SPAN Pilots Home AI Data Centers, Targeting 80,000 Nodes and 1 Gigawatt by 2027
Updated
Updated · Ars Technica · May 12
SPAN Pilots Home AI Data Centers, Targeting 80,000 Nodes and 1 Gigawatt by 2027
1 articles · Updated · Ars Technica · May 12
SPAN has begun pilot testing AI data-center nodes installed alongside new homes, with a 100-home trial planned later this year.
Thousands of XFRA units would use liquid-cooled Nvidia RTX Pro 6000 Blackwell Server Edition GPUs, tapping spare household power to add AI compute without building large centralized facilities.
The startup says hosts would get subsidized electricity, internet access and backup batteries, while the quieter setup could avoid the land, water and neighborhood backlash tied to conventional data centers.
SPAN told CNBC it could deploy 8,000 units at one-fifth the cost of a typical 100-megawatt data center with similar compute capacity.
Starting in 2027, SPAN aims to expand to 80,000 nodes nationwide and deliver more than 1 gigawatt of distributed compute for AI inference, cloud gaming and streaming rather than model training.
Is scattering mini data centers in suburbs a green solution or just hiding AI's energy problem in plain sight?
Could turning homes into data centers create a massive cybersecurity backdoor for hackers targeting critical AI infrastructure?
When your home becomes a computing node, who truly profits from the untapped power sitting in your walls?
SPAN XFRA: Deploying 80,000 Home-Based AI Data Centers for 1 GW Distributed Compute by 2027
Overview
SPAN is launching its XFRA initiative in May 2026, moving from concept to a pilot program that will integrate AI data center capabilities into 100 homes. This marks a major step in decentralizing AI infrastructure by establishing a network of home-based data centers. XFRA aims to deploy 80,000 nodes to reach 1 gigawatt of distributed compute capacity by 2027. The shift is driven by the rising demand for AI compute power and the challenges faced by traditional, centralized data centers. By lowering deployment friction and diversifying infrastructure, SPAN is setting the stage for a more resilient and scalable AI ecosystem.