Home-Based Mini Data Centers to Accelerate AI Computing, SPAN Offers Electricity Bill Subsidies
Startup SPAN announces plans to deploy distributed data center nodes with liquid-cooled GPUs for home use. The company will subsidize residents' electricity and internet costs to rapidly scale AI computing.
Data Centers in Homes? A New Approach to AI Computing
To rapidly scale the computational resources needed for training and inference of AI models, a bold plan has emerged to place data center nodes directly in residences. San Francisco-based startup SPAN has announced a “distributed data center solution” that integrates liquid-cooled GPU nodes into newly built homes, aiming to avoid the costs and delays associated with constructing large-scale data centers.
Technical Details and Benefits for Residents
The “XFRA” nodes deployed by SPAN are equipped with liquid-cooled Nvidia RTX Pro 6000 Blackwell Server Edition GPUs, which are said to minimize operational noise. These nodes leverage the surplus power capacity of each household to build computational resources for AI workloads.
The biggest benefit for residents is the subsidy on electricity and internet bills. SPAN will cover the electricity and internet costs for each household, allowing residents to use these services at a low fixed rate (an example of $150 per month is suggested) or potentially for free. This is expected to alleviate concerns about rising electricity bills caused by data centers.
Cost Reduction and Scaling Plans
SPAN claims that the cost of installing 8,000 XFRA nodes is one-fifth the construction cost of a typical 100-megawatt data center with equivalent computing capacity. Furthermore, the company revealed plans to deploy over 80,000 nodes across the United States starting in 2027, providing a total computing capacity exceeding 1 gigawatt.
However, this distributed network is not intended to replace the large, centralized data centers built by hyperscalers like Google or Microsoft. SPAN positions its system as better suited for cloud gaming, content streaming, and “AI inference”—applying trained models to real-world tasks—rather than training AI models.
Future Outlook and Challenges
Pilot tests are currently underway, with a trial operation targeting 100 households planned within the year. While future plans include retrofitting existing homes and offering larger configurations for commercial customers, the initial focus will be on installation in newly built homes.
This approach could potentially avoid the land use and water consumption issues associated with large-scale data centers and reduce community opposition. However, challenges remain for practical implementation, including securing installation space within homes, long-term maintenance, and concerns about resident privacy.
Frequently Asked Questions
- Will residents' electricity bills really become cheaper with the installation of home data center nodes?
- Under SPAN's plan, the company will cover the electricity and internet costs for each household. Residents may be able to use services at a low fixed rate or even for free, which is expected to reduce their utility expenses. However, specific pricing plans will be finalized later.
- Can these nodes be installed in existing homes?
- At this stage, the initial phase of the plan prioritizes installation in newly built homes. SPAN is considering retrofitting existing homes in the future, but specific timelines or methods have not yet been announced.
- Is it safe to have a data center node inside a home?
- SPAN explains that the XFRA nodes are liquid-cooled, low-noise, and carefully designed. Additionally, the company will handle installation and operation entirely, so residents do not need to be directly involved in technical management. However, the long-term safety and reliability will depend on the results of future demonstration tests.
Comments