AI data centers to add 44 million tons of CO2 by 2030

Data centers are fueling a fossil fuel comeback

Tech giants are installing massive fleets of natural gas turbines to power AI data centers, a move that could add 44 million tons of carbon dioxide to the atmosphere by 2030. This surge in emissions is equivalent to putting an additional 10 million combustion-engine cars on the road for a full year. The shift toward on-site fossil fuel generation marks a sharp departure from the industry’s long-standing promises to run exclusively on carbon-free energy.

A new report from the nonprofit organization Truthout highlights the environmental cost of the current artificial intelligence boom. Training and running large language models requires an unprecedented amount of electricity that existing power grids cannot provide. As a result, hyperscalers are turning to gas-fired turbines as the fastest way to bring new compute capacity online.

The scale of this energy demand is staggering. Deloitte Insights recently projected that the power required by U.S. data centers could grow by more than 30 times over the next decade. This rapid expansion is forcing companies to prioritize speed and reliability over their stated climate goals, leading to a massive reinvestment in carbon-intensive infrastructure.

Tech companies are repurposing jet engines for power

The demand for power generation hardware has become so intense that it has triggered a global shortage of standard industrial gas turbines. This scarcity is driving data center operators to source power from unconventional places, including the aerospace industry. Companies are now repurposing aircraft engines to generate electricity directly on-site at their server farms.

Boom Supersonic, the startup developing the Overture supersonic passenger jet, has pivoted to meet this demand. The company is now building power turbines based on its Symphony supersonic engine design. These engines are specifically modified to provide high-density power for the massive server racks required for AI training.

Crusoe, a specialized "neocloud" operator, is the first major customer for this aerospace-derived hardware. The company has ordered 29 turbines from Boom Supersonic to deploy at its data centers across the United States. These turbines allow Crusoe to bypass the years-long wait times for traditional grid connections by burning natural gas on-site.

• 29 turbines ordered by Crusoe from Boom Supersonic
• Symphony engine architecture repurposed for stationary power
• 1,000 gigawatts of gas-fired power currently in development worldwide
• 31 percent increase in global gas power projects in the last 12 months

Meta and Microsoft lean into natural gas

Meta is leading the charge in building massive, self-sustaining energy hubs for its AI operations. The company is currently developing its Hyperion campus in Louisiana, which is designed to reach a total capacity of 5 gigawatts. To power this site, Meta has commissioned the utility company Entergy to build three combined-cycle combustion turbine plants.

These three plants will supply 2.26 gigawatts of dedicated gas power to Meta’s servers. While combined-cycle turbines are more efficient than older gas plants, they still rely entirely on fossil fuels. This project represents one of the largest private investments in new gas-fired generation in the United States.

Microsoft is following a similar path, despite its public commitment to become carbon-negative by 2030. The company has stated that while renewables are the goal, natural gas remains the only viable near-term solution for the "always-on" power AI requires. Microsoft continues to use gas generators to bridge the gap in regions where wind and solar capacity are insufficient to meet the 24/7 demands of its Azure cloud.

Industry analysts at Gartner now advise data center operators that they cannot rely on the public grid to keep the lights on. Gartner recommends that all major facilities maintain their own on-site generating capacity to avoid brownouts. Schneider Electric, a major player in energy infrastructure, also points to gas turbines as the most practical solution for immediate energy needs.

Coal generation sees a policy-driven resurgence

Natural gas is not the only fossil fuel making a comeback to support the AI industry. Coal-fired generation increased by nearly 20 percent last year, driven largely by the spike in electricity demand from data centers. This trend is being accelerated by federal policy changes that prioritize domestic energy production over carbon reduction.

The Trump administration recently signed an Executive Order directing military installations and defense facilities to sign long-term power purchase agreements with coal-fired plants. This order aims to stabilize the grid as data centers consume an increasing share of available electricity. By locking in coal power for defense sites, the administration is freeing up other energy resources for the private tech sector.

The resurgence of coal represents a significant setback for the global energy transition. For the past decade, coal plants have been shutting down in favor of cheaper natural gas and renewables. However, the sheer volume of power required for AI has made every available megawatt valuable, regardless of the source.

The numbers don't lie: the tech industry is trading its green reputation for raw compute power. While companies talk about carbon credits and offsets, the physical reality on the ground involves burning more coal and gas than the industry has seen in years.

AI competition takes priority over climate goals

The shift back to fossil fuels is supported by a growing consensus in Washington that the "AI arms race" is a matter of national security. U.S. Interior Secretary Doug Burgum recently told a natural gas industry event that the primary existential threat to the world is America losing its lead in AI, not climate change. This perspective suggests that the government will continue to support fossil fuel expansion if it ensures U.S. tech dominance.

This policy environment makes it difficult for environmental organizations like Truthout to gain traction. When the government views AI development as a geopolitical necessity, environmental regulations often become secondary concerns. The result is a massive build-out of gas infrastructure that will likely operate for 30 to 40 years.

The global impact of this shift extends far beyond the United States. Projects totaling more than 1,000 gigawatts of new gas-fired power are currently in development worldwide. This represents a 31 percent jump in just one year, as other nations scramble to build the infrastructure necessary to compete in the AI market.

• 44 million tons of CO2 added by 2030 from new gas turbines
• 20 percent increase in coal-fired generation last year
• 5 gigawatts planned capacity for Meta’s Hyperion campus
• 30x increase in U.S. data center power demand by 2034

Tech companies are caught in a paradox. They are marketing AI as a tool to solve climate change through better efficiency and material science, yet the very process of building that AI is driving a fossil fuel renaissance. The industry is effectively burning the planet to build the engine that is supposed to save it.

As long as the demand for compute remains "inexhaustible," the reliance on natural gas and coal will likely continue. The speed at which gas turbines can be deployed makes them the default choice for a sector that cannot afford to wait for the grid to modernize. The 44 million tons of carbon projected by 2030 may only be the beginning if the AI boom continues at its current pace.