The AI Data Center Boom: Strategies for Sustainable Growth and Risk Management

Generative artificial intelligence (GenAI) services like ChatGPT are driving a new era of growth along the entire supporting infrastructure chain for these technologies. Since most of the AI workloads today live in the cloud, as the need for AI grows, so will the demand for cloud infrastructure. This is seen in the amount of data center capacity that cloud providers are planning to add.¹

This rising demand for data centers is also fueling a surge in data center construction, both in key markets and non-traditional locations around the world. Hyperscalers, large cloud computing service providers — including tech and IT giants — investing heavily in GenAI are leading the charge. Other industries driving data centers include entertainment, finance and banking, healthcare, retail and e-commerce. 

Sourcing power is a key priority for data center developers across all regions. However, land and power supply restraints, in addition to construction delays, can stand in their way.² In the near-term, developers and operators must also contend with the challenge of adhering to sustainability commitments, alongside managing many other external and project-specific risks while meeting the high energy demands of data centers. 

The early involvement of risk advisory partners is therefore crucial to manage complex risks through each phase of the data center project life cycle. It also supports cost-effective risk management and risk transfer programs.

The Promise of AI: Opportunities and Challenges

Meeting energy demands from data centers is encouraging developments aimed at boosting power grid efficiency and establishing systems that can handle and distribute renewable energy resources. 

“The advance of AI is a catalyst for problem solving that is necessary to increase grid efficiency,” says Liz Henderson, global head of Aon’s Climate Risk Advisory. 

Examples of advancements in data center development that focus on sustainability include the following:

• Researchers are designing specialized hardware, such as new accelerators, and new technologies, like 3D chips, that offer improved performance.
• New cooling technologies and sites that can perform more computations when power is cheaper, more available and more sustainable are also being explored to make energy use at data centers more efficient.³
• Data center operators are exploring alternative energy sources to power sites, including hydrogen and nuclear. In December 2024, Meta Platforms Inc. requested proposals from nuclear energy developers to provide 1 GW to 4 GW of new nuclear capacity in the U.S. to power its AI operations. Additionally, developers and operators are investing in emerging technology, such as carbon removal to pull CO2 out of the air and store it safely.⁴

While these developments are promising, the current data center landscape can be challenging to navigate considering the pressure on power and land availability, and the velocity of construction of buildings, phasing and project pipelines. 

The growth in data centers is a global trend, but the scale of development differs per region:

Key Risks in Design, Development and Construction

Power Availability 

Power is now a key consideration for data center operators conducting site selection. Operators have begun to canvas counties and utility providers across regions in search of large tracts of untapped, ready-to-go power, as well as markets with available power from renewable and sustainable sources, such as solar, wind, hydroelectric and micronuclear. Nonetheless, traditional power is still very much needed to power data centers. While developers are exploring how renewable energy can play an ever-increasing role, these projects remain closely tied to those power sources.

Many utility providers are suggesting wait times of several years for sizable power to be delivered to their developments. The challenge of available power has led operators to invest in new power technologies, such as battery storage facilities and geothermal power.  

Power supply challenges have been seen in data centers in North Virginia, for example. While transmission line construction has progressed, power availability and permitting processes have had a significant impact on data center construction timelines in the state. Meanwhile in Silicon Valley, several developers who purchased property to build a data center have been notified that they won’t receive utility power for over a decade, and must decide whether to wait, sell, reposition or pursue alternative power sources.¹⁰

Due to the increased use of GenAI, data centers are also growing in scale, with developers building power production, such as natural gas burning generation, at the site itself. On-site campuses with their own power needs are likewise being built to house permanent workers, which developers must consider in site selection, alongside other power-related considerations. For instance, in Malaysia, data centers must be built within a certain range of metro areas due to a decrease in power availability when located too far away. 

Land Availability 

As cloud demand from AI increases, operators are looking to build larger developments to satisfy hyperscale users. In addition to demanding more power, owners must find larger acreages to lock in a land price that protects exposure to speculative demand once a major market development is announced. This also helps operators control the phasing of the development over a period of several years. 

