Powering AI: Why Data Centers Are a Board-Level Risk

Artificial intelligence is driving one of the fastest infrastructure expansions in modern history. As companies race to scale computing power, a more complex challenge is emerging: Data centers are no longer just an IT backbone. They are becoming a critical source of enterprise risk.

Once seen as back-office infrastructure, data centers sit at the heart of AI, which will become important to revenue, operations and customer experience. If a data center fails, the impact is immediate, from lost income and disrupted operations to reputational damage. For boards and senior executives, the issue is no longer technical. It is strategic.

The scale of the shift is significant. Bain & Company’s Sixth Annual Technology Report suggests that as much as $2 trillion in annual revenue will be needed to fund the computing power to meet AI demand by 2030. Global AI-related power consumption is expected to hit 200 gigawatts by the end of 2030, with the United States responsible for half of that energy draw.

As demand accelerates, so does the complexity — and the risk.

Global demands, local constraints

While AI demand fuels data center construction worldwide, the facilities have an outsized impact locally. Data centers require significant amounts of energy and water, potentially straining utility grids.

Access to ample power and water for cooling systems is driving the clustering of data centers, concentrating risk in specific geographies and increasing exposure to regional constraints. Virginia, Texas, California and Ohio account for around 40% of data center capacity in the United States, with 807 sites operational and many more planned, according to data intelligence firm Aterio.

For boards, this raises critical questions around concentration risk. Geographic clustering can expose organizations to regional disruptions — whether from extreme weather, grid instability or regulatory constraints tied to energy and water usage.

Building at speed, managing complexity

The race to expand capacity is also introducing new forms of execution risk. Data centers are capital-intensive, technically complex projects, often delivered under compressed timelines.

Organizations face a combination of supply chain bottlenecks, long lead times for specialized equipment and shortages of skilled labor. At the same time, design and integration risks, particularly around power, cooling and redundancy systems, are becoming more pronounced as facilities scale in size and density.

Delays, cost overruns and stranded capital are risks in many types of construction projects, but these risks are particularly acute in high-value data centers.

Operational resilience in high-density environments

Once operational, data centers introduce a different set of risks. High-density computing environments generate large amounts of heat and depend on uninterruptible power sources and sophisticated cooling systems. This creates tightly coupled systems where small failures can cascade quickly.

As facilities become more connected and automated, reliance on digital monitoring and control systems is also expanding the cyber threat. Extreme weather and site-specific vulnerabilities similarly add exposure. In colocation environments, where multiple tenants share infrastructure, an outage can quickly amplify the economic impact of business interruption.

For organizations relying on third-party providers, these risks are often less visible, but no less significant. Dependencies on external infrastructure can create blind spots in resilience planning.

A board-level risk agenda

As data centers increasingly become strategic assets, boards need to approach them as a core element of enterprise risk, not just a technical function

That starts with asking the right questions:

· Where are our concentration risks across geographies, providers and infrastructure?
· What is the true cost of downtime across our operations and value chain?
· How resilient are our systems to failures in power, cooling or network connectivity?
· Do we fully understand our dependencies on third-party infrastructure?

The rapid expansion of AI infrastructure presents significant opportunities, but also introduces new and evolving risks that require attention. For boards, this means treating data centers as a core enterprise risk, understanding risk accumulation and stress-testing resilience across both owned and third-party infrastructure.

Scott Gunter is Chief Executive Officer of AXA XL, AXA’s property, casualty and specialty insurance and reinsurance division, and a member of AXA’s management committee.