Skip to main content
Marsh logo

Expertise

Data center power risk and resilience

Know what power failure could cost before it reaches the balance sheet.

Download the report

Power risk starts with infrastructure and contracts, but the consequences rarely stay there.

Power has become one of the main keystones of the data center ecosystem. It can influence where a site can be built, how quickly capacity can come online, what tenants can be promised, and how investors and lenders assess the asset.

You may already have the engineering answer: on-site, hybrid or behind-the-meter generation; N+1 or 2N power architecture; uninterruptible power supply (UPS) systems; generators; backup fuel arrangements. You may also be negotiating or renewing power purchase agreements (PPAs), or carrying service-level agreement (SLA) exposure tied to uptime.

There is no getting around those questions. But just because you've got answers to them, doesn't mean your data center project is immune to power risk. What happens if a power event did occur?

Resilience engineering can reduce the probability of failure. It does not reduce what the failure costs.  Revenue disruption, SLA penalties, tenant disruption, emergency costs, reputational damage and financing pressure can still sit on the balance sheet. That is why power risk needs to be looked at as a power system, a revenue dependency, a contract exposure and a capital question at the same time.

The data center power risk landscape

The data center power risk landscape does not start and end with an outage. It runs across utility supply, generation, backup systems, fuel supply, equipment and capacity. A site can be engineered for resilience and still carry dependencies that only become visible when the system is under pressure.

And this is not a theoretical outage risk. According to the Uptime Institute’s Annual Outage Analysis 2025, power is the leading cause of significant data center outages, accounting for roughly 45%, most often involving UPS-related issues. That is why the failure mode is important because a utility outage is not the same as undervoltage, a UPS battery-string failure, a generator failure-to-start or fuel delivery disruption.

Grid risk is the obvious starting point. Utility outage, undervoltage, capacity constraints or interconnection delay can all affect whether a facility receives the power it needs, when it needs it. As demand rises, the grid is no longer background infrastructure that can simply be assumed.

Backup architecture carries its own failure points as well. UPS battery-string failure, generator failure-to-start, switchgear issues, temporary power constraints and maintenance gaps can all affect continuity. Redundancy helps, but it does not make every dependency disappear.

On-site, hybrid and behind-the-meter generation add another layer. These models may help operators manage grid pressure or secure supply, but they also make the power risk map more complex than it used to be.

Why the power risk profile is changing now

The power risk profile is changing because the power model itself is changing. AI compute density is pushing load requirements up, grid constraints are making capacity harder to secure quickly, and the energy transition is changing how power is sourced, priced and contracted. Put those together and the old data center power assumption starts to look thin with the idea that grid supply, backup systems and property cover are enough.

Higher-density computing does not just need more power. It changes the pressure on cooling, electrical infrastructure, site selection and speed to market. The IEA projects that global data center electricity consumption will more than double to around 945 TWh by 2030, which shows the scale of the shift now sitting behind these decisions.

At the same time, grid certainty is harder to take for granted. Interconnection delays, local capacity limits and infrastructure pressure can push operators toward on-site, hybrid or behind-the-meter generation. That may help secure supply, but it also brings the operator into a more complicated power position: generation assets, fuel arrangements, equipment dependencies, PPA terms, credit exposure and more parties carrying pieces of the risk.

That is why coverage and contracts written for the old power profile may not fit cleanly. The issue is not only whether the site can keep the lights on. It is whether the insurance program, power agreements and capital plan have been tested against how power is actually being sourced and consumed now.

Where resilience investment stops and financial exposure begins

Most data center operators have already invested heavily in resilience, and they should have. N+1 or 2N design, UPS systems, generators, switchgear, fuel arrangements and maintenance protocols all reduce the chance of failure, help facilities stay online through stress events, and give tenants, investors and lenders some confidence that continuity has been engineered into the site.

But that investment does not decide what a failure costs when one still gets through.

A generator failure-to-start does not stay in the plant room. A UPS battery-string failure does not stay inside the UPS system. A fuel delivery disruption does not stay with the supplier. In a data center, those failures can move quickly into service credits, lost revenue, tenant obligations, emergency procurement costs, lender questions and reputational damage.

That is the exposure operators need to map back to their own profit and loss, because a power event can move from equipment failure to financial consequence very quickly.

The equipment may be where the event begins, but the balance sheet is often where the consequence lands. That is why the power risk conversation has to separate probability from cost. Engineering resilience lowers how often failure happens; risk transfer and capital planning deal with what the failure costs when it does. For data center owners, operators, developers, investors and lenders, both sides need to be tested together before a power event tests them in real time.

The insurance dimension of power risk

Power risk becomes an insurance placement question once the failure mode is clear. Property damage, business interruption and equipment breakdown cover may all be relevant, but each turns on different issues that includes what asset was affected, whether there is a physical damage trigger, what policy wording applies, and which part of the interruption is covered.

Modern data center power profiles make those questions less straightforward. Standard commercial property policies were not always written for on-site assets, hybrid or behind-the-meter generation, UPS dependencies, generator systems and temporary power arrangements. These are not small details to discover after an outage. They can decide whether the loss sits inside the program or outside it.

Power as a contract and capital risk

By the time power is committed, it is already part of the site’s economics. Price, duration, delivery obligations, connection milestones, counterparty performance and credit support can all shape what the project can promise to tenants, and how investors or lenders assess the project’s commercial potential.

That matters because data center power demand can change after the agreement is signed. Tenant requirements, AI workloads, equipment refresh cycles, energy pricing and grid constraints can all move across the life of the asset. A PPA that supports one phase of development may create a different risk profile as the site scales, refinances or brings new capacity online.