While larger acreages can provide the opportunity for substation and renewable energy development, securing large parcels of land in many metropolitan areas can be difficult. In Paris, for example, securing land for new data centers is affected by zoning restrictions implemented for the 2024 Olympics. On the other side of the world in Tokyo, limited land alongside power constraints have increased costs and delayed new developments.¹¹

Growing Use of Sustainable Materials

Data center owners are increasingly trying to use low-carbon materials to decarbonize both the centers and construction operations. This approach includes concrete that permanently traps carbon dioxide and steel, which is powered using renewable energy. Microsoft is now building its first data centers made with structural mass timber to slash the use of steel and concrete, which are among the most significant sources of carbon emissions.¹²     

Alternative cooling methods are also being considered to reduce energy use when a center is in operation. This includes central district cooling, where a handful of data center facilities within a district are cooled by one underground facility. The result is more efficient cooling for the entire district. 

Fire Risks

Fires in data centers are typically caused by a breakdown of machinery, plant or equipment. A fire that spreads quickly can result in significant financial losses and business interruption. 

While the structures for data centers often have concrete frames that are not significantly impacted by fires, it’s the high-value equipment that drives losses — from cooling technology to high-end computer servers or graphic card components. 

“The degree of damage that can occur stems from the actual fire, and direct and indirect losses due to heat. The smoke and soot also cause significant damage to electrical components. And then, water coming from extinguishing the fire can worsen the extent of damage,” says Clarence Ting, Aon’s global claims leader for Construction & Infrastructure. 

When these three components come together, the direct impact following an incident can cause a full replacement or close to full replacement loss scenario. If this type of fire damage occurs toward the end of a project, it often means a complete reset to day one. Even more, these scenarios usually require full replacement of equipment, in turn causing a start-up delay.

Many of the larger data centers are built in phases.  While expensive server equipment is not necessarily fitted within a building still under heavy construction, workers move around the site, and flammable material will sometimes be stored while finishing works are ongoing, adding further to fire risk. 

Pre-loss risk mitigation and related regulations are evolving to prevent associated losses. “Due to the high concentration of electrical components, and more combustible and flammable materials within data centers, there is an increasing focus on general risk management, separation of Li-ion batteries from the uninterrupted power supply systems and enhanced monitoring systems, which we expect to see grow in the coming year,” adds Ting.

The Broader Risk Horizon Facing the Data Center Pipeline

Extreme Heat in a Changing Climate

Extreme weather events can affect all life cycle stages of data centers, from construction to decommissioning. Data center developers account for this growing risk during site selection, as they look for structures that have adequate strength to survive earthquakes (if the building is going to be repurposed for a data center) and consider sites for new builds that are on elevated ground to avoid flooding.

Climate change has also introduced heat risk to projects, the impacts of which are far-reaching, from reduction in employee productivity, to increased potential for long-term disability and mental health stress. In response, regulations across Europe and the U.S. are emerging to protect worker safety, which is relevant for all construction projects, including data centers. 

“As infrastructure assets are being built, it is pertinent to consider impacts of extreme heat on workforce wellbeing, which in turn will account for potential disruptions and schedule delays to a project,” says Jon Chapman, a practice leader for Aon’s Construction & Infrastructure team in Europe, Middle East and Africa. “Contracts will generally provide for time relief under force majeure provisions, but are very unlikely to trigger a compensation event.” 

Data centers are cooled using water cooling, air cooling, refrigerants or combinations of these methods, depending on the data center size and location. Extreme heat events often lead to increased energy use for cooling, overheating and failure of equipment, reduced efficiency, and shutdowns or outages due to heat-related power disruptions. “Extreme heat can also damage the data center building infrastructure and water scarcity arising from prolonged drought can lead to decreases in cooling capacity as well as operational disruptions,” says Carol Stark, a managing director and North American renewable energy practice leader at Aon.

As a result, it’s important to consider future climate projections in data center design and planning to avoid surprises further down the road.

The Shifting Political and Regulatory Landscape

Energy sources to power AI data centers include fossil fuels and renewables. Therefore, changing policies impacting any one part of the energy sector could, in turn, impact the availability of sources, alongside the changing attitudes of communities toward data centers, which can be an impediment to their construction. Community opposition has mirrored similar movements against wind turbines and other developments, with specific concerns about the massive power and water demands that data center projects can impose on local infrastructure.¹³

In the U.S., the future of President Biden’s Executive Order 14008 and commitment to a carbon neutral economy, as well as the Inflation Reduction Act, remain in question.¹⁴ Potential tariffs on offshore wind projects could also be coming down the pike from President-Elect Donald Trump.

