Part 3: The 100-Gigawatt Answer Hiding Inside America’s Power Grid

The biggest obstacle facing new data centers may not be land, water, or even money. It may be the time it takes to connect them to the electrical grid.

A large data center can wait five to seven years for new power lines, substations, or generation to be built. That delay is a problem for technology companies, but it is also a problem for local communities. When a utility decides it needs new infrastructure for one enormous customer, residents may worry that part of the cost will eventually show up on their monthly bills.

At first, the answer seems obvious: build more power plants.

But there may be another option hiding inside the grid we already have.

The Grid Is Built for the Worst Day, Not the Average Day

Think about a grain elevator during harvest season.

For a few weeks, trucks may line up from morning until night. The equipment, roads, and staffing have to be ready for the busiest days of the year. But outside harvest season, much of that same capacity sits unused.

The electrical grid works in a similar way.

Utilities must build enough power capacity to survive the hottest summer afternoon, when air conditioners are running across entire regions at the same time. If the grid cannot handle that peak, homes and businesses can lose power.

The system therefore has to be built for the worst few hours of the year.

During most other hours, some of that capacity is still available.

That unused room is the main idea behind the 100-gigawatt answer.

Researchers examined large regional grid systems across the country and asked a simple question:

How much new electricity demand could the existing grid handle if very large users agreed to temporarily reduce their usage during the most stressful hours?

Their answer was roughly 98 to 100 gigawatts.

That is an enormous amount of electricity. In the source material, it was described as more power than all data centers worldwide were using at the time.

The surprising part is how little curtailment would be required.

If a large customer agreed to reduce its demand for about 0.5% of the year, that would equal fewer than 44 hours annually. In many cases, those interruptions would come in short periods of roughly two hours.

In plain English, the data center would use the grid normally almost all year. During a handful of extreme peak events, it would agree to step back.

Why This Matters to a Local Community

Without this kind of agreement, a utility may say it needs a new power plant, new transmission lines, or a new substation before the data center can connect.

Those projects are expensive and slow.

A flexible-load agreement can change the conversation.

Instead of building everything from scratch, the utility can connect the data center to capacity that already exists during most hours. The data center gets power sooner. The utility uses equipment it has already paid for. Residents are less likely to be asked to support new infrastructure built mainly for one private customer.

That does not mean the connection is free or that the company should avoid paying its share. The company should still pay dedicated large-load rates, fund the equipment it directly requires, and accept long-term financial commitments.

The point is that not every new data center automatically requires an entirely new power plant.

Sometimes the better answer is to use existing capacity more intelligently.

What “Curtailment” Actually Means

Curtailment is an industry word for temporarily using less electricity.

That does not necessarily mean shutting down the entire data center.

Some computer work must continue without interruption. Hospital systems, emergency communications, payroll services, and live customer traffic cannot simply stop because it is hot outside.

Other work is more flexible.

Training a large artificial intelligence model can take weeks or months. Pausing that work for two hours usually does not destroy the project. The work can resume later that evening when demand on the grid drops.

A company may also move certain computer jobs through fiber connections to another data center in a region where electricity demand is lower at that moment.

This is sometimes called workload shifting, but the idea is straightforward: move the work instead of building a new power plant for a few difficult hours.

Battery storage may provide another option. A facility could use stored power during a peak event and recharge later.

The important question is not simply whether the data center leaves the grid. The question is how it reduces its demand.

The Diesel Generator Loophole

This is where the contracts matter.

Imagine that a data center promises to reduce its grid use by 50 megawatts during an emergency.

The utility calls and says the system is under stress. The data center disconnects from the grid, then starts a field of diesel generators so its servers can continue running at full speed.

On paper, the company reduced its demand from the grid.

For the nearby community, the problem may have become worse. Residents now have diesel exhaust, engine noise, and fuel trucks operating close to homes and businesses.

The pollution was not eliminated. It was moved.

A proper agreement must therefore explain how curtailment will happen. It should favor actual workload reductions, shifting work to another location, or using cleaner stored energy. It should not allow a company to claim success by replacing grid electricity with large amounts of local diesel generation.

That distinction may sound technical, but it determines whether the policy protects the community or merely protects the utility’s paperwork.

Do Residents Need to Know PJM, MISO, or ERCOT?

Probably not in detail.

Those names refer to large regional organizations that help manage electricity supply and demand. They act somewhat like air-traffic controllers for the power grid. They watch how much electricity is being used and make sure enough generation is available.

Different parts of the country have different organizations:

PJM serves a large portion of the Mid-Atlantic and parts of the Midwest.

MISO manages a broad section of the central United States.

ERCOT manages most of Texas.

Other regions have their own operators.

The names matter to engineers and utility planners because each region has different limits and available capacity. For the average resident, the larger point is enough:

Several major regions appear to have unused room during most hours of the year.

The source material gave examples of estimated headroom in different regions, including roughly 18 gigawatts in PJM, 15 in MISO, and 10 in ERCOT. Those numbers help show that the opportunity is not limited to one state.

But a local decision should never rely on a national estimate alone.

The community needs its own independent study. The question is not, “Does the country have spare capacity?” The question is, “Does our part of the grid have enough capacity, under what conditions, and who carries the cost if the estimate is wrong?”

The Same Principle Applies to Water

This idea of shifting demand away from peak hours also works for water systems.

A data center may need water throughout the day for cooling. But it does not always need to pull that water from municipal pipes at the same moment it uses it.

If the facility builds large storage tanks, it can fill them overnight when household demand is low. Then it can draw from those tanks during the evening, when residents are cooking, bathing, washing clothes, and watering livestock or gardens.

This may not reduce the total amount of water used over the entire year. It can still reduce pressure on the town’s treatment plant and distribution system during the busiest hours.

The same basic rule applies to both electricity and water:

Do not judge infrastructure only by total annual use. Pay attention to when the demand happens.

Peak demand is often what forces a community to build expensive new systems.

A Better Question for City Leaders

The wrong question is:

“How much power does this data center need?”

The better questions are:

How much does it need during ordinary hours?

How much does it need during the hottest day of the year?

Which computer jobs can pause or move?

How many hours of curtailment will the company accept?

Will diesel generators be allowed to satisfy that promise?

Who pays for new substations, transmission lines, and backup capacity?

What happens if the company leaves after the utility has already built the infrastructure?

Those questions turn a vague promise into an enforceable operating plan.

The Practical Opportunity

The 100-gigawatt idea does not prove that every data center can connect tomorrow. Local conditions still matter. Some regions may have enough generation but not enough transmission. Some communities may face limits at the substation or distribution level. Some facilities may carry workloads that cannot be paused.

The idea does prove something important: the country may not need to build every new data center around a brand-new power plant.

There is room to negotiate.

A company that accepts limited, carefully controlled flexibility may connect faster. A utility may avoid unnecessary construction. Residents may avoid paying for equipment that serves one customer. The grid may become more efficient because existing capacity is used more fully.

That is the real message behind the calculations.

The power system is not an empty bucket that must be filled from scratch every time a large customer arrives. It is a system built with extra room for the worst few hours of the year.

The opportunity is to use that room without putting reliability at risk.

For farmers, residents, and local leaders, the conclusion can be stated without an acronym:

Before approving a new power plant for a data center, first determine whether the company can use the grid we already have—and agree to step back during the few hours when everyone else needs it most.