Since the turn of the 21st century, annual load growth in the U.S. has slowed to an average of 0.5%, ushering in an era of power demand stagnation. However, with the recent advancement of AI and its associated data center rollout, the U.S. stands on an unprecedented precipice with the possibility of massive load increases in only a few years to meet the growing AI demand. This load increase will have wide-ranging effects on the U.S. energy ecosystem, most notably with a push for more natural gas-fired generation and possibly the pushback of retirements for previously uneconomic coal plants. Check out our more detailed analysis available to FactSet users, and be sure to register for the FactSet Energy team’s upcoming webinar diving into how the rapid expansion of AI data centers is driving electricity load growth and reshaping energy landscapes across the United States.
Historically, U.S. electricity load growth has remained relatively steady, with the Lower 48 experiencing an average annual increase of just 0.8% since 2016. However, utilities have been consistently overestimating electricity demand over this same time period, evidenced by forecast errors averaging between 3% and 10%. This tendency is not without cause though, as it allows utilities to maintain adequate supply during peak periods and can support increased investment in the grid. On a regional level, ERCOT accounts for 44% of the total U.S. load growth seen since 2016, driven by rapid population growth and booming oil and gas production. On the other hand, CAISO, NYISO, and ISONE all show decreasing load for the past decade due to a combination of higher power prices, stable or shrinking populations, and ambitious energy efficiency efforts.
Even though we’ve seen load overestimations in the past, the explosion of AI in recent years has provided strong reasoning to suspect that load will grow quickly over the next five years. A range of data center forecasts have been developed to address the uncertainty surrounding the future demand. Estimates for 2030 span from a low scenario of about 65 GW of data center capacity to a high scenario exceeding 100 GW. Even at the lower end, such growth would significantly increase electricity demand, and the higher end could present major challenges for utilities nationwide. This anticipated rise in load is likely to lead to reserve margins tightening in key ISOs, like PJM and ERCOT, wherein utilities may face difficulties bringing enough new generation online fast enough to keep pace with demand.
As reserve margins tighten, reliability concerns for grid operators will likely intensify, prompting an increased reliance on natural gas-fired generation to meet rising demand from data centers. Through 2027, natural gas consumption is expected to remain steady as increased data center demand is offset by rising solar generation. But, after 2027, gas burn is expected to see increases by at least 2 Bcf/d under both the Mid and High data center growth scenarios. This rise is not only due to data center demand increasing, but also the expiration of renewable tax credits enacted on the One Big Beautiful Bill Act (OBBBA). With the need for increased gas burn, pipeline bottlenecks may begin to become more apparent, leading to potential price spikes in electricity prices in some regions.
The surge in both AI-driven demand and the associated rollout of data centers are driving an unprecedented era of load growth in which utilities may be challenged to meet and maintain adequate reserve margins. Based on the data center capacity increases projected under the Mid and High scenarios over the next five years, large increases to gas burn are expected. Combined with ongoing pipeline constraints, these pressures could contribute to higher electricity prices and may even prompt a renewed reliance on coal to offset the resulting high gas prices. Ultimately, these elevated energy costs could serve as a moderating force, potentially even slowing the pace of data center development. Whether that will be a strong enough deterrent to the rapid expansion of data centers remains to be seen.
Be sure to register for the FactSet Energy team’s upcoming webinar diving into how this rapid expansion of AI data centers could vastly reshape energy landscapes across the United States. If you’re already a FactSet user, check out the full report on data center load growth and its effects here.
This blog post is for informational purposes only. The information contained in this blog post is not legal, tax, or investment advice. FactSet does not endorse or recommend any investments and assumes no liability for any consequence relating directly or indirectly to any action or inaction taken based on the information contained in this article.