The Virginia Incident, 2024

On 10 July 2024, in Loudoun County, Virginia, approximately 60 data centres collectively dropped from the grid within seconds. About 1,500 MW of load disappeared in response to a transmission disturbance. Not because the grid failed, but because the data centres did the prudent thing and disconnected to protect themselves.

The result was the opposite of prudent at system scale. Excess generation, no load to absorb it, and an immediate frequency excursion that PJM and Dominion Energy then spent hours managing.

The Virginia incident has become a touchstone for grid operators worldwide. It is the case study the AEMC's Improving the NEM Access Standards – Package 2 references when explaining why disturbance ride-through and recovery obligations for large inverter-based loads matter.

The structural lesson is clear. AI data centres are now system-security-significant loads. Their behaviour during disturbances has system-wide consequences. The choice of whether to ride through, when to disconnect, how quickly to recover, and how to coordinate with grid operators is no longer an internal engineering decision. It is a system-stability question.

The technical standards being developed in jurisdictions including Australia, Texas, Ireland, and Finland are the right response. The complementary question is how to verify that those standards are actually being met in real time, without depending solely on the operator's self-report.

The Virginia incident is the reason that question is no longer abstract. The next 1,500 MW disappearance is no longer a hypothetical. The framework that prepares for it requires both credible technical performance standards and credible evidence that those standards are being honoured.