Key Takeaways:
Analysts from Bank of America predict that the U.S. will require over 230 GW of new power generation capacity in the next five years. However, regulated utilities are only anticipated to add around 93 GW, leaving a shortfall of more than 100 GW.
Data centers are projected to contribute about 125 GW to the U.S. electricity demand during this period, increasing overall electricity demand growth to a rate of 4.1% per year from 2026 to 2030.
- With large gas turbines mostly sold out until 2030, developers of data centers might resort to on-site gas engines while utilities extend coal plant operations and enhance battery storage and transmission systems.
Insights:
The rapid expansion in AI computing infrastructure, particularly through specialized chips and servers, is a primary factor behind this assessment from Bank of America. Analysts have noted that utilities have continually upped their demand forecasts over the last three years due to quicker-than-expected electricity demand linked to AI.
As the pace of new power capacity growth struggles to keep up, more data center developers are looking towards on-site generation solutions. More than 7.5 GW of these data center projects equipped with on-site generation are currently underway, with over 60 GW more in planning stages. These facilities are likely to utilize both self-generated power and traditional grid resources to enhance reliability and speed up project timelines.
The rise of AI infrastructure is altering long-term electricity needs following a decade of relatively stagnant growth in consumption, driven by efficiency measures like LED lighting and solar energy. Analysts caution that planned generation expansions might misrepresent actual supply as intermittent energy sources, such as wind and solar, may not deliver as much energy during peak demand as expected.
Natural gas will play a crucial role in satisfying this new demand. However, the limited availability of equipment poses a challenge. While large gas turbines remain the go-to choice for flexible energy, their production capacity is largely tied up until 2030, and new units can be slow to come online. This situation has sparked more interest in quicker-to-install natural gas reciprocating engines, with manufacturers like Caterpillar and Rolls-Royce ramping up production to meet soaring demand.
Utilities and regulators are also opting to keep existing power generation assets in operation for longer, which helps maintain reliability. The report highlights the postponement or cancellation of retirement plans for coal plants in several states, including Maryland and Wisconsin, to retain critical capacity.
To tackle reliability challenges, solutions such as battery storage, expanded transmission lines, and updated regulatory frameworks that make full use of existing facilities may prove effective. However, the analysts warn that transmission projects often take years to gain approval and be constructed, using the Champlain Hudson Power Express as an example of lengthy development timelines.
The surge in AI-centric data centers in the United States has led utilities and regulators to reassess how to accommodate large new energy loads and distribute the costs for new infrastructure. Although higher electricity prices could reduce demand from some customers, research indicates that electricity consumption tends to be relatively resistant to short-term price changes. For instance, a 10% rise in electricity prices could lead to a mere 1% to 2% drop in usage.
“The market is no longer limited by demand—it is limited by the capacity to deliver power,” noted Bank of America analysts in their Global Research Report.