Tools + Intel.

The rapid expansion of artificial intelligence is reshaping not only the digital economy but also the physical infrastructure that powers it. Hyperscale data center construction is accelerating globally, driven by AI workloads that require unprecedented computing density, energy consumption, and cooling capacity. Analysts estimate nearly $3 trillion will be spent globally on data centers by 2029, with projections approaching $7 trillion by 2030.

While data centers themselves are not new, AI-ready facilities represent a materially different property risk profile. Higher rack density, extreme power demand, advanced cooling systems, on-site energy generation, and geographic concentration are redefining underwriting considerations. For property carriers and brokers alike, understanding these evolving exposures is essential.

Global data center investment is projected to reach $3T by 2029 and could approach $7T by 2030.¹

THE HEAT PROBLEM: HIGHER RACK DENSITY = HIGHER PROPERTY EXPOSURE

AI workloads demand extraordinary computing power. Generative AI training clusters can consume seven to eight times more energy than traditional computing workloads.² As chip density increases, server racks are being configured with dramatically higher power loads to support Graphics Processing Units (GPUs) and advanced processing capabilities.

In fact, power density within server racks may increase by as much as 50 times compared to traditional configurations.³ This concentration of energy significantly amplifies property exposure. Higher rack density produces elevated heat loads, increasing:

• Fire risk from overheated equipment
• Mechanical strain on HVAC systems
• Equipment fatigue and premature failure
• Cascading losses when cooling systems fail

Lithium-ion batteries, increasingly integrated directly into Uninterruptible Power Supply (UPS) systems, add another dimension of concern. Thermal runaway events can escalate rapidly, generating extreme temperatures and flammable gases. While fires remain relatively infrequent, when they occur, the severity potential is significant.

Cooling failure in an AI-ready facility can force emergency shutdowns within minutes. Unlike traditional commercial properties, downtime tolerance is extremely limited. Property damage is often secondary to business interruption loss.

Global data center electricity demand is projected to more than double by 2030.⁴

POWER SUPPLY + UTILITY FAILURES AS LOSS DRIVERS

Reliable power is the foundation of data center resilience. Yet global electricity demand from data centers is expected to more than double by 2030, reaching approximately 945 TWh — roughly equivalent to Japan’s current total annual electricity consumption.⁴ AI-optimized facilities are projected to more than quadruple their electricity demand within that timeframe.⁴

This surge places enormous strain on aging grid infrastructure. According to the Uptime Institute, 45% of impactful data center outages are caused by power issues.⁵ Transformer failures, substation disruptions, and switchgear breakdowns are increasingly central underwriting concerns.

Operators are responding by deploying behind-the-meter power solutions, including gas turbines, battery storage systems, microgrids, and, in some cases, nuclear energy agreements.⁶ While this enhances resilience, it introduces new property exposures, including:

• On-site generation equipment breakdown
• Environmental liability exposures
• Construction and commissioning risks
• Fuel supply interruption

The scale of AI-related infrastructure investment is prompting industry observers to question whether insurance markets can keep pace with the growth in exposure.⁷ Large-scale utility outages also create systemic accumulation risk, particularly in concentrated data center hubs. From a property perspective, underwriting must extend beyond the building envelope to include electrical redundancy design, transformer procurement timelines, and backup generation adequacy.

By 2030, 52% of global data center hubs may face high or very high water stress.⁹

CAT CONCENTRATION: WHY LOCATION MATTERS MORE THAN EVER

Data center development is clustering in key US hubs including Northern Virginia, Dallas-Fort Worth, Phoenix, Chicago, and Atlanta.⁸ While clustering offers network advantages, it creates concentration risk for property insurers.

Many new campuses are also being built in regions exposed to severe convective storms, hail, wildfire, and flooding. The combination of geographic expansion and climate intensification heightens catastrophe exposure.

Wildfire presents both direct property damage risk and time element exposure. Smoke infiltration, power shutoffs, and evacuation zones can halt operations even without structural damage.

Water stress is another emerging concern. Large data centers can consume millions of gallons of water daily for cooling. Research indicates that 52% of global data center hubs may face high or very high water stress by 2030.⁹ Regulatory restrictions or community opposition tied to water consumption can materially impact operations.

