Nvidia, Invenergy Propose Principles for AI Data Center Grid

Officials from Heron Power, NVIDIA, Invenergy, and Emerald AI have published a set of three principles for integrating AI data centers into the electric grid. The proposal, published on September 1, 2026, in Utility Dive, responds to incidents where data centers disconnected from the grid, including a July 22 event in Northern Virginia where roughly 3,100 MW of demand vanished after a transmission line fault. The authors argue that with current technology, AI facilities can become 'good grid citizens' that advance energy affordability and grid reliability rather than threaten it.

The call to action follows a May alert from the North American Electric Reliability Corp., which issued its highest-level warning due to urgent threats to the power system. Federal regulators have also issued new directives for connecting large loads. The authors represent companies involved in building power supply, computing systems, power electronics, and orchestration software for AI infrastructure.

Do No Harm to the Grid

The foundational principle is that an AI data center should do no harm. Current facilities often fail this standard in two key ways. First, they disconnect entirely from the grid during brief voltage dips, switching to backup power and creating a massive, instantaneous drop in demand. The 3,100 MW shock in Virginia originated from this type of disconnection. Second, the power draw of modern AI workloads can swing by tens of megawatts in seconds, creating fast and large power swings that stress the local grid.

The authors state that the technology exists to solve these problems. Batteries at data centers can provide backup power to graphics processing units and also export power to maintain grid stability during voltage disturbances. This allows a facility to 'ride through' a disturbance rather than abandoning the grid. "Do no harm is the floor, but built into the AI data center rather than promised by its operator, it becomes something the grid can actually rely on," they write.

Provide Flexibility During Stress

The second obligation is for AI data centers to offer flexibility when the grid system is stressed. The traditional power system has operated on a one-way assumption where demand is largely inflexible. AI data centers can change this dynamic. The authors claim these facilities can be designed to modulate their power consumption in response to grid signals, providing a valuable grid service.

This flexibility can help balance supply and demand during peak periods or when renewable generation is variable. By reducing consumption when the grid is strained, data centers can alleviate stress and potentially prevent more costly interventions. The principle moves beyond passive grid connection to active participation in maintaining stability.

Add to Grid Capacity

The most ambitious principle is that AI data centers should actively help bring new generation, storage, and transmission infrastructure into service. The authors argue that these large, flexible loads can be strategically used to strengthen the overall system. By coordinating their development with new energy projects, data centers can provide the demand certainty that helps finance and build new grid assets.

This approach could accelerate the deployment of renewable energy and storage by creating a stable, large-scale customer for their output. It could also support the business case for new transmission lines. The authors state that all three obligations can be met with commercially available technologies today and, for the first time, can be integrated within a single architecture that their four companies can offer together.

The article frames the rapid AI buildout not solely as a grid challenge but as a potential catalyst for a more resilient and affordable electric system. The authors acknowledge community anxiety about rising energy bills and reliability but contend that intentional design can align data center growth with broader grid benefits. The proposal comes as communities and regulators increasingly scrutinize the impact of large new loads on the existing power infrastructure.