In northeast Louisiana, a quiet stretch of land in Richland Parish is about to become the heart of the AI revolution. Once overlooked by auto manufacturers and industrial developers, the 1,200-acre Franklin Farm megasite will soon host a $10 billion Meta data center—the company’s largest in the world. Slated to begin operations in 2028, the site will require a staggering two gigawatts of power for computation alone.
But what fuels this massive leap in AI capability might not be green energy. To support the constant energy needs of Meta’s facility, utility company Entergy is proposing a $3.2 billion plan to build three large natural-gas power plants and modernize the grid. While Meta has promised to invest in 1.5 gigawatts of new solar projects, the immediate solution relies heavily on fossil fuels—a move criticized by environmental groups and U.S. lawmakers alike.
This tension highlights a growing national challenge. As AI infrastructure expands, so does its energy footprint. And increasingly, utilities across the United States are turning to natural gas as the most “reliable” option to meet the soaring electricity demand from AI data centers. Natural gas, which now accounts for about 40% of U.S. electricity generation, is cheap, abundant, and scalable. In Louisiana, it’s even more dominant, powering 72% of the state’s grid.
A Bridge Fuel That Became a Crutch
Natural gas was once hailed as a “bridge” to a cleaner energy future—a temporary solution as the country moved away from coal. And in many ways, it delivered. The U.S. power sector has significantly reduced its emissions over the last two decades, thanks in large part to the switch from coal to natural gas.
But the picture is now more complex. Natural gas still emits a significant amount of carbon dioxide. Methane leaks during extraction and transportation remain a major concern, as methane is a far more potent greenhouse gas than CO2. And with utilities now proposing more than 20 gigawatts of new natural-gas plants across southern states like Virginia, Georgia, and the Carolinas, the risk is that this “bridge” fuel could lock the U.S. into decades more of carbon emissions.
AI data centers are a major driver of this resurgence. In some states, data centers account for up to 85% of projected growth in electricity demand. And as these facilities run 24/7, with enormous computing and cooling needs, utilities argue that renewables and battery storage aren’t yet capable of meeting their requirements reliably or cost-effectively.
A Future Locked in Fossil Fuels?
Entergy, Meta’s utility partner in Louisiana, says the new gas plants are the only viable way to deliver constant power quickly. Alternatives like nuclear are too slow and expensive. Solar with batteries, while technically possible, is deemed prohibitively costly. As a result, the three gas plants are expected to operate for 30 years, extending Louisiana’s dependence on fossil fuels well into the 2050s.
U.S. Senator Sheldon Whitehouse recently criticized Meta’s plan in a public letter, citing the company’s “vague” offsetting measures like funding carbon capture and solar as insufficient. Advocacy groups argue that Entergy failed to conduct a rigorous analysis of clean alternatives, instead relying on Meta’s projections and treating around-the-clock demand as a fixed, non-negotiable requirement.
The concern is that these plants might not only accelerate emissions but also burden local residents with higher electricity bills, especially if AI demand fails to materialize as projected. Meta has signed a 15-year contract to purchase electricity from Entergy, but questions remain about what happens after that period ends—and who will pay for any underused capacity.
The Case for Flexibility
While AI is currently synonymous with high energy consumption, experts say it doesn’t have to be this way. Data centers, particularly those used for training large models, can be more flexible than assumed. By strategically reducing consumption during peak hours or shifting computation across regions, AI firms could help ease the strain on the grid.
A recent Duke University study suggests that if data centers curtailed usage by just half during high-stress periods, the U.S. grid could handle an additional 76 gigawatts of demand without building new plants. This would be more than enough to cover forecasted AI-related growth through 2029.
Projects are now underway to explore these solutions. The Electric Power Research Institute (EPRI), in collaboration with Meta, Google, and grid operators, is testing how real-time demand shifting could work. However, both utilities and data centers must learn to coordinate such efforts seamlessly to avoid disruptions.
Regulating the AI-Energy Nexus
Much of the oversight falls not to federal policymakers but to state utility commissions. In Louisiana, the Public Service Commission must now decide whether to approve Entergy’s gas plant proposal. Environmental groups like the Alliance for Affordable Energy argue that regulators need to demand better data and greater transparency, especially since public infrastructure is at stake.
There is also growing scrutiny of who pays for the new energy infrastructure. While Meta claims it covers its share of utility costs, critics fear that residential and small business customers could be left footing the bill if contracts falter or usage falls short.
In a recent paper, Harvard Law School researchers warned that utilities often mask the true cost impact of serving tech giants, making it “nearly impossible” for the public to assess the long-term financial risk.
AI, Power, and Responsibility
The explosion of AI has given Big Tech a new and powerful role in shaping America’s energy future. These companies have the money, influence, and demand to accelerate the clean energy transition—or delay it. They can be partners in grid modernization or contributors to long-term fossil fuel dependency.
The Meta data center in Louisiana may bring jobs and growth to a long-struggling region. But its legacy will depend on the energy decisions made today—and whether the promise of AI will be powered by innovation, or by the same old emissions-heavy status quo.
Prepared by Navruzakhon Burieva