This is where PPA credit risk mitigation can become important. The issue is not only whether power has been secured, but who carries the exposure if the counterparty fails, power is delayed, demand assumptions shift or the economics of the agreement no longer match the project.

SLA exposure can compound that position. If uptime commitments are tied to customer contracts, a power-related outage can move quickly into service credits, revenue loss, liability and questions from lenders or investors.

Power therefore belongs in capital planning, not just facilities. Managing it means looking at operational resilience, insurance response, contract exposure and credit risk together, so the project’s power strategy is tested against the commitments and capital structure sitting behind it.

LEARN MORE

Build a more resilient data center risk strategy

Specialist guidance can help you map your exposures, pressure-test your current program, and identify strategies to strengthen protection across the data center lifecycle. Contact us to learn more more about how to better assess data center exposures, right-size coverage, and strengthen resilience from construction through operations.

Contact us

How Marsh helps you manage power risk

Marsh helps data center owners, operators, developers, investors and lenders test the exposure left behind by the power architecture, the insurance program and the contracts sitting around it.

The work is not to re-engineer the power system. It is to understand what the current power position leaves exposed. A power risk assessment and coverage adequacy review can help identify where interruption could occur, how loss could move through the site, and whether property, business interruption or equipment breakdown coverage has been tested against credible failure scenarios. It can also bring PPA credit risk mitigation into the same conversation, so power supply, counterparty exposure and financing assumptions are not reviewed in isolation.

Marsh’s integrated ARC model, which stands for assets, revenue and contracts, helps coordinate those interdependencies. For power risk, that means looking at the generating asset, tenant revenue, SLA terms, PPA credit exposure, insurance response and financing assumptions together, rather than treating engineering, insurance and contracts as separate exercises.

Marsh also brings together wider Marsh McLennan capability, including Marsh Risk placement, Guy Carpenter for reinsurance and capital solutions, Mercer for alternative capital sourcing, and Oliver Wyman for capital and strategy advisory. For data center clients, that means power risk can be assessed across operations, insurance and capital, rather than handled as a facilities issue after the major decisions have already been made.

FAQs

The main power risks facing data centers include grid failure, utility outage or undervoltage, capacity constraints, interconnection delays, UPS failure, generator failure-to-start, fuel delivery disruption, equipment breakdown and the added complexity of on-site, hybrid or behind-the-meter generation.

Power is often cited as a leading cause of significant data center outages, with UPS-related issues frequently involved. For operators, the important point is not only how often power failures occur, but what financial consequence remains when redundancy does not prevent an outage.

Standard property insurance may respond to certain power-related losses, but coverage depends on the cause of loss, policy wording, physical damage triggers, insured assets and resulting financial loss. More complex data center power models can raise coverage questions that should be reviewed before an outage occurs.

Assessing revenue exposure against a power-related outage requires more than engineering redundancy. Operators should assess business interruption exposure, SLA penalties, tenant impacts, emergency costs and whether existing insurance and contractual arrangements respond to realistic power failure scenarios.

Power purchase agreements can carry credit risk as well as operational exposure. Long-term commitments, financial guarantees, grid connection requirements, changing demand assumptions and counterparty obligations can all affect how a PPA sits within the wider capital structure.

Power redundancy is designed to reduce the probability of a power event causing an outage. Power risk protection looks at the financial consequence if that event still happens, including lost revenue, business interruption, SLA exposure, equipment breakdown and coverage response.

UPS or generator failure can cause outages when the backup system does not perform as expected during a power event. This may involve battery-string failure, generator failure-to-start, switchgear issues, maintenance gaps, fuel disruption or temporary power constraints.

Why Marsh

Marsh currently works with 75% of the top 25 companies engaged in data centers and are a partner to 10 of the largest global technology companies by revenue. Our dedicated digital infrastructure integrates our extensive experience with energy providers — including grid, nuclear, and renewables — and deep relationships with contractors, private equity investors, and technology suppliers into one seamless practice.

Speak with a Marsh data center specialist

If you are reviewing power architecture, PPA terms or redundancy assumptions, assess the potential financial impact before  a disruption occurs. Speak with a Marsh data center specialist about a power risk and coverage review, or download the report to see how power exposure moves across downtime, contracts, capital and the wider data center lifecycle.

Related insights

Man looking at a tablet in a server room

Report

Data center operational risks

09/22/2026
Placeholder Image

Article

The new age of contractual risk allocation in data center development

09/17/2026
Rows of data storage units are present in a large facility at night. The lights from the units create a blue glow, showing the technology in use. The setting highlights the scale of storage.

Report

Data center risk across the lifecycle

09/09/2026
Server racks in server room cloud data center. Datacenter hardware cluster. Backup, hosting, mainframe, mining, farm and computer rack with storage information. 3D rendering. 3D illustration

Report

Data center planning and construction risks

09/02/2026
View more

Download the report

Unlocking the Digital Infrastructure Opportunity

Digital infrastructure powers today’s economy and tomorrow’s innovations—but with rapid growth comes accelerated risks.

Marsh’s digital infrastructure risk report outlines core risk management tools that stakeholders can leverage which may help better manage risk across the digital infrastructure lifecycle.

Our report includes:
  • Comprehensive insights across the digital infrastructure ecosystem, offering an end-to-end perspective on risk
  • Novel risk management solutions that enable you to free up your capital
  • Resilience- building strategies to support long-term success
  • Core tools to help protect your investments and enable growth across every phase of the digital infrastructure lifecycle

Fill out the form to download your copy of the report.