission infrastructure, with reporting pointing to natural-gas generation as part of the broader power plan. Reuters

That is the reality of the AI power conversation.

The market wants reliable, fast, affordable, lower-carbon electricity. In many regions, those objectives do not arrive on the same timeline.

So developers are assembling portfolios: utility supply, renewables, storage, gas, nuclear discussions, demand flexibility, and more direct control over generation.

What Matters: not the press-release energy label, but whether the project has a credible path to firm, deliverable power.

Policy & Community: “Who Pays?” Is the New Permitting Question

The projects getting serious attention are no longer evaluated solely on investment dollars and job counts.

Communities Want to Know:

• Who pays for new substations and transmission?

• Who pays if the water system needs expansion?

• What happens to residential electricity bills?

• How many permanent jobs will exist after construction?

• What commitments are enforceable?

• Can the data center reduce load during emergencies?

OpenAI says the Ohio project could create 35,000 construction jobs over a six-year buildout and 2,500 long-term operating jobs. SB Energy and OpenAI have announced a combined $80 million in community-benefits commitments. OpenAI | SB Energy

But community acceptance is not automatic. Reuters reported that at least 75 projects worth roughly $130 billion faced local opposition in the first quarter of 2026, citing Data Center Watch research. Reuters

The takeaway is not that data centers cannot be built.

It is that local consent is becoming an execution variable, right alongside power, capital, and permitting.

IOI Take: Community engagement is not the final communications workstream. It is part of the project’s bankability.

The Number to Remember: $4.2 Billion

That is the planned investment in regional grid infrastructure associated with the PORTS-Pike campus.

Not the data center shell.

Not the chips.

Not the total capital stack.

The grid infrastructure.

That is why the AI race is becoming an infrastructure race.

What IOI Is Watching

1. PJM’s Large-Load Framework:

Project funding is one thing; approvals, equipment, routes, and construction are another.

2. Ohio’s Transmission Milestones:

A $10 billion headline means less if the agreement can be delayed, reduced or canceled. Deposits, credit quality and enforceable commitments matter.

3. The Generation Mix Behind PORTS-Pike:

A campus of this size needs a detailed, phased, firm-power plan.

4. Who Follows NVIDIA iInto Physical-Capacity Underwriting:

Other chipmakers, cloud platforms, and asset managers are watching closely.

5. Ratepayer-Protection Structures:

The market is moving toward a new standard: build fast, but pay your own way.

The Bottom Line

This week did not just produce a bigger AI-data-center number.

It showed the new shape of the deal behind the number.

The AI tenant needs compute. The chip company needs deployment. The developer needs land and capital. The utility needs a grid plan. The community needs protection and return.

The next winners in the Intelligence Economy will be the ones that can coordinate all of it.

The headline gives you the number. IOI maps the systems required to make the number real

Infrastructure of Intelligence™ provides research and analysis for informational purposes only. Nothing in this article constitutes investment advice or a recommendation to buy or sell any security

Workforce

Policy

Capital

Sustainability