The Data-Center Divide: The Cloud Has a Physical Address

Before we judge data centers, we should understand what they are, what they do, and why artificial intelligence is changing the infrastructure behind our digital lives.

Most of us can go an entire day without thinking about a data center. We check email, stream music, open a banking app, search Google Maps, join a video call, upload a photograph, or ask an AI assistant a question. The experience feels almost weightless. Tap a screen, and information appears.

But the digital world is not weightless.

Behind those everyday actions are physical buildings filled with computers, storage systems, fiber connections, electrical equipment, cooling systems, backup power, security, and people keeping all of it running.

That is a data center.

A useful way to think about one is as a utility building for the digital economy — a place where the services we casually call “the cloud” become physical.

And before communities decide whether these facilities are good, bad, welcome, or unwanted, there is a more basic step:

First understand the thing. Then judge it.

WHAT MATTERS: WHERE THE DIGITAL WORLD BECOMES PHYSICAL

A data center is a facility built to run large amounts of computing equipment reliably, securely, and continuously.

Some are relatively small. Others stretch across enormous campuses. Some serve one company. Others host equipment for many customers. Increasingly, the largest facilities support cloud computing and artificial intelligence at extraordinary scale.

Yet the basic job is simple.

Data centers process information, store information, and move information.

Inside, the operation can be understood through four basic systems:

Compute — the brains.
Servers perform calculations and run the software behind websites, business applications, financial transactions, cloud services, and AI.

Storage — the library.
Storage systems hold photographs, documents, videos, databases, backups, applications, and the information modern organizations depend on.

Networking — the roads.
Fiber-optic cables, switches, and routers move information between computers, between data centers, and ultimately to the device or organization requesting it.

Power + cooling + resilience — the life-support system.
Computers need electricity and produce heat. Data centers therefore require electrical distribution, cooling, backup systems, monitoring, fire protection, physical security, and redundancy designed to keep services available when equipment fails.

That fourth system matters to the public debate.

A data center is not simply a warehouse where someone plugs thousands of computers into the wall. The supporting infrastructure is part of the facility.

That is why conversations about data centers quickly become conversations about power, substations, cooling, water, fiber, generators, and land.

WHY WE RARELY NOTICE THEM

One of the strange things about data centers is that their success makes them almost invisible.

We rarely notice data centers precisely because they are designed so that we do not.

If one server fails, another can often take over. If a power source is interrupted, backup systems are designed to respond. Major digital services can be spread across multiple servers and sometimes multiple facilities or regions.

Reliability is part of the product.

Imagine tapping “play” on a movie.

Your request travels across a network. A server receives it. Software processes the request. Storage provides the content. Networking equipment sends the data back to you. Electrical and cooling systems keep everything operating.

Moments later, the movie is playing.

Now multiply that process by billions of searches, transactions, messages, videos, business operations, photographs, and AI prompts.

The “cloud” starts looking a lot less like a cloud.

It looks like infrastructure.

FOUR DATA-CENTER TERMS WORTH KNOWING

The industry can become technical quickly, but citizens do not need an engineering degree to understand the basic landscape.

Cloud

Cloud computing allows people and businesses to use computing power, storage, and software over a network instead of owning and maintaining all of the underlying equipment themselves.

Think of it like purchasing electricity rather than owning your own power plant.

The physical computers still exist. They are simply housed and managed in data centers operated by companies such as Amazon, Microsoft, and Google.

Colocation

A colocation facility rents secure space, power, cooling, and connectivity to customers that place their own computing equipment inside.

Think of it as specialized commercial real estate for computers.

Hyperscale

Hyperscale data centers are the giants — facilities or campuses designed to support enormous volumes of cloud computing, storage, and increasingly artificial intelligence.

These projects tend to dominate headlines because their physical size and infrastructure requirements can be substantial.

Edge

Edge computing moves some processing closer to the people, devices, or machines generating the data. That can reduce delay for applications such as industrial systems, logistics, connected devices, telemedicine, and real-time analytics.

These categories can overlap. The point is simply to recognize that “data center” describes an ecosystem, not one identical type of building.

