By Aiden Warrick
Finance Intern
Artificial intelligence (AI) is becoming one of the most important technologies in the modern age, but behind the promise of more intelligent systems and faster progress is a growing problem: AI needs massive amounts of electricity, which could push the United States power grid to its limits.
Communities around the country welcomed data centers. These large buildings filled with servers brought with them jobs, tax revenue, and a sense of progress. As AI has become more advanced, the amount of electricity required to run these data centers has rapidly increased. Now, many areas are starting to push back, worried that the power demand is too high.
A somber prognosis
According to McKinsey & Company, the electricity consumption of U.S. data centers is expected to increase by 9% annually, reaching 35 gigawatts by 2030. That means data centers alone could use 7.5% of the country’s electricity, up from the current 2.5%. A major reason for this surge is the advent of more powerful computer chips, like Nvidia’s new H100 GPU. Each chip draws 700 watts of constant power, nearly double that of its predecessor. A single rack filled with H100 GPU servers can consume as much electricity as dozens of American homes.
States and cities are being forced to reconsider due to this rapid growth. Community leaders have postponed or even rejected the approval of new facilities in Northern Virginia, which is home to the world’s largest collection of data centers. Communities in Georgia, Texas, and Iowa are starting to question whether the benefits of having these centers outweigh the risks of potential blackouts and rising energy costs.
The power sources in play
Currently, natural gas is the primary fuel powering most AI data centers. In 2023, it accounted for 43% of all U.S. electricity generation due to its reliability and relatively low cost. Renewable energy sources, such as wind and solar, are growing rapidly; however, their output depends on the weather. Nuclear power is dependable and carbon-free but expensive to build and slow to put into use.
Nuclear could play a bigger role in the years ahead. Under President Trump, new policies were introduced to speed up permit approvals and expand domestic uranium mining. The aim is to stabilize and grow the U.S. nuclear fuel supply. Some companies are already moving in that direction. Amazon, for example, recently bought a 960-megawatt data center campus next to the Susquehanna nuclear power plant in Pennsylvania. This is an indicator that big tech is looking for a consistent energy source.
However, experts say that nuclear power won’t scale fast enough to meet the growing energy needs of AI in the next decade. Most of the extra demand will likely continue to be met by natural gas. According to Tortoise Advisors, AI could require an additional 7 to 16 billion cubic feet of natural gas per day by 2030, which will put even more strain on pipelines and storage systems that are already under pressure.
Searching for solutions
Some researchers say that improved chip efficiency and better cooling technologies could slow the growth in energy demand. However, the trend is clear: next-generation chips are doubling the amount of electricity used inside data centers, which can consume up to 100 megawatts each.
This indicates a definite reality. Although AI has lots of potential, if a significant amount of infrastructure improvements are not made soon, it could strain the grid. The United States must find a balance between the potential of AI and the physical limitations of the power it consumes and determine whether the trade-offs are ultimately worth it as communities, utility companies, and policymakers look for solutions.
References
Kim, Tae. (2024). How AI is Sparking a Change in Power. Barron’s.
McKinsey & Company. (2024). AI Power: Expanding Data Center Capacity to Meet Growing Demand.
NVIDIA Corporation. (2022). NVIDIA H100 Tensor Core GPU: Technical overview.
Tortoise Capital. (2025). Energy Infrastructure is Powering the Future of Artificial Intelligence.
U.S. Energy Information Administration. (2024). Electricity explained: Electricity in the United States.
Miller, R. (2024). AWS eyes 960 MW for newly acquired nuclear-power data center in Pennsylvania. Data Center Frontier.