The size and quantity of data centers have attracted further scrutiny from governments, local citizens and environmental groups. Governments and utilities have restricted data center development in markets such as Dublin, Singapore, the Netherlands, Beijing and Shanghai.

“In some regions, data centers need to specifically comply with energy efficiency standards. This will require implementation in the design at an early stage or perhaps even a change to the design later on,” explains Vincent Banton, head of Construction & Infrastructure for Aon in Asia Pacific. “Both processes may extend the overall project timeline.” For example, in Singapore there is a green certification code for buildings, which would include data centers. 

“Developers need to consider timelines and upfront costs of complying with regulations ahead of a project breaking ground, particularly in more developed countries like Singapore, Japan or Korea,” adds Banton. 

In the U.S., China and the European Union, data centers account for around 2-4 percent of total electricity consumption, but because they tend to be spatially concentrated, their local impact is notable.¹⁵ Further growth of data centers could add more strain on local power networks due to the mismatch between rapid data center construction times and the slower pace of expanding and strengthening grids and generation capacity.¹⁶ In the UK, the building of new houses in West London was hampered by a lack of energy capacity, as power was consumed by nearby data centers. In 2023, Thames Water threatened to restrict the water services provided to data centers if they did not curb their consumption.¹⁷ 

FAST-Infra Label  and other groups are developing guidelines for sustainable performance of infrastructure assets, including energy efficiency of data centers.¹⁸ Similarly, best practice guidance on sustainable data center operations from the European Code of Conduct for Data Centers could have global relevance. Regulators, including the European Parliament, are beginning to establish requirements for systems to be designed with the capability of logging their energy consumption, ultimately aiming to improve transparency around energy usage.¹⁹

A Holistic Approach to Project Risk Management

Developers and owners can consider the following initial solutions for the various risk scenarios impacting a data center over the course of the project life cycle.

Innovation is Key to Navigate the Insurance Market

Due to the ever-developing property and casualty insurance markets, data center developers and operators must be well-prepared to address the complex risks their projects will face during all phases. This is crucial as claims continue to evolve for these projects and large losses relating to data centers increase in frequency. 

In this environment, risk managers can look to lessons learned from the power generation industry and claims trends stemming from losses. Developers and operators can also consider alternative risk options to supplement traditional insurance cover and access the capacity needed for projects. 

With parametric solutions, for example, insureds show financial loss to trigger a policy. “Even if they don't have physical damage to a data center asset, they may have suffered a financial loss because their access to local resources may now be scarcer,” says Hunter Nash, a business development leader for Aon’s Construction & Infrastructure team in the United States. “The parametric product provides much needed capital to subsidize lost income or increased costs from a local event that could impact their access to consistent power, water, workforce and more, which are crucial to carry on operation of the local assets. It is also a product that can creatively complement an existing insurance program, providing more confidence to the business operations.”  

Aon experts see claims payouts approved through parametric solutions consistently within weeks of the triggered event. This early cashflow can help a business expedite their recovery plans.

However, developing solutions for the current risk environment in which data centers operate is not enough, as innovation in this sector is occurring at a rapid pace. Moreover, taking a long-term view is critical when it comes to risk management and risk transfer for these projects, as the life span of data centers can reach several decades. Hyperscalers continue to support research into methods that can help generate power on-site, reduce the impact of capital costs and provide power away from large grid systems. 

“Entities that are leasing a data center over a period of 15 to 20 years understand that they have to help solve the energy problem to continue to grow data center capacity,” says Brian Hearst, a managing director in Aon’s Construction & Infrastructure team in the United States. “They have already begun pouring enormous amounts of money into research on micronuclear facilities to generate power on-site, shortening timelines for implementation from 10 years to now as early as five years. The insurance industry must get its arms around those developments from a risk transfer perspective.”

AI technology is also going to make managing complex data centers and the environments they operate in more efficient by reducing operational costs, optimizing energy usage, monitoring equipment and predicting potential issues before they happen, all while streamlining the construction phase. The companies building data centers are often the owners of the AI engines and plan to use that technology to make projects more efficient.

In turn, the insurance industry needs to innovate to help overcome sustainability challenges and ensure market capacity is available for these projects and the necessary solutions to get them up and running. 

“From a risk capital standpoint, we need to use creativity in underwriting practices to help data center clients support their green ambitions," notes Natalia Moudrak, global managing director for Aon’s Climate Risk Advisory. "Data center construction projects are becoming more complex and expensive. It will be important to continue to grow the insurance and reinsurance market capacity for green materials, such as mass timber, to match a growing demand.”

