PJM Introduces Two Mechanisms to Address Soaring Power Demand from AI Data Centers
PJM Interconnection, the largest grid operator in the United States, recently announced plans to propose two new mechanisms to the Federal Energy Regulatory Commission (FERC) to address the rapid expansion of power demand driven by AI data centers. PJM estimates that by 2038, large-scale power demand within its service territory will increase by 70 GW, with the vast majority attributed to high-energy-consuming facilities such as AI data centers. However, since 2022, PJM’s grid has seen 15 GW of generation capacity retire, and the latest 2028/2029 capacity auction revealed a 6.8 GW shortfall, indicating that the growth in power supply capacity is failing to keep pace with rising demand.
The first mechanism proposed by PJM is the Reliability Backstop Procurement, designed to attract new generation resources to the market through a one-time procurement process and promote bilateral power purchase agreements to accelerate the filling of future supply gaps. The second mechanism, the Large Load Registry, will require new large-scale power consumers to register their demand. If the grid faces supply constraints and these consumers lack independent power sources or backup capacity, they may be prioritized for load reduction to maintain overall grid stability.
New Mechanisms Focus on Cost Allocation and Grid Reliability
PJM emphasized that these two new mechanisms align with the Ratepayer Protection principle, which has garnered support from the White House, governors of PJM member states, and some data center operators. Under this principle, new large-scale power consumers are expected to bear the costs of additional supply needs driven by their demand, preventing these expenses from being passed on to residential consumers and existing businesses. This approach aims to alleviate grid pressure while ensuring fairness and reliability in power supply.
Market observers note that AI data centers typically aim to complete construction and come online within 2 to 3 years, whereas the development of large power plants, transmission lines, and substations often takes 5 to 10 years or longer. This mismatch in timelines makes it difficult to meet the power demands of AI data centers through traditional grid expansion in the short term. In response to this challenge, tech giants such as Google, Amazon, Microsoft, Meta, and Oracle have begun investing in self-built energy systems in recent years, including solar energy, energy storage systems, natural gas turbines, fuel cells, and small modular reactors (SMRs). Oracle’s Project Jupiter even plans to use fuel cells as its primary power source to ensure stable electricity supply for its data centers.
PJM’s service territory covers 13 U.S. states and the District of Columbia, serving over 65 million people. Facing the surge in power demand from AI data centers, PJM stated that, until grid infrastructure expansion is completed, the market will continue to rely on transitional solutions such as natural gas power generation, energy storage systems, microgrids, and demand management to shorten the time required for data centers to secure power. However, the effectiveness of these new mechanisms in attracting generation resources and preventing cost-shifting to consumers remains subject to FERC approval and real-world market validation. Industry experts widely agree that balancing grid stability with the power needs of emerging industries will be a critical challenge for future energy policy.
PJM’s latest proposals reflect the unprecedented challenges facing the U.S. power market. As AI technology advances rapidly, the power demand from data centers continues to rise, while traditional grid expansion struggles to keep up. Against this backdrop, PJM hopes that through these new mechanisms, it can ensure grid reliability while requiring high-energy-consuming industries to bear the costs of their power needs, thereby reducing the burden on general consumers. However, the actual effectiveness of these measures remains uncertain, particularly regarding FERC approval and market acceptance.
