When a heat dome settled across the Eastern Seaboard this July, it revealed something the energy industry has been reckoning with: expanding AI data centers have changed what it means for the American grid to manage extreme weather.
With massive amounts of power coming online and continuous electricity demands increasing, energy executives and grid operators are realizing that AI doesn't just increase energy demand, it amplifies the grid's sensitivity to weather.
And, in regions of the grid where renewable generation is doing more of the work, forecast precision has become just as important as generation capacity.
Understanding this shift means looking not only at where AI infrastructure is being built, but at what it demands of the weather intelligence that supports the grid beneath it.
A study published this month by climate analytics firm First Street found that 79% of global data center capacity is located in regions that face high levels of risk from severe weather, including flooding, extreme winds and wildfires.
In addition to the straightforward challenges of increasingly severe weather, the locations of new data center infrastructure also introduce exposure to new climates and differing capacities for accurate weather forecasts.
For example, roughly 64% of data center capacity currently under construction in America has moved beyond traditional hubs like Northern Virginia and into what the industry is calling "frontier markets": West Texas, Tennessee, Wisconsin and Ohio. While these regions offer benefits like more affordable land, they also carry heightened exposure to tornadoes, severe storms and high winds and suffer from limited low-level weather observations or weather radar gaps.
For utilities that have long relied on weather intelligence to manage their operations, this summer's events haven't changed the form of the queries being asked. Instead, they've changed the stakes.
For example, two of the core questions facing grid operators have always been:
Now, with more AI data centers drawing power continuously, operators must apply these questions to facilities that can't easily or quickly dial back their consumption when the grid runs tight. Therefore, the weather data they're working off of becomes more than just operationally useful and turns into one of the few variables still available to act on.
For grids with diverse energy sources, wind and solar generation also shift directly with conditions, including temperature, cloud cover and wind patterns. That output can change significantly over the course of a few hours, such as during a heatwave, when solar efficiency can fall off just as cooling demand is spiking across the region.
As a result, the forecast for what weather will do over the next six hours immediately shapes the decisions being made in real time. And while these forecasts have always mattered to how utilities operate, as the U.S. grid becomes more complex the cost of overlooking them has grown accordingly.
The risk extends beyond peak demand moments. Heat waves, convective storms, wildfire smoke and flooding affect how power is generated, how it moves across transmission lines, and where it ultimately arrives. Data centers sit at the end of those lines, running through the same terrain that burns, floods, and loses power in severe weather. The industry has largely treated weather as an external variable. This summer is demonstrating that it's an operational one, on both ends of the wire.
NERC's most recent long-term reliability assessment flags AI data center load as one of the most significant near-term threats to grid stability. The forecasting challenge is at the heart of managing that risk.
When utilities and energy traders have visibility into what conditions will look like over the next several hours, they can respond to the forecast rather than the event, positioning resources before the heat peaks rather than after the grid is already running tight. When that precision isn't there, the margin for error is already gone before the event arrives.
The events of this summer also clarify something that has been building quietly for a few years: weather risk for the AI build-out is also a utility company's problem, not just a data center operator's problem.
The model that utilities like CenterPoint Energy have developed is instructive here. Rather than treating meteorological capability as a storm-season function, leading utilities are embedding weather intelligence across their operations: infrastructure planning, crew positioning, real-time grid management. This approach, built around the assumption that weather affects daily operations and not just emergency response, has become essential for managing a grid where AI load is a permanent feature of the baseline.
For grid planners and energy executives, the areas where data centers are being built fastest are often the same areas where the weather data is thinnest. That gap has always mattered — it matters considerably more now.
This summer's events were not anomalies. As AI data centers multiply across the country's fastest-growing markets, the quality of weather data has become one of the more consequential variables in how reliably the grid performs.
Focusing on more accurate long and short-term forecasts and treating weather intelligence as a daily operational input rather than a storm-season function are the practical steps that close the gap. You can't manage 21st century compute loads on 20th century weather infrastructure, and the AI build-out has made closing that gap more urgent than it has ever been.
Chris Goode is the Founder and CEO of Climavision, where he leads the only comprehensive supplemental radar network in the United States and a suite of high-resolution forecasting technologies that improve decision-making in energy and utility operations. With more than 30 years of weather industry experience, he has guided the company’s rapid expansion since its 2021 launch with backing from TPG’s The Rise Fund. An Air Force Weather veteran and former executive at The Weather Channel Companies, AirDat, and Enterprise Electronics Corporation, Chris brings deep operational expertise to advancing real-time weather intelligence for grid resilience, energy trading, and other weather-sensitive sectors.
