A WIRED investigation has quantified a stark environmental cost of the AI boom: 11 natural gas projects powering US data centers could emit more greenhouse gases annually than the entire country of Morocco emitted in 2024. The analysis, based on air permit documents, estimates these behind-the-meter gas projects have the potential to release over 129 million tons of CO2 equivalent per year.
These projects are being developed to serve some of the most powerful AI companies — OpenAI, Meta, Microsoft, and xAI — as demand for compute skyrockets. The infrastructure is designed to bypass the electrical grid entirely, a trend known as behind-the-meter power, as data center developers face multi-year waits for grid interconnection and rising public resistance to higher energy bills.
Key Takeaways
- WIRED investigation reveals 11 behind-the-meter natural gas projects for AI data centers could emit 129 million tons of greenhouse gases annually, surpassing Morocco's 2024 emissions.
- Projects tied to OpenAI, Meta, Microsoft, and xAI bypass traditional grids.
The Scale of the Problem
Michael Thomas, founder of clean energy research firm Cleanview, who has been tracking gas permits for data centers nationwide, described behind-the-meter power as "a crazy acceleration of emissions."
"It's almost like we thought we were on the downside of the Industrial Revolution, retiring coal and gas, and now we have a new hump where we're going to rise," Thomas told WIRED.
One of the most prominent examples is xAI's Colossus campuses. Air permit applications for the Colossus 1 campus in Memphis, Tennessee, and the Colossus 2 campus in Southaven, Mississippi, show gas turbines at each site could generate more than 6.4 million tons of CO2 equivalents per year. Combined, that is roughly equivalent to the emissions from more than 30 average-size natural gas plants, or enough energy to power 1.5 million homes.
These campuses have already drawn community opposition. Residents in the low-income Black community around the Memphis campus protested the turbines. The EPA ultimately approved their use in 2024. Last month, regulators granted a permit for an xAI affiliate in Southaven despite widespread opposition. The NAACP filed suit against xAI last week, claiming the company was illegally operating the turbines.
Behind-the-Meter: A Growing Trend
The WIRED investigation identified 11 gas projects tied to data center campuses, but the scope may be larger. The analysis covers only projects that have submitted air permit applications, meaning many more could be in earlier stages of planning.
Behind-the-meter power allows data centers to avoid grid interconnection queues that can stretch 5-7 years. It also insulates them from rising electricity costs and public scrutiny over grid impacts. However, these projects largely bypass environmental review processes that would apply to new power plants serving the general grid.
The Broader Context
This is not just a US issue. Globally, AI's energy demand is projected to grow dramatically. The International Energy Agency has warned that data center electricity consumption could double by 2026, with AI workloads being a primary driver.
The emissions from these 11 projects alone — 129 million tons annually — would place them among the top 50 national emitters globally, comparable to countries like Morocco, which emitted approximately 130 million tons in 2024.
Company Responses
Microsoft, Meta, and OpenAI have all made public commitments to carbon reduction and renewable energy. Microsoft has pledged to be carbon-negative by 2030. Meta has committed to net-zero emissions across its value chain by 2030. However, the rapid buildout of AI infrastructure appears to conflict with these goals.
WIRED reports that the companies did not respond to requests for comment on the specific projects identified in the analysis.
What This Means in Practice
For AI practitioners and infrastructure planners, this investigation highlights a fundamental tension: the compute demands of frontier AI models are outstripping the ability of clean energy infrastructure to scale. Until renewable generation, grid capacity, and energy storage can catch up, behind-the-meter gas will remain an attractive option for companies needing power quickly. This creates a direct trade-off between AI capability advancement and climate commitments.
Frequently Asked Questions
How much greenhouse gas do AI data centers emit?
According to a WIRED investigation, 11 planned gas-powered data center campuses in the US could emit over 129 million tons of CO2 equivalent annually. That is more than the entire country of Morocco emitted in 2024.
What is behind-the-meter power for data centers?
Behind-the-meter power refers to electricity generation that is built on-site and directly powers a facility, bypassing the public electrical grid. Data center developers are increasingly using this approach to avoid multi-year grid interconnection delays and rising energy costs.
Which AI companies are building these gas-powered data centers?
Projects tied to OpenAI, Meta, Microsoft, and xAI have been identified in air permit documents reviewed by WIRED. xAI's Colossus campuses in Memphis and Southaven are among the most prominent examples.
Can AI data centers run on renewable energy?
Yes, many companies have pledged to use renewable energy. However, the rapid pace of AI infrastructure buildout is outstripping the availability of clean power. Behind-the-meter gas offers faster deployment than waiting for new renewable projects and grid upgrades to come online.
