Who Insures a $20 Billion AI Data Center?

Building a large AI data center can cost more than $20 billion.

And that may only be the beginning. Once construction is finished, the facility has to be filled with GPUs, servers, cooling systems, transformers, batteries, and other expensive equipment.

That raises a simple question:

What happens if something goes seriously wrong?

If your car is damaged, the answer is fairly straightforward. Your insurance company pays for the covered loss.

A $20 billion data center is different. The risk can be too large for one insurer to comfortably take on alone.

So the risk gets shared.

How Do You Insure Something This Expensive?

Imagine a huge AI data center project.

Instead of one insurance company taking all the risk, several insurers can participate. One takes a portion, another takes another portion, and more insurers may join.

Those insurers can also pass some of their own risk to reinsurance companies.

Reinsurance is basically insurance for insurance companies.

This means a single data center can have a surprisingly large financial network sitting behind it. What looks like one building from the outside may have many companies sharing the risk.

This is not new. Very large assets such as aircraft, power plants, and skyscrapers have long used insurance and reinsurance to spread large risks.

What is changing is the size of the data centers.

AI Is Making Data Centers Much More Valuable

Traditional data centers already contained expensive servers and electrical equipment.

AI data centers take this much further.

Thousands of high-performance GPUs can sit inside a single facility. Those GPUs consume enormous amounts of electricity and produce enormous amounts of heat.

That means the facility also needs more electrical infrastructure and more powerful cooling systems. Some projects may include large batteries and their own power generation.

Swiss Re says construction costs for large AI data centers can exceed $20 billion. Once the technology equipment is installed, the total value at the site can be much higher.

For insurers, this creates a simple problem:

There is suddenly much more money sitting in one place.

The Accidents Are Often Surprisingly Ordinary

The risks are not limited to dramatic cyberattacks or catastrophic explosions.

Water leaks. Electrical equipment fails. Cooling systems malfunction. Fires happen.

Allianz Commercial analyzed around €700 million in data center insurance losses and found that fires accounted for more than half of the value of those losses.

A serious fire can damage more than servers. Power systems can fail. Cooling can stop. The facility may not be able to serve customers while repairs are being made.

So the financial loss can extend beyond the cost of replacing damaged equipment. There is also the money lost while the data center is not operating.

AI introduces another interesting problem: liquid.

Modern AI chips generate so much heat that data centers are increasingly using liquid cooling systems to keep them running efficiently.

Putting liquid close to billions of dollars of computing equipment is manageable, but it creates another risk that insurers have to understand and price.

The more powerful the computers become, the more complicated the infrastructure around them becomes too.

One Storm Can Hit More Than One Data Center

Data centers are not spread evenly across the world.

They tend to cluster in places with access to electricity, fiber networks, land, and major customers.

That creates concentration risk.

A major storm, flood, wildfire, or other natural disaster may not damage just one data center. It can affect several facilities in the same region.

Swiss Re has pointed to this problem in the United States, where significant amounts of data center capacity are located in areas exposed to hazards such as hail and tornadoes.

For an insurer, that changes the calculation.

Insuring one expensive building is one thing. Insuring several expensive buildings that could all be damaged by the same event is another.

The Insurance Industry Is Bringing in More Capital

The insurance market is already responding.

In August 2026, Marsh launched a new insurance market called Stratus for large digital infrastructure risks.

It can provide access to as much as $10 billion of property insurance capacity in a single placement.

That does not mean one insurance company is promising to pay $10 billion.

Stratus brings together 30 traditional insurers and alternative capital providers. Each can evaluate the risk and decide how much it is willing to take.

That structure tells us something important about the AI infrastructure boom.

The buildings have become so valuable that the insurance industry is finding new ways to bring more capital together to cover them.

And even $10 billion is an interesting number when large AI data center projects can cost more than $20 billion.

That does not mean the remaining $10 billion is uninsured. Insurance limits do not simply have to equal the total value of a project. Coverage depends on the type of risk, the stage of construction, expected maximum losses, and the terms of each policy.

But it shows why insurance capacity itself is becoming an important part of the data center business.

Insurance Can Matter Before Anything Goes Wrong

There is another reason insurance matters.

A company building a $20 billion data center is unlikely to pay for the entire project from its own bank account.

It may borrow from banks or raise money from other lenders and investors.

Now think about the project from the lender's side.

You lend billions of dollars to help build a data center. Halfway through construction, a major fire destroys part of the project.

Your money is at risk too.

That is why lenders care about insurance.

Allianz Commercial says comprehensive insurance coverage is becoming a prerequisite for financing many large AI infrastructure projects.

So insurance is not only something that matters after an accident. It can help make the project possible in the first place.

An AI company needs more computing power.

A data center company builds a new facility.

Banks and investors provide some of the capital.

Insurers and reinsurers take on some of the risks that could damage that investment.

AI data centers need more than GPUs and electricity.

They also need someone willing to take the financial risk if things go wrong.

AI Is Creating a Bigger Insurance Market

The global data center insurance market is estimated at roughly $11 billion today.

It is expected to grow to more than $24 billion by 2030.

That makes sense when you look at what is happening inside these facilities.

More GPUs. More electricity. More cooling. More batteries. More valuable equipment concentrated in the same place.

The AI boom is creating demand for chips, transformers, power plants, and cooling equipment.

But all of those expensive assets create something else too:

risk.

So the next time you see plans for a $10 billion or $20 billion AI data center, there is one more question worth asking.

Who is supplying the GPUs?

Who is supplying the electricity?

Who is providing the financing?

And who has agreed to pay if something goes wrong?

There is an entire industry behind that last question.

BEYOND THE OBVIOUS.

Sources

Swiss Re — Insuring AI: Data Centre Value Accumulation Risks

Allianz Commercial — The Data Center Construction Boom: Risks and Claims Trends

Marsh — Stratus: $10 Billion Property Insurance Exchange for Digital Infrastructure