U.S. residential solar and storage provider Sunrun is advancing the pilot of the so-called "distributed AI data center." CEO Mary Powell said on Monday that for users participating in this program, they may be able to earn several hundred U.S. dollars per month by selling computing power.

It is understood that Sunrun has more than 1 million users in the United States who have already deployed rooftop solar systems. Facing the AI industry’s frenzy-driven pursuit of computing power, solar companies have seen an opportunity to "sell shovels."

Powell told the media that, as U.S. energy demand continues to hit new highs, it is projected to grow by another 40% over the next decade. The company has found that its own business is indeed potentially an important part of solving this problem: by working with customers to deploy nodes in their homes, these households can effectively function as "micro data centers" and receive corresponding compensation as a result.

More precisely, the company is targeting the AI inference track. Unlike AI training, which requires large-scale and tightly synchronized systems, inference is modular, geographically distributed, and highly latency-sensitive. This makes it naturally well-suited for deployment at the edge, close to end users. At the same time, deploying computing infrastructure in residential homes can also save on land acquisition for data center construction, transmission grid build-out, and electricity congestion queues.