Geographic underwriting now requires evaluation of:

• Floodplain positioning
• Wildfire defensibility and vegetation management
• Grid reliability in storm-prone regions
• Water availability and regulatory climate
• Regional accumulation exposure

Equipment breakdown is emerging as a leading severity driver in AI-ready data centers, with long transformer lead times and human error contributing to extended outages.^{5 10}

EQUIPMENT BREAKDOWN AS A PRIMARY SEVERITY DRIVER

AI-ready facilities incorporate high-performance computing equipment, advanced HVAC systems, redundant electrical feeds, and sophisticated fire suppression systems. As infrastructure complexity increases, equipment breakdown becomes a leading severity driver.

Transformer procurement lead times can exceed one year, extending potential downtime following failure.¹⁰ Custom-built components further complicate restoration timelines.

Human error remains a material contributor to outage events. Configuration and change management failures account for 50% of major network-related outages.⁵ Improper maintenance, installation defects, and system misconfigurations can trigger multi-million-dollar losses.

Equipment breakdown coverage and sublimits should be evaluated carefully in facilities reliant on high-density, high-value hardware.

DELAY IN STARTUP + TIME ELEMENT EXPOSURES

Construction costs for hyperscale facilities now routinely exceed hundreds of millions, and in some cases billions, of dollars.¹ Delays tied to transformer procurement, permitting constraints, labor shortages, or grid interconnection bottlenecks can materially impact financial outcomes.

Traditional property forms may not adequately address Delay in Startup (DSU) exposures for AI-driven facilities. Given the revenue expectations tied to AI compute capacity, even minor delays can produce significant financial consequences.

Similarly, contingent business interruption exposure is growing. A third-party substation failure or upstream utility disruption can impact thousands of customers simultaneously. Time element underwriting should consider:

• Revenue concentration within AI tenants
• Dependency on single utility providers
• Redundancy across campuses
• Contractual service level agreements
• Supply chain interdependencies

Hyperscale data centers now cost hundreds of millions, or even billions, making DSU and construction delays a critical financial exposure.¹

BOTTOM LINE

AI is not simply increasing data center demand; it is fundamentally altering the physical risk profile of these facilities. The property market remains disciplined, but underwriting conversations have shifted. The focus is no longer just on construction class and square footage. It is focused on resilience engineering, grid dependency, cooling redundancy, and systemic exposure. Brokers and agents working with AI-driven risks must ask deeper questions about infrastructure design, catastrophe positioning, and time element adequacy.

CRC Property specialists are actively working with carriers to navigate the evolving data center risk landscape, including exclusive solutions such as Insurisk’s Data Center product offering up to $500 million in combined property, including Builders Risk, and casualty capacity for these complex risks. If your client is developing, expanding, or retrofitting AI-ready infrastructure, connect with your CRC Specialty producer to review:

• Equipment breakdown adequacy
• DSU and time element exposures
• CAT concentration considerations
• Power redundancy and on-site generation risks
• Builder’s risk and operational transition coverage

AI-ready facilities demand AI-ready underwriting, and access to scalable, specialized capacity, so Team CRC can help you structure it correctly.

CONTRIBUTOR

• Daniela Mills is CRC’s National Property Practice Director.

END NOTES

1. The Cost of Compute: A $7 Trillion Race to Scale Data Centers, McKinsey & Company, April 2025.
2. Explained: Generative AI’s Environmental Impact, MIT News, January 2025.
3. Rising Power Density Disrupts AI Infrastructure, Goldman Sachs Research, May 2025.
4. Energy and AI, International Energy Agency (IEA), April 2025.
5. Global Data Center Survey 2025 and Annual Outage Analysis 2025, Uptime Institute, 2025.
6. Amazon, Google Sign Pledge to Support Tripling of Nuclear Energy Capacity by 2050, Reuters, March 2025.
7. AI Data Centers Are in Urgent Demand — Can Insurance Keep Up?, Insurance Business America, 2025.
8. Global Data Center Trends 2025, CBRE, June 2025.
9. Water Stress Index 2050 SSP585, Verisk Maplecroft, 2025.
10. AI Data Center Boom Has to Contend with Realities of Tough Labor Market, CNBC, September 2025.