Texas is booming as a host for data centers used to power AI’s growing infrastructure. With over 400 active centers, Texas already accounts for about 15% of the nation’s total energy use, and the state’s industry demands are on the rise. And it will soon be home to the largest data center campus in the country, which will be permitted to emit up to 12,000 tons of air pollutants and 33 million tons of greenhouse gas emissions a year from its natural gas plants.
Many Texans are concerned about the impacts the growing industry will have on air quality, water consumption, land usage, and grid stress. Dr. Vijay Gadepally, an MIT scientist and speaker on South by Southwest’s The Climate Paradox of AI Education panel, said they should be.
A single ChatGPT prompt requires 10 times the power of a classic Google search, and just one AI server uses 15 kW of power, enough energy to power a few homes, Gadepally said. Power isn’t the only issue. “A single ChatGPT conversation uses one bottle of water just from cooling,” he added. Estimates conclude that Texas data centers consumed around 25 billion gallons of water in 2025, which could increase to upward of 160 billion gallons by 2030.
“It’s no secret that AI requires a lot of energy and a lot of water,” said Anurag Bajpayee, a water entrepreneur and speaker on South by Southwest’s Dirty Data: The Hidden Climate Cost of Our Digital Lives panel. But regardless of the strain that the developments have on natural resources, Bajpayee says “it’s here to stay.”
His water solutions company, Gradiant, is working to implement recycling and treatment strategies to mitigate industrial impacts and waste through new technology. A combination approach of chemical solutions, recycling technology, and monitoring systems allows Gradiant to recycle up to 99% of a facility’s water, Bajpayee said. Just last year, the company signed commercial operations contracts with two data center companies, and have a third on the way – and he’s confident of the impact that can have on the industry.
“We can touch so many data centers just with these three companies,” he said. Recycling water can scale down impacts by magnitudes of hundreds. “If we can do that in every single place, the water problem with AI will be mitigated.”
Some recent legislation gives experts hope for the role regulation may play. Senate Bill 6 requires data centers to increase energy transparency and backup power generation, as well as emergency energy shutoffs for centers during grid emergencies, like 2021’s Winter Storm Uri, which left more than 4.5 million Texans without power. “This is a good example of where policymakers can help design laws that enable and incentivize flexibility,” Gadepally said.
“The faster we can balance this equation in a sustainable way, the better it is,” he said. While Gadepally knows AI is the root of the problem, he also thinks it may be what helps us solve it. “AI is creating a problem on one end, but it may also be the thing that helps us get past it – That’s the paradox,” he said.
In his companies and labs, Gadepally implements the concept of “environmentally aware computing” by incorporating external feedback through AI software systems to optimize system efficiency, identify systemic weaknesses in buildings, and make facilities more energy resilient through automated power adjustments. Meanwhile, Gradiant’s AI monitoring system is allowing them to track usage efficiency across companies to reduce water footprints. At the global scale, AI is already playing a large role in tackling climate solutions, as scientists can now track and analyze environmental data faster than ever before.
The Data-Driven EnviroLab out of UNC-Chapel Hill is using AI to analyze large data to influence and track climate policy, said founder Dr. Angel Hsu, who will join Gadepally on the Climate Paradox of AI Education panel. Currently, the lab is working on their ChatNetZero approach, an AI assistant program designed to make climate pledges and status more comprehensible.
AI is especially useful in large-scale data analysis thanks to its ability to manage and assess large datasets very quickly, Hsu said. Their work aims to “evaluate, demonstrate, and translate innovations [including AI] into evidence and tools that policymakers and practitioners can use responsibly,” she said. They are also investigating “how large tech companies account for the growing energy and water footprint of AI itself, as well as how user behavior impacts the overall energy consumption of AI,” outlined in the organization’s recent working paper.
Hsu said she is “cautiously optimistic” about AI’s impact, noting the beneficial role of analysis but also warning of the challenges that it poses, such as amplifying “existing biases and blind spots in the underlying data,” making “inequities worse while giving the illusion of objectivity.” Still, she believes sustainable AI systems that have more environmentally friendly underlying infrastructure, clear guidelines for use protocol, and active monitoring that can help mitigate impacts.
Experts believe that system flexibility, reducing energy peaks, and rethinking designs are key ways to mitigate current environmental loads. Bajpayee said it comes down to how much we care about human society, and the work needed now to safeguard the future.
“The planet will be here long after we are gone,” he said. But if the AI industry continues to treat water, energy, and land as a mere resource to be used and then thrown away, “we’re screwed.”
The Climate Paradox of AI Education
Cities & climate track
Thursday 12, 11:30am, Austin Marriott Downtown, Waller Ballroom C
Dirty Data: The Hidden Climate Cost of Our Digital Lives
cities & climate track
Tuesday 17, 2:30pm, JW Marriott, Room 201-202
This article appears in SXSW 2026 Festival Guide.