Tech1 hr ago

SPAN Targets 80,000 Home‑Based Data Centers by 2027, Claiming Five‑Fold Cost Savings

SPAN aims to install 80,000 quiet, low‑cost XFRA nodes in U.S. homes, delivering over 1 GW of AI compute and cutting costs five‑fold versus traditional data centers.

Alex Mercer/3 min/US

Senior Tech Correspondent

TweetLinkedIn
SPAN Targets 80,000 Home‑Based Data Centers by 2027, Claiming Five‑Fold Cost Savings
Source: SpanOriginal source

SPAN will roll out 80,000 home‑based XFRA data nodes by 2027, promising five‑times lower cost than a traditional 100‑MW data center and more than 1 GW of distributed AI compute.

Context The United States faces a surge in demand for AI inference power, while traditional warehouses strain land, water and local electricity grids. SPAN, a San Francisco startup, proposes to embed compute in residential neighborhoods, pairing each node with subsidized power, broadband and backup batteries. A pilot program will test the model in 100 homes later this year.

Key Facts - Each XFRA node houses liquid‑cooled Nvidia RTX Pro 6000 Blackwell Server Edition GPUs, engineered for minimal noise and a compact footprint. - Chris Lander, SPAN’s vice president of XFRA, describes the units as “quiet, discreet, and makes energy more affordable for the host and community.” - SPAN estimates that installing 8,000 XFRA units costs five times less than building a conventional 100‑megawatt data center that delivers the same compute capacity. - The company’s roadmap calls for 80,000 XFRA nodes nationwide by 2027, collectively providing over 1 gigawatt (1,000 megawatts) of distributed computing power. - The network is aimed at workloads such as cloud gaming, content streaming and AI inference, not the massive training jobs handled by hyperscalers like Google or Microsoft.

What It Means If SPAN’s cost model holds, homeowners could earn subsidies while keeping electricity bills low, and communities would avoid the visual and acoustic impact of large facilities. Distributed compute also reduces latency for edge‑focused services, potentially improving real‑time AI applications. However, the approach relies on widespread homeowner participation and stable grid capacity, factors that will shape adoption.

Looking Ahead Watch for the outcome of the 100‑home trial and any regulatory response to large‑scale residential power use as SPAN moves toward its 2027 deployment goal.

TweetLinkedIn

More in this thread

Reader notes

Loading comments...