· 12 min read
Introduction
Because AI is expected to enhance productivity and innovation, many people are optimistic about its future impact on economic growth in the U.S. Goldman Sachs thinks that this could begin to happen as soon as 2027 with more pronounced impacts possible over the long-term. To make this feasible, primarily technology and real estate companies will continue to make sizable investments in new data centers while government entities, utility companies, and regional grid operators will expand and adapt the electric grid. To give them more control over power, over time increasing numbers of data centers are expected to separate themselves from the grid by generating their own power.
Average hourly U.S. data center electricity demand
BloombergNEF forecasts that U.S. data center electricity demand will grow by ~206% between 2024 and 2035 while S&P Global Commodity Insights forecasts that total market electricity demand will increase by ~28%. Data centers’ higher growth rate of electricity demand compared to the total electricity market is expected to grow their proportion of electricity demand from 3.5% in 2024 to 8.6% in 2035.
Having reliable access to electricity is fundamental to the success of any data center. What may seem reliable access can become unreliable quickly in a world that is changing rapidly. As Roger Spitz observes, “in an unpredictable world change is slow, until it isn’t.”
In this environment, data center operators should try to ensure that they have access to resilient electricity. According to the OECD, “sustainable and resilient infrastructure is designed and built to withstand and recover from disasters and disruptions, such as extreme weather events or socioeconomic challenges.”
In this paper, I will focus on how more extreme weather events can threaten a data center’s source of electricity and lay out how to fortify resilience against them.
Increasingly volatile weather
The U.S. has been experiencing a growing number of large-scale natural disasters. The 1980–2024 annual average was 9.0 events while for the most recent five years (2020–2024) the annual average increased to 23.0 events.
U.S. billion-dollar disaster events 1980-2024 (CPI-adjusted)
Source: NOAA National Centers for Environmental Information (NCEI) U.S. Billion-Dollar Weather and Climate Disasters (2025), DOI: 10.25921/stkw-7w73
Since 2000, the number of major weather-related U.S. power outages has also grown. Between 2011 and 2021, the average annual number of weather-related power outages increased by roughly 78% compared to 2000-2010.
Major U.S. power outages
In the future, if companies can better protect electric equipment from the damage caused by large-scale storms, the number of major power outages might be reduced. Preparation includes weatherization of equipment such as construction of enclosure buildings, installation of insulation to protect from cold and heat, and waterproofing or elevating to protect from floods.
The degree to which preparation occurs depends to a large extent on resilience requirements in the market in which a utility operates as well as returns on investment. The resilience requirements, source of funding for resilience, and general impacts on resilience in the three types of electric markets are as follows:
Market type |
Resilience requirements |
Funding for resilience |
Observed results |
Fully Regulated |
State Public Utilities Commissions (PUCs) often require storm-hardening, vegetation management, weatherization, and long-term planning for climate impacts |
Costs added to rate base |
Outages still occur but duration often shorter than in deregulated markets |
Fully Restructured / Deregulated |
Transmission & Distribution (T&D) reliability standards enforced; generator weatherization may be voluntary unless mandated post-crisis |
T&D upgrades funded via regulated charges; generators fund from market revenues—no guaranteed recovery, so upgrades may be deferred |
Lower resilience to rare/extreme events; large-scale outages possible (e.g., 2021 Winter Storm Uri); quick restoration harder due to fragmented responsibilities |
Partially Restructured / Hybrid |
Generally more climate-specific rules than deregulated markets |
Utility-owned assets: costs recovered in rates; merchant assets: must comply with Independent System Operator (ISO)/state mandates without guaranteed recovery |
Better coordination than deregulated markets; mixed results — wildfire and heatwave blackouts still occur (e.g., California Public Safety Power Shutoff (PSPS) events), but planning tends to address seasonal risks proactively |
Source: ChatGPT
Because of regulations and mandates, all types of assets operating in fully regulated and partially restructured/hybrid markets are generally required to invest in weather resilience steps. In contrast, only transmission and distribution assets in fully restructured/deregulated markets tend to be required to do the same. As a result, generation assets in fully restructured/deregulated markets are often less resilient.
Without external sources of funds to finance resilience efforts, some owners of the latter assets have not seen the benefit of investing in resilience efforts. Costs of resilience efforts are added to most electricity rates in fully regulated markets, partially restructured/hybrid markets, and transmission and distribution assets in fully restructured/deregulated markets. However, they generally are not added to rates of generating assets in fully restructured/deregulated markets, nor do these assets often receive capacity payments.
