A startup called Span is offering homeowners money to mount small AI computing units on the exterior walls of their homes, in a push to spread AI infrastructure beyond traditional data centers.
The program has ties to NVIDIA, according to one report, which describes the hardware as roughly the size of a standard exterior unit. Reports diverge sharply on the financial terms, however. One source describes payments of over $220,000 per year, while another puts the figure at over $22,000 annually — a tenfold gap that the available sources do not explain.
Fast Company took a skeptical angle, framing its coverage around the question of "who really pays" — a pointed hint that homeowners might find their electricity bills, heat exposure, or liability quietly eating into whatever payments they receive. Hosting high-powered computing hardware around the clock is not free, and the real economics for a typical household remain murky.
The broader idea is not new. Distributed or "edge" computing pushes processing closer to where data is generated or consumed, cutting the need to pipe everything through distant, centralized facilities. As AI workloads explode in scale, companies are under growing pressure to find creative ways to add computing capacity quickly and cheaply — and residential walls represent untapped real estate.
For homeowners, the pitch sounds attractive on its face, but the wildly inconsistent payment figures circulating in coverage suggest the program's terms are still poorly understood, or that multiple distinct offers are being conflated.
If distributed home-hosted AI infrastructure scales, it could quietly turn residential neighborhoods into nodes of the AI supply chain — shifting some of the energy and infrastructure burden of the AI boom directly onto private property.