Overview
- Sunrun began a distributed AI compute pilot that installs inference compute nodes in participating customer homes, with the company paying homeowners to host the hardware and sell the compute capacity to enterprise buyers.
- The pilot, announced Wednesday, expands a proof-of-concept into a multi-month test that will evaluate node performance, electricity rate structures, and the homeowner experience before any decision on wider deployment.
- Sunrun plans to leverage its roughly 1.1 million solar and battery installations to operate the network and to run nodes behind customers’ meters so units can keep processing during some grid outages using home battery power.
- The offering targets AI inference workloads, which Sunrun cites McKinsey as projecting to grow about 35% annually and to outpace training by 2030, making geographically distributed, low-latency nodes attractive to AI firms.
- The pilot complements Sunrun’s separate 16 GW residential flexibility framework with Tesla and Renew Home, but scale and economics remain uncertain because the company has not disclosed rollout timelines, revenue targets, or how regulators and utilities will approve grid interactions.