Tom Dakich

A city rejects a billion-dollar project. Fine. Sometimes that's the right decision. But what happens to the vacant industrial property? The aging infrastructure? The shrinking tax base? The jobs that might have been created? Those problems don't disappear because the project does. Communities should scrutinize every major development. They should also understand the cost of saying no.

The AI boom has created a new type of real estate expert. Someone who owns land near a transmission line and suddenly has a “500 MW data center campus.” No committed power. No customer. No financing. But the rendering looks fantastic. I've been around development long enough to know the difference between an opportunity and a project. The next few years will expose that difference.

One of the biggest mistakes in infrastructure development is treating the community as the last step before approval. Buy the land. Secure power. Design the project. Then introduce yourself to the people who live there. That's backwards. At Digital Crossroads in Hammond, working with local leadership and the building trades was part of development from the beginning. You can accelerate engineering and construction. You can't always accelerate trust. And developers who ignore that usually discover how expensive the difference can be.

A billion-dollar investor arrives in a small town with lawyers, engineers, consultants, and capital. But nobody understands the town. The utility relationships. The history of the property. The local workforce. The political realities. The promises previous developers failed to keep. That's where experienced local developers become invaluable. Capital can get you into a market. Local knowledge helps you succeed there. And I think the AI infrastructure industry is seriously underestimating that distinction.

A billion-dollar rendering is not a billion-dollar asset. I'm seeing enormous AI campuses marketed based on projected capacity, projected demand, and projected valuations. Fine. But who's the customer? Where's the deliverable power? Who's financing construction? What happens if the economics change? I've spent enough time in infrastructure development to know that the market eventually separates presentations from projects. The next AI winners won't necessarily be the biggest announcers. They'll be the ones who can actually deliver.

The biggest risk in the AI data center boom may not be building too little. It may be building the wrong infrastructure. AI hardware and computing demand can change dramatically in a few years. Buildings, substations, and transmission infrastructure last decades. That mismatch should concern every serious developer. Build in phases. Secure real demand. Design for flexibility. The goal isn't to build the biggest campus. It's to build one that still makes economic sense ten years from now.

A city rejects a billion-dollar project. Fine. Sometimes that's the right decision. But what happens to the vacant industrial property? The aging infrastructure? The shrinking tax base? The jobs that might have been created? Those problems don't disappear because the project does. Communities should scrutinize every major development. They should also understand the cost of saying no.

Owning 200 acres near a substation doesn't make you a data center developer. Neither does a rendering of six buildings labeled “500 MW AI Campus.” Can the utility deliver the power? Is the fiber there? Is the project financeable? Is there a real customer? I've seen plenty of impressive development presentations. The industry is about to find out who can actually build what's in them.

Here's a question more people should be asking about AI data centers: What happens when a utility builds a new substation or transmission infrastructure for a 300 MW project that never gets built? Somebody still has to pay for it. The technology company? Utility shareholders? Existing ratepayers? If developers want utilities to make long-term infrastructure commitments, they should be prepared to make binding financial commitments in return. AI moves fast. The electric grid can't afford to operate on speculation.

A billion-dollar project comes to town. Great headline. Big groundbreaking. Lots of speeches. Now ask who owns the development, who gets the contracts, where the suppliers come from, and where the profits go. Outside investment is important. But a community needs local developers, contractors, and businesses capable of participating in it. Attracting capital and capturing opportunity are not the same thing.

Somebody is going to lose a fortune building data centers for demand that never materializes. AI is real. The infrastructure opportunity is real. But a proposed 500 MW campus isn't a business until there's deliverable power, financing, and customers willing to pay for the capacity. I've seen enough development cycles to know that the biggest announcements don't always produce the best investments. The AI boom will create winners. It will also expose people who confused a presentation with a project.

AI is driving enormous centralized data centers. At the same time, AI is moving into factories, hospitals, vehicles, cameras and physical infrastructure. Those two trends will eventually meet. Not every workload belongs hundreds of miles away. That's why I think edge computing is being underestimated. Not every city needs a 1,000 MW campus to have a place in the AI economy.

The most expensive megawatt may be the one somebody promised you five years from now. AI developers are chasing enormous power commitments. But theoretical capacity isn't the same as deliverable power. I'd rather have 20 MW I can actually use than 200 MW that exists in a presentation. In the next phase of data center development, certainty may become more valuable than size.

I think we're building some data centers backwards. Find land → secure power → design project → get permits → explain it to the community. Try reversing the thinking. Understand the place first. Its infrastructure. Industry. Workforce. Existing properties. Actual needs. Then decide what belongs there. A good project needs a reason for being there, not just a reason for being built.

Construction eventually ends. The more interesting economic-development question is what remains afterward. Did local contractors gain new capabilities? Did trade schools develop new programs? Did suppliers enter new markets? Did local developers become better positioned for the next investment? A successful project should do more than create activity while it's being built. It should leave the local economy stronger.

A feasibility study can tell you about land, power, zoning and construction costs. It can't tell you everything about how a community actually works. That's where experienced local developers become valuable. They understand the people, history, workforce and relationships behind the numbers. Local knowledge is an infrastructure asset in its own right.

A billion-dollar project can come to town without creating a billion dollars' worth of local opportunity. The difference often comes down to what's already there to capture it: local developers, contractors, skilled trades, suppliers and training pipelines. Attracting capital is one challenge. Making sure the community participates in the opportunity is another.

One factor the industry can’t ignore anymore is community resistance. The pushback against large-scale data center projects is becoming more organized and more visible. Some concerns are reasonable. Some are emotional. Some are political. But regardless of the reason, local alignment is becoming just as important as financing and infrastructure planning.

Servers become obsolete. Good infrastructure doesn't have to. Twenty years after a technology project opens, the real legacy may be the power infrastructure, fiber, skilled workforce, productive property and expanded tax base it leaves behind. Don't only ask what's being built inside the fence. Ask what value remains outside it.

Not every computing problem needs a 500-acre campus. Some workloads benefit from being closer to factories, hospitals, transportation systems and the communities generating the data. Hyperscale has a role. So does regional compute. So does the edge. The future of AI infrastructure may be far more distributed than the headlines make it look.