TECHNOLOGY — U.S. residential solar and energy storage provider Sunrun is testing a “distributed AI data center” model that could use computing nodes deployed in homes to support artificial intelligence inference workloads.

Sunrun CEO Mary Powell said Monday that participating households could potentially earn hundreds of dollars per month by providing computing capacity. The business model, however, remains in the pilot stage.

Sunrun has more than one million customers with rooftop solar systems across the United States. The company is exploring the deployment of computing equipment in customers’ homes, effectively turning participating residences into small-scale computing nodes.

Powell said the company is examining how its existing residential solar and battery network could help support growing energy and computing demand in the United States.

The pilot is primarily focused on AI inference. Unlike large-scale AI training, some inference workloads can be distributed geographically and processed closer to end users.

Sunrun says deploying computing infrastructure in homes could potentially reduce some of the land, transmission, and grid interconnection challenges associated with building large data centers.

The company has held discussions with several potential buyers of computing capacity. Over the coming months, Sunrun plans to evaluate computing performance, business objectives, and the customer experience before determining whether to expand the program.

The initiative follows other efforts by Sunrun to aggregate residential energy resources. The company previously announced partnerships, including with Tesla, involving more than 16 gigawatts of flexible home energy capacity for utilities and large-scale cloud computing customers.