6 Strategies to Manage Data Center Risks Effectively 

Managing risks in data centers is crucial for ensuring smooth operation and sustainability. Here are key recommendations that can help mitigate potential risks and enhance the overall resilience of data center projects.

1. Start early. Engage your insurance risk advisor early in project development to get assistance in risk identification and mitigation strategies with the help of available data while mapping estimated cost projections.
2. Conduct pre-loss preparations. Nominating strong loss adjusting partners to ensure that the right experts are on board for construction projects is particularly important for data centers. Using loss simulations through the project cycle also helps to improve the level of preparedness of the project team when an incident, such as a fire, occurs.
3. Use a holistic approach. Work with a broker that has the team, knowledge, technology, global presence and market relationships necessary to take a total project life cycle approach to data center risk management, rather than breaking the project into three separate work streams — from construction to operations and decommissioning.
4. Understand stakeholder risks. Risk and insurance advisors can engage in contract negotiations with architects, contractors and lenders to allocate and address risks properly. They can also provide guidance around the business challenges of power and utility producers and independent power producers involved in projects. Because partnerships on these projects are complex and diverse, the division of risk needs to be clearly understood.
5. Look to non-traditional insurance solutions. In light of a changing climate, alternative risk transfer products like captives and parametric insurance are effective solutions for data center developers and owners to use and supplement traditional insurance coverages, particularly if the data center is in an area prone to weather perils.
6. Use climate risk analysis. This review will help developers access an informed view of how the changing climate will affect natural catastrophe-related risks that the data center is exposed to over the next 10 to 15 years. 

“Data center developers need to carefully consider the broader outlook for their projects,” says Brandon Schoper, project risk advisor and casualty broker for Aon’s Construction & Infrastructure team in the United States. “Aon’s value proposition means we can provide support across the entire project pipeline and program, and be strategic about helping projects move forward, leveraging not only our cross-solution expertise, but also our market relationships, innovative platforms and global reach.”

Aon’s Thought Leaders

Vincent Banton 
Head of Construction & Infrastructure, Asia Pacific 

Jon Chapman
Practice Leader, Construction & Infrastructure, Europe, Middle East and Africa 

Brian Hearst
Managing Director, Construction & Infrastructure, United States

Liz Henderson 
Global Head of Climate Risk Advisory

Natalia Moudrak
Managing Director, Climate Risk Advisory, Global 

Hunter Nash
Business Development Leader, Construction & Infrastructure, United States

Brandon Schoper
Project Risk Advisor and Casualty Broker, Construction & Infrastructure, United States

Carol Stark
Managing Director, North American Renewable Energy Practice Leader

Tariq Taherbhai
Global Chief Commercial Officer, Construction & Infrastructure

Clarence Ting
Global Claims Leader, Construction & Infrastructure

¹ 2024 Trends in Datacenter Services & Infrastructure, S&P Global Market Intelligence
² Global Data Center Trends 2024, CBRE
³ AI and energy: Will AI help reduce emissions or increase demand? Here's what to know, World Economic Forum 
⁴ AI and energy: Will AI help reduce emissions or increase demand? Here's what to know, World Economic Forum
⁵ Strategic Industries Surging: Driving US Power Demand, Grid Strategies
⁶ Global Data Center Trends 2024, CBRE
⁷ 2024 Trends in Datacenter Services & Infrastructure, S&P Global Market Intelligence
⁸ Global Data Center Trends 2024, CBRE
⁹ EMEA Data Center H1 2024 Update, Cushman & Wakefield
¹⁰ Global Data Center Trends 2024, CBRE
¹¹ Global Data Center Trends 2024, CBRE
¹² Microsoft builds first datacenters with wood to slash carbon emissions, Microsoft
¹³ Fighting back against data centers, one small town at a time, Washington Post
¹⁴ United States of America Construction Outlook, Q2 2024, Oxford Economics
¹⁵ What the data center and AI boom could mean for the energy sector, International Energy Agency
¹⁶ What the data center and AI boom could mean for the energy sector, International Energy Agency
¹⁷ Not just nimbys – why data center protests are on the rise, Raconteur
¹⁸ Label Development, FAST-Infra Label
¹⁹ AI and energy: Will AI help reduce emissions or increase demand? Here's what to know, World Economic Forum