In fully restructured/deregulated markets, businesses seek to purchase electricity at the lowest price with many preferring to hedge their prices so they can be protected from electricity rate spikes due to extreme weather or other events. Generators active in these markets compete against other generators and deliver electricity at prices set in advance or at spot. If extreme weather is unusual, it may be difficult for a generator to pass on the cost of any resilience efforts to customers. Many customers will not find it worthwhile to pay a premium to be protected from what they perceive to be a long shot when they can receive electricity from others at a relative discount.
The 2021 power market failure during severe cold weather from Winter Storm Uri in Texas’ principal grid (a restructured/deregulated market managed by the Electric Reliability Council of Texas (ERCOT)) and other grids in the South Central U.S. is a case in point of what can happen to a grid in a restructured/deregulated market when there is extreme weather. The power failure was triggered by above normal demand for electricity due to record-breaking low temperatures. Because many power generators in ERCOT were not incentivized to improve resilience in the years before 2021, some equipment was not able to weather the cold weather.
As a result, approximately 40% of the grid's capacity went offline, demand for electricity exceeded supply, and rolling blackouts ensued. Because the rolling blackouts started with industrial users, there was a vicious cycle. Industrial natural gas compressors could not produce natural gas used to produce electricity for various consumers and many consumers depending directly or indirectly on electricity produced using natural gas could not operate normally. Electricity companies that could still generate electricity, whether from natural gas or other generation types, likely profited as ERCOT left prices at maximum levels for a few days.
In their assessments of the causes and impacts of the power failure across Texas and other regions, the Federal Energy Regulatory Commission (FERC), North American Electric Reliability Corporation (NERC) and NERC’s regional entities determined that 44.2% of generating outages were due to freezing issues, of which 81% occurred at temperatures above the units’ stated ambient design temperature. Of note, according to these findings, a majority of the assets should not have failed simply due to the temperature. 31.4% was due to fuel issues, of which 87% were related to natural gas.
In the aftermath of this crisis, ERCOT took several actions aimed at improving grid reliability in the event of colder-than-normal weather, including:
• Inspections of generating and transmission assets for weatherization
• Proof of weather readiness from generation and transmission equipment owners
• Increasing operational reserves
• Requirements for some on-site fuel supply
• Unannounced testing of generation resources
However these efforts have not eliminated risks and according to the Federal Energy Regulatory Commission (FERC), much remains to be done.
Given that ERCOT accounts for approximately 12% of national generation, and Google GPT/Chat GPT estimate that 15-30% of U.S. generation comes from fully restructured/deregulated markets, there are a significant number of generating assets with resilience obstacles.
Can data centers, especially those located in fully restructured/deregulated markets, improve their resilience to extreme weather while maintaining or enhancing profits? Below are some threats that generators in these markets face from increasing extreme weather events and possible productive solutions:
Threat |
Solution |
Inaccurate weather forecasting of extreme weather events
|
Better understanding of the timing and magnitude of potential future weather conditions impacting relevant assets can improve strategic decisions:
• The European Centre for Medium-Range Weather Forecasts, a European forecaster, recently released an AI-powered weather model that forecasts many weather elements more accurately, faster, and using less power than other products. • Tomorrow.io uses proprietary AI models to develop up to the minute forecasts
|
Renewable generating assets’ reliability may become more volatile due to climate change |
Determine the most prudent type of generators to directly use or depend on:
|
Volatile electricity prices
|
By integrating extreme weather and the below drivers into analyses of electricity prices, decision makers can pursue more successful strategies:
|
Higher insurance prices
|
Owners can diminish the impact by choosing:
• Parametric insurance can cost less than traditional insurance and offer faster payouts by using event parameters like fire intensity or wind speed to determine payouts. However, payouts can be smaller. • Higher or aggregate deductibles (i.e. single deductible to all claims) • Partial self-insurance where owners assume a portion of the risk |
Debt funding limitations |
Data centers in these markets can address some of these concerns by:
|
Both the magnitude and timing of these threats are far from certain, making it harder for decision makers to choose which if any solutions to pursue.
To improve their chances of taking their organizations down a sensible path, it is imperative that decision makers not only apply traditional metrics like return on investment but also new metrics that capture increasingly important drivers of value. Those that the National Renewable Energy Laboratory (NREL) are good examples. Their resilience analysis tools and frameworks help determine the value of resilience over time such as economic losses within a community over an outage duration, or the avoided cost of lost utility revenue.
Conclusion
In order to maximize the chance of delivering solid returns to shareholders, decision makers should keep in mind how increasingly numerous and disruptive extreme weather conditions can affect performance. By taking appropriate actions, they can better manage risk and create value in an increasingly volatile world.
illuminem Voices is a democratic space presenting the thoughts and opinions of leading Sustainability & Energy writers, their opinions do not necessarily represent those of illuminem.
Track the real‑world impact behind the sustainability headlines. illuminem's Data Hub™ offers transparent performance data and climate targets of companies driving the transition.