The US data center boom is becoming an economic force, but its environmental cost is not evenly distributed. A new analysis published in Nature Communications argues that location will play a major role in how much carbon and water pressure the next wave of AI infrastructure creates.
The core finding is simple: data centers should not be treated as interchangeable buildings. Their impact depends on the electricity grid they plug into, the water available for cooling, and whether local resources can absorb rapid growth.
Why location now matters more
Tech companies have committed enormous sums to US infrastructure as the AI race accelerates. Meta chief Mark Zuckerberg told President Donald Trump last week that the company would spend $600 billion on US infrastructure, including data centers, by 2028. OpenAI has already committed to spending $1.4 trillion.
That scale changes the policy question. The issue is no longer only whether companies can build enough computing capacity. It is also where that capacity can be added without putting the heaviest strain on electricity grids, water systems, and climate goals.
The Nature Communications study uses data including demand for AI chips, state electricity information, and water scarcity to project possible environmental impacts through the end of the decade. It also models several scenarios for how data centers could affect the US and the planet.
Fengqi You, a professor in energy systems engineering at Cornell and one of the authors of the analysis, said the work began three years ago and arrives at “a perfect time to understand how AI is making an impact on climate systems and water usage and consumption.” He also said the AI industry “is growing much faster than we expected.”
The states the analysis favors
The analysis identifies Texas, Montana, Nebraska, and South Dakota as “optimal candidates for AI server installations.” These states stand out because they balance two major inputs: cleaner or improving electricity supply and lower water scarcity risk.
Electricity matters because data centers need large amounts of power. A state grid with more renewable energy, or one making progress toward cleaner energy, can reduce the emissions linked to running servers. Water matters because cooling is a major part of data center operation, and cooling also contributes to energy use.
That does not mean every favored state is already a mature data center market. Texas has built a large industry and is now the second-most-popular state for data centers in the country, according to Data Center Map. Data center construction in Texas quadrupled between 2023 and 2024.
Montana, Nebraska, and South Dakota are in a different position. They still have relatively few data centers, though You says their numbers are rising as the industry expands across the country. Nebraska has 39 data centers, including recently opened facilities owned by Meta and Google. South Dakota has just five facilities, the second-lowest in the country, according to Data Center Map, while an LA-based company is looking to build the first hyperscale data center there.
Why older hubs are under pressure
Much of the US buildout has historically centered on places such as Virginia and Northern California. These regions offered proximity to Washington, DC, and Silicon Valley, dense fiber connectivity, and skilled workforces. Virginia also offered substantial tax breaks for data centers for years.
Data Center Map, an industry tool that tracks data center development, says the US has more than 4,000 data centers. More than 650 are in Virginia, the most in the country, while California has more than 320, ranking third.
The problem is that successful hubs can become resource bottlenecks. Virginia does not suffer from water scarcity, but advocates have warned that data center power demand could interfere with the state’s goal to source 100 percent clean energy by 2045. California faces a different concern because its long-running water issues could become a bigger problem if data centers continue expanding there.
In October, Governor Gavin Newsom vetoed a bill that would have required data center operators to disclose water use, saying the state is “well positioned to support the development” of data centers. Arizona, which has more than 160 data centers, is named in the analysis as one of the states facing “severe water scarcity issues.”
The industry is still planning major projects in favored regions. Virginia has massive projects planned, including a $9 billion investment from Google announced in August. But You warned that continued construction in already-stressed places could exceed natural resource capacity.
The uncertainty around AI demand
Any forecast about AI infrastructure carries uncertainty. The study acknowledges that model efficiency, cooling technology, and changes in the mix of energy added to the grid could all change future water and electricity demand.
There is also the possibility that the AI bubble may burst, leaving half-built projects and contracts around the country. Power issues are already leaving some data centers sitting idle for years.
Still, the downside scenario is significant. If the grid does not move toward renewable energy while data centers keep expanding, emissions could rise sharply. In the most extreme scenarios, the analysis finds that US data center growth could generate up to an extra 44 million tons of carbon dioxide equivalent each year, more than entire countries like Hungary, Portugal, and New Zealand each generated in 2022.
Political choices also shape the outcome. Nebraska has massive potential for wind energy, one reason it is highlighted as a good data center location, but it has not yet built out that capacity. Nebraska’s utilities have instead invested heavily into more natural gas this year. The Trump administration has spent months putting its finger on the scale to promote fossil fuels to power the AI boom.
What the buildout should prioritize
The analysis points toward a practical framework for data center planning. Companies and policymakers should look beyond tax incentives, land availability, and existing tech clusters. They should also weigh whether a state can support growth without worsening emissions or water stress.
- Power source: data centers have lower carbon impact when the grid is cleaner or moving toward more renewable energy.
- Water scarcity: cooling needs make water availability a central siting issue.
- Resource capacity: popular hubs can become stressed if too much development concentrates in the same place.
- Future demand: AI growth could outpace efficiency gains, making siting choices more consequential.
The study does not claim that the environmental problem disappears if companies build in Texas, Montana, Nebraska, or South Dakota. It shows that some locations are better positioned than others if the US continues building data centers at speed.
The bigger lesson is that AI infrastructure is physical infrastructure. It competes for electricity, requires cooling, and depends on public decisions about energy systems. Where it gets built will help determine whether the AI boom deepens environmental strain or reduces avoidable pressure.