What Hyperscale Means, and Why These Data Center Campuses Keep Getting Bigger

On the evening of December 18, 2025, the auditorium of the Colleton County civic center in South Carolina was full. Residents, environmental groups, a state senator, and a candidate for governor had come to a three-hour zoning hearing over a single project: an 860-acre, nine-building data center campus proposed near Walterboro, at a reported cost of about $6 billion. At times the meeting grew loud enough that the zoning board warned it might stop. If the campus is built, it would draw about 1,000 megawatts of electricity, which would make it the first "gigawatt" data center in the state.

That word, gigawatt, is the reason to pay attention. A decade ago, a data center that used 50 megawatts of power was considered large. Today, companies are planning campuses that use twenty times that. This piece explains, from scratch, what a "hyperscale" data center is, why these sites keep getting bigger, and why the growth in size is making the local fights over them harder.

First, What a Data Center Actually Is

Start with the basics. A data center is a building, or a group of buildings, filled with computers that run all day, every day. The computers are called servers: machines with no keyboard or screen, built to store data and do the work behind the apps and websites people use. When you stream a show, send an email, back up a photo, or ask a chatbot a question, a server in a data center somewhere is doing part of that job.

The servers sit in tall metal frames called racks, lined up in rows like library shelves. Each rack holds many servers stacked on top of one another. A large facility can hold thousands of racks.

Most of what people call "the cloud" is just data centers like these. The cloud is not floating anywhere. It is physical buildings, owned or rented by companies, that store your files and run software so you do not have to keep it all on your own device. Some data centers are run by one company for its own use. Others are "colocation" sites, meaning a landlord builds the building and the power and cooling, then rents space inside to many different customers who bring their own servers. The word just means many tenants sharing one location.

What "Hyperscale" Means

There is no single legal definition of a hyperscale data center. No law sets the line, and no agency hands out the label. It is an industry term, and different people draw the boundary in different places.

The rough idea is size at an extreme. The name comes from "scale," and it points to facilities run by the handful of companies that operate the world's largest cloud and computing services. Historically the label attached to sites with tens of thousands of servers. More recently, people in the industry have tied it to electricity, because power is now the hardest thing to get. A common rule of thumb treats a campus that draws in the low hundreds of megawatts as hyperscale.

A quick word on the units, because the whole story runs on them. A watt is a measure of electric power. A megawatt is one million watts, very roughly enough to power several hundred homes. A gigawatt is one thousand megawatts, closer to the draw of a small city. When a project is described in gigawatts, it is asking for city-sized amounts of electricity in one place.

How a Hyperscale Campus Differs From an Ordinary One

The gap between a normal data center and a hyperscale one is not a small step. It is a different category of building.

An ordinary business data center might use a few megawatts and fit on a normal commercial lot. A hyperscale campus can cover hundreds or thousands of acres and use hundreds of megawatts, sometimes more. The Colleton County proposal sits on 860 acres. In Prince William County, Virginia, a project known as the Prince William Digital Gateway was approved in 2023 for up to 37 data center buildings across roughly 2,000 acres of rural land near the Manassas National Battlefield.

The other difference is who runs them. A small group of very large technology companies operate most hyperscale capacity worldwide. The research firm Synergy Research Group, which tracks the sector, counts companies such as Amazon, Microsoft, Google, and Meta among the main hyperscale operators. These are the firms behind the biggest cloud platforms and, increasingly, the biggest artificial intelligence systems.

Why They Keep Getting Bigger

Two forces are pushing the size up. Both are worth understanding plainly.

The first is simple growth in demand. Synergy Research counted more than 1,100 hyperscale data centers in operation worldwide at the end of 2024, with 137 new ones opened that year alone. The firm reported that total hyperscale capacity had doubled in under four years and was on track to double again in a similar span. The United States held about 54 percent of the world's capacity. Synergy's chief analyst, John Dinsdale, said the trend had "become supercharged in the last few quarters as companies build out AI-oriented infrastructure."

The second force is what happens inside the building. Artificial intelligence, the software behind chatbots and image generators, runs on specialized chips that draw far more power than the chips in an ordinary server. That changes the math of a single rack. According to AFCOM, an industry association that surveys data center operators each year, the average rack drew about 7 kilowatts of power in 2021 and roughly 16 kilowatts by 2025. A rack packed with the newest AI chips can draw well over 100 kilowatts on its own, many times what a conventional rack uses, according to industry reporting.

Put those together and the size climbs from both directions. There are more data centers, and each new AI-focused one wants far more power per square foot than the last. That is why the frontier has moved from the megawatt campus to the gigawatt campus in only a few years, a shift the trade press has been tracking closely.

Why Bigger Campuses Make the Local Fights Harder

A larger campus is not simply a larger version of the same argument. Size changes the substance of what a town is being asked to approve, and that is where the recent opposition tends to concentrate.

The first pressure is electricity. A gigawatt campus needs city-sized power, and that power has to come from somewhere. In Colleton County, opponents pointed out that the proposed campus would rely on a natural gas plant and a pipeline that are not expected to be operating until the 2030s. When a single customer asks for that much power, residents often ask who pays to build the new lines and plants, and whether their own electricity bills move as a result.

The second pressure is water. Many data centers use water to cool their equipment, and hotter, denser AI racks need more cooling. In the Colleton County case, the developer said the site would use about 12,000 gallons of water a day, which it compared to the use of 14 homes, while a state senator argued the draw would still pose a risk to the local river system. Reporting the claim and the objection side by side, each attributed to the person who made it, is the honest way to handle a disputed number.

The third pressure is land. A campus measured in hundreds or thousands of acres reshapes the area around it. In Prince William County, the objections centered on the rural setting, the nearby Civil War battlefield, and the water table. After a 29-hour public hearing, the county approved the rezoning in December 2023. Rezoning means changing what a piece of land is legally allowed to be used for, in this case from rural to data center use. Residents sued, and in 2025 a Virginia judge voided the rezonings, finding the county had not followed the state's public notice rules. An appeals court agreed, and the developer, QTS, later ended the project.

None of this settles whether any single campus is good or bad for the town that hosts it. That is a decision for the people and officials in each place. The point is narrower: as the buildings scale from megawatts toward gigawatts, the demands they place on power, water, and land scale with them, and so does the size of the public argument.

The Bottom Line

Hyperscale has no fixed definition, but the direction is clear. The largest data centers have moved from tens of megawatts to hundreds, and the newest AI projects are being drawn up in gigawatts, a scale that a decade ago would have described a power plant rather than a computing site. That growth is why the same three words keep coming up in town halls from South Carolina to Virginia: power, water, and land.