Texas' rapidly growing data center industry could erase much of the state's progress on air quality by the end of the decade, according to a new report from the Houston Advanced Research Center covered by The Texas Tribune. The findings land as the state juggles the largest AI infrastructure build-out in the country with a governor-ordered audit of the industry's demands on the power grid.
The report, released Tuesday, models how different power choices for data centers produce wildly different emissions futures. The outlook is far worse if facilities rely on natural gas plants built on site rather than connecting to the grid and adding renewable generation.
Two Decades of Progress, One Decade to Unwind
Between 2002 and 2024, Texas sharply cut its major air pollutants — driven by transportation, industry and power generation — even as the state's population grew by millions and its economy roughly doubled, according to the report, which drew on federal and state emissions data.
The projected reversal comes in two flavors. By 2031, statewide nitrogen oxide emissions could rise 35% even under a cleaner scenario in which data centers connect to the grid and deploy wind and solar on site. Under a scenario where facilities instead choose traditional on-site natural gas power, those emissions could climb nearly 400%.
Nitrogen oxides are not a cosmetic pollutant. Long-term exposure can damage the human respiratory system and contribute to chronic lung disease, which is why the report's authors frame the trajectory as a public health question, not just an environmental one.
"What we are showcasing here is that there is a significant increase in different pollutant emissions that needs to be addressed because otherwise this could have strong implications for the population and environment," said Carlos Gamarra, vice president for energy, air and IT at HARC, whose team produced the air quality scenarios.
The Scale of the Build-Out
The stakes are unusually high in Texas. At least 288 data center projects are planned across the state, according to a recent Texas Tribune analysis. Proposed facilities range from five megawatts to 11,000 megawatts of planned capacity, and the estimated electricity that large projects could draw from the grid adds up to 439 gigawatts — roughly a third of all power generation in the United States.
The grid operator, the Electric Reliability Council of Texas, has received more than 500 requests to connect large electricity users over the last two years, many of them data centers, compared with roughly two dozen such requests in the year before that, according to state records cited by the Tribune.
That queue is exactly why the emissions scenarios diverge so sharply. Waiting for a grid connection is slow, and developers that cannot wait look for alternatives — which today usually means on-site natural gas generation, the option the report identifies as the worst case for air quality.
Gov. Greg Abbott has temporarily halted new data centers from connecting to the state's power grid pending an audit examining each facility's power demand, water and cooling use, public financial assistance and community impacts. The pause adds another incentive for projects to weigh self-supplied power, raising the stakes of the report's modeling.
Treat Them Like Industrial Facilities
Gamarra's central argument is regulatory: data centers should be held to the same emissions standards as the industrial facilities they effectively are.
"We need to treat data centers as the industrial facilities they are … to hold them accountable to the same industrial and power generation emissions standards that produced Texas progress," he said.
The report's researchers argue the outcome is not locked in. Instead of defaulting to the cheapest and fastest power options, state lawmakers and industry leaders could shape the trajectory by pushing cleaner choices such as grid connections paired with renewables, and by implementing regulation that internalizes the emissions cost of on-site fossil generation.
The physics pushes in the other direction. Data centers need electricity for servers and, critically, for cooling — and some of the most water-efficient cooling technologies consume the most electricity. As AI chips grow more powerful and run at hotter temperatures, cooling demand rises, and with it energy use.
A Test Case for the National AI Build-Out
Texas is not the only state confronting this trade-off, but it is the largest and fastest-moving, which makes the HARC report a template for fights brewing elsewhere. The same question — whether the AI boom's power demand is met with renewables, grid upgrades and storage, or with quick-build gas turbines — is playing out in statehouses and county commissions across the country.
For continuous coverage of the energy side of the AI boom, AI Buzz Wire tracks policy, infrastructure and industry news as it develops. The Texas report suggests the window for shaping that outcome is measured in years, not decades: the emissions divergence in the modeling opens up by 2031, within the typical build timeline of a single large data center campus.
The report's authors put the choice plainly. The industry's environmental footprint is not fixed by the technology itself but by procurement and regulation decisions being made right now, county by county and interconnection request by interconnection request.
Stay Ahead of AI
Follow the intersection of AI, energy and policy with AI Buzz Wire's independent coverage of the latest AI developments.
Read more AI news →