Solmar Insights
US grid reliability is facing new pressure as AI data centers grow, with major incidents in Virginia highlighting risks from large-scale, rapid power disconnections. Leaders from Heron Power, NVIDIA, Invenergy, and Emerald AI are calling for three core requirements to ensure that facilities aid grid resilience rather than undermine it.
Key figures
3,100 MW of demand disconnected in Northern Virginia incident
60 data centers switched to backup in an earlier regional event
NERC highest-level alert issued in May 2026
Incidents drive regulatory response
On July 22, a routine transmission line fault in Northern Virginia caused 3,100 MW of demand to abruptly disconnect as data centers transferred to backup generation. This event, along with a similar incident involving 60 data centers two years prior, propagated voltage disturbances across the broader PJM footprint, reaching as far as Chicago. While bulk system reliability remained intact, the magnitude and speed of the disruption have alarmed regulators, underscoring the escalating risk as hyperscale data center loads proliferate.
In response, the North American Electric Reliability Corporation (NERC) issued its highest-level grid alert in May, flagging stress from large, sudden load swings as an urgent reliability threat. At the same time, the Federal Energy Regulatory Commission (FERC) expedited new directives for integrating large loads, targeting safe and orderly interconnection for energy-intensive AI infrastructure.
These developments follow mounting concern among local communities, faced with the prospect of higher energy bills and weaker service reliability as next-generation data centers ramp up across major markets including Virginia and Texas.
The three principles defined
Industry leaders representing power supply, computing hardware, grid interface technology, and AI orchestration have articulated a phased definition of “good citizenship” for data centers. The first requirement is to do no harm: facilities must not amplify grid instability, particularly during voltage events or power disturbances.
The second principle is to provide grid flexibility. Data centers are encouraged, when technically feasible, to adapt their operations and support grid balancing in periods of stress, such as responding dynamically during tight power supply conditions. The third and most ambitious requirement is to actively add to grid capabilities by helping bring additional generation, battery storage, and transmission online, collaborating with regional system operators to reinforce grid strength and flexibility.
Notably, the technology to meet all three obligations is described as commercially available now, with the four companies indicating their collective ability to deliver such an integrated data center solution.
How conventional data centers fall short
Traditional data centers often fail even the foundational obligation to do no harm. One key weakness is their so-called “ride-through” behavior: when a grid voltage drops briefly, most facilities automatically switch entirely to backup power, resulting in a multi-hundred or multi-thousand-megawatt load disappearing from the grid in seconds. The July 22 Virginia incident makes clear the cascading effect this can have on neighboring grid regions.
Additionally, workloads in modern AI data centers can fluctuate by tens of megawatts within seconds. These rapid power swings can further complicate local power quality and make system planning more difficult for grid operators such as PJM, ERCOT, and others overseeing high-demand corridors in the United States.
The combination of high sensitivity to local disturbances and volatile power usage has made data centers a central topic for both reliability engineers and energy market analysts concerned with maintaining balance and avoiding service interruptions as digital infrastructure demand grows.
Pathways to enhanced grid integration
According to leaders from Heron Power, NVIDIA, Invenergy, and Emerald AI, solutions to these reliability challenges are emerging from a combination of software and hardware advances. At the facility level, batteries can play a crucial role in ride-through, allowing centers to stay online during short disturbances without disconnecting, thus preventing a sudden loss of grid load.
Other promising strategies involve power electronics and smart orchestration software that can fine-tune computing loads in real time, responding to system operator signals and helping smooth aggregate demand. This supports both grid reliability and economic efficiency, as energy arbitrage and flexible loads become more valuable for balancing intermittent renewable resources.
By proactively contributing to new generation, storage, and local transmission development, AI data centers can go beyond simply avoiding harm and become central partners in building out U.S. grid capacity, an outcome seen as vital for both energy transition and continued digital growth.
What this means for buyers
Data center capacity and grid interconnection rights in PJM and Texas are directly impacted by new federal reliability directives. The July 22 event, where 3,100 MW of data center load disconnected, signals regulatory scrutiny and potential requirements for better ride-through and flexibility. Buyers should factor in evolving integration technologies and operational standards in due diligence for new data center acquisitions this quarter.
Reporting via the original publisher
