Connecting a new power plant or solar farm to the grid can take years, largely because operators must run slow, complex studies before approving each interconnection request. On April 10, 2025, Google announced a partnership with PJM Interconnection - North America’s largest grid operator, serving 67 million people across the District of Columbia and 13 Mid-Atlantic and Midwest states - and Tapestry, an Alphabet-incubated moonshot, to apply AI to this bottleneck.
Tapestry, powered by Google Cloud and Google DeepMind, is building AI tools aimed at three goals: speeding the interconnection application and verification process, improving planning efficiency by integrating dozens of existing databases into unified models, and supporting the addition of more variable energy resources such as wind and solar. The stated aim is to shorten project approval timelines while keeping the grid reliable and affordable.
This matters because the pace of grid connection is now a major constraint on deploying new energy - including the power that AI data centres themselves demand
- so using AI to plan the grid is an unusually direct case of the technology being turned on its own infrastructure problem.