US Power Grid Under Stress from AI Boom: Data Centers Strain PJM Interconnection

The US power grid, managed by PJM Interconnection LLC, is facing stress due to the booming growth of AI data centers. The grid, which covers 13 states, no longer has a spare power supply, and tech companies must bring their own generating plants. PJM issued nine level 1 energy emergency alerts in the past five weeks, with two alerts attributed to unplanned power plant outages. The annual capacity auction resulted in a record $16.1 billion payout and a 208.7-megawatt shortfall. FERC Chairman Mark Christie warned that the reliability threat is not in the future but is present now.

The US power grid, managed by PJM Interconnection LLC, is facing unprecedented stress due to the booming growth of AI data centers. The grid, which covers 13 states, no longer has a spare power supply, and tech companies are being forced to bring their own generating plants to meet the demand. In the past five weeks, PJM issued nine level 1 energy emergency alerts, with two alerts attributed to unplanned power plant outages [1].

The situation is particularly concerning during the peak consumption season for US power suppliers. The first warning this year came in late June when extreme heat sent electricity demand soaring to a 14-year high. However, not all alerts have been due to high temperatures; two were attributed to unplanned power plant outages [1].

The annual capacity auction, held by PJM, further underscores the strain on the grid. The auction resulted in a record $16.1 billion payout and a 208.7-megawatt shortfall. This shortfall is significant as it marks the first time the auction did not meet its target, according to Monitoring Analytics, the PJM watchdog [1].

Federal Energy Regulatory Commission (FERC) Chairman Mark Christie warned that the reliability threat is not on the future horizon but is actually present now. "The reliability threat is not on the future horizon. It is actually here now," Christie said last week [1].

The growing demand for AI data centers is putting immense pressure on the power grid. As AI workloads continue to grow, the need for clean power is becoming increasingly urgent. The U.S. Department of Energy predicts that data centers could consume 12% of the country's total electricity by 2028, up from 4.4% in 2023. Much of this growth will come from AI-related processing [2].

In response to the challenges, Oracle has partnered with Bloom Energy to bring clean, reliable power to its AI data centers in the US using advanced fuel cell systems. These systems can generate on-site electricity in under 90 days, helping Oracle avoid grid limitations while reducing emissions [2].

The partnership is part of Oracle's broader strategy to reduce emissions. Oracle plans to reach net-zero emissions across Scope 1, 2, and 3 by 2050, with a 50% reduction by 2030 based on 2020 levels. By 2025, Oracle wants all of its operations, including all Oracle Cloud Infrastructure (OCI) data centers, to run entirely on renewable energy [2].

While the partnership with Bloom Energy is a step in the right direction, the broader issue of power grid strain remains. The rapid growth of AI data centers is likely to continue, putting further pressure on the grid. As such, a more comprehensive and coordinated effort from both tech companies and policymakers will be necessary to ensure the reliability and sustainability of the power grid.

References:
[1] https://www.bloomberg.com/news/newsletters/2025-07-29/the-boom-in-ai-is-stretching-the-largest-us-power-grid-to-capacity-this-summer
[2] https://carboncredits.com/oracle-orcl-stock-surges-due-to-ai-momentum-taps-bloom-energy-for-net-zero/