YOU PROBABLY USED A DATA CENTER BEFORE BREAKFAST

The easiest way to understand why data centers matter is to stop thinking about the buildings and think about what they enable.

Email. Online banking. Streaming. Navigation. Cloud-based business software. School portals. Telehealth. Digital payments. Photo backups. Social media. Government systems. Artificial intelligence.

Somewhere behind each service is physical computing infrastructure.

Businesses increasingly depend on cloud systems rather than maintaining every server themselves. Governments and institutions rely on digital records and applications. Financial networks process transactions continuously. AI adds another rapidly growing layer of computing demand.

Data centers are not important because technology companies say they are important.

They are important because modern life increasingly runs through them.

THEN AI CHANGED THE SCALE

Data centers are not new. Artificial intelligence did not invent them. What AI is changing is the concentration of computing inside them.

Traditional data centers were built primarily around conventional servers handling websites, databases, business software, cloud applications, storage, and other digital workloads.

AI adds a different challenge.

Training advanced AI models can require large numbers of specialized processors working together on enormous datasets. Once a model has been trained, inference is the process of putting that model to work — answering a question, recognizing an image, recommending a product, or detecting fraud.

A simple way to remember the difference is:

Training is teaching. Inference is using what was learned.

That leads to one of the central ideas of this series:

AI does not change what a data center is. It changes how much computing, power, cooling, and networking can be concentrated inside it.

WHY DENSITY CHANGES THE CONVERSATION

For years, most data-center racks operated at relatively modest power levels. Uptime Institute’s 2025 industry survey found that most operators still reported common rack densities in the single digits, with the average near 9 kilowatts.

Leading AI systems occupy a very different category. Current AI racks can exceed 40 or 50 kilowatts, and some exceed 100 kilowatts. NVIDIA’s GB200 NVL72 rack-scale system, for example, consumes roughly 120 kilowatts.

The exact numbers matter less than the direction.

We are packing vastly more computing into the same physical space.

More computing creates more heat. More heat requires more sophisticated cooling. Higher electrical loads require different power infrastructure. At the highest densities, traditional air cooling increasingly gives way to liquid systems that remove heat directly from the hottest components.

This helps explain why communities are suddenly hearing much more about transmission lines, substations, cooling technology, and infrastructure planning.

WHY ARE SO MANY BEING BUILT NOW?

AI is the accelerant, but it is not the whole story.

Cloud adoption was already growing. Streaming, online commerce, cybersecurity, connected devices, digital payments, social media, remote work, and enormous volumes of data were already increasing demand for computing.

AI arrived on top of those trends.

Lawrence Berkeley National Laboratory found that U.S. data-center electricity consumption grew sharply through 2023 and projects substantial additional growth through 2028, with the future range heavily influenced by the number and use of AI servers.

The expansion is not occurring because someone suddenly decided America needed more giant buildings.

The buildings are a physical response to growing demand for digital services.

That does not mean every proposed facility belongs in every location. It does mean the underlying infrastructure demand is real.

FIRST UNDERSTAND THE THING. THEN JUDGE IT.

This is where the public debate should begin.

Data centers consume land. They need electricity. Cooling matters. Infrastructure has to be built. Communities have legitimate questions about water, noise, utility costs, taxes, jobs, environmental impacts, and who should pay for supporting infrastructure.

Those questions deserve serious answers.

But they should be asked with a clear understanding of what the facility actually is and why it exists.

A data center is where the digital world becomes physical.

It is where the cloud becomes servers, where an AI prompt becomes computation, and where the convenience of modern digital life becomes buildings, wires, cooling systems, and infrastructure somewhere in the real world.

That creates tradeoffs. And tradeoffs deserve something better than reflexive boosterism or reflexive opposition.

In Part II of The Data-Center Divide, we will examine the honest ledger: what data centers actually cost communities, what benefits they can create, which criticisms are supported by evidence, which claims leave out important context, and what communities can do to protect themselves while capturing the upside.

Part I does not ask the reader to support data centers. It asks the reader to understand them well enough to evaluate the debate intelligently.

Because understanding should come before judgment.

And better decisions begin with seeing the whole picture.

— Matt Cucinotta | Growth Solutions KC | Inspire · Inform · Ignite