Startups are deploying cabinet-sized data center units in residential homes and small businesses, promising to reduce strain on the electrical grid while generating revenue for homeowners. However, leading researchers caution that these distributed systems could inadvertently encourage even greater data center expansion, ultimately worsening environmental impact rather than solving it.

What Are These Home Data Centers, and How Do They Work?

California-based Span, partnering with Nvidia, has begun deploying prototype data center "nodes" called XFRA in Northern California. These cabinet-sized units install on the sides of homes and small businesses, requiring no cooling fans, which eliminates the noise pollution that has frustrated residents near traditional warehouse data centers. Span charges homeowners a flat monthly fee of about $150, and in return, the company covers electricity and internet bills while using the computing power generated from the nodes to serve hyperscalers and artificial intelligence (AI) companies.

The company estimates XFRA will generate about one to two megawatts of computing capacity later this year, with plans to scale to more than one gigawatt annually beginning next year. PulteGroup, one of the largest homebuilders in the United States, is testing the system. Nvidia will supply liquid-cooled RTX PRO 6000 Blackwell Server Edition graphics processing units (GPUs) for the infrastructure.

"We do see a path to being able to contribute on an annual basis hundreds of megawatts, if not gigawatts, of scale compute capacity, while doing so in a deflationary-to-energy-price way," said Ryan Harris, chief revenue officer of Span.

Ryan Harris, Chief Revenue Officer at Span

Span can install nodes at six times the speed of centralized 100-megawatt data centers and at about one-fifth of the construction cost. A UK-based startup called Heata takes a different approach, installing servers that process cloud computing workloads while using thermal conductors to carry heat from computer processors to cylinders filled with water for home heating. Heata has installed units in about 100 homes and claims to have saved approximately one gigawatt-hour of energy, with about 70 percent of savings coming from reduced need for domestic gas or electric heating systems.

Why Are Companies Pushing Home Data Centers Now?

The push for distributed data center infrastructure reflects mounting tensions between hyperscalers and communities. McKinsey projected in April 2025 that AI infrastructure would require $7 trillion in capital expenditures by 2030. Traditional warehouse-scale data centers, some as large as dozens of football fields combined, have strained the United States' electrical grid system and potentially driven up electric bills by 6 percent over the next year, according to Goldman Sachs research. Beyond energy concerns, data centers consume massive amounts of water for cooling; two data center developments in Arizona and Georgia took public water without authorization, and a recent study by the Houston Advanced Research Center projected that data centers would drain as much as 399 billion gallons of water in Texas alone by 2030.

Home-based systems promise a solution by distributing computing load across residential areas, using existing electrical capacity that would otherwise sit idle. This approach appeals to both companies seeking to expand AI infrastructure and communities seeking to avoid large industrial facilities in their neighborhoods.

What Do Experts Say About the Real Environmental Impact?

Despite the appeal of home data centers, a critical analysis from Utah State University physics professor Robert Davies raises serious concerns about whether these systems actually solve the underlying problem. In a preliminary analysis, Davies calculated that only 30 to 40 percent of homes may be suitable for mini data centers due to integration constraints, the need for stable internet connectivity, and residents' willingness to have the technology installed. Additionally, only 2 to 3 percent of homes could realistically be heated through alternative energy-harnessing technologies because of constraints on how much waste heat can be collected, and heating energy often goes to waste in many geographies where heating is a seasonal need.

"These projects tend to be heavy on the benefit analysis and very light on the cost analysis. And you don't actually get a full sense of the cost until you do a whole systems analysis. These are multi-generational challenges, and are they solving problems that we really need solved?" said Robert Davies, physics professor at Utah State University.

Robert Davies, Physics Professor at Utah State University

Davies invoked Jevons paradox, a 160-year-old economic theory named after English economist William Stanley Jevons. The principle states that as a resource becomes more efficient to use, people tend to consume more of it rather than less. Jevons observed that better steam engines made coal cheaper, which subsequently increased total coal consumption. Davies applied this logic to data center efficiency: "We now need about 45 percent less energy to do the same thing that we needed 35 years ago. So that seems awesome. Are we using 45 percent less energy than we were 30 years ago? The answer is, no. Turns out, we're using about 70 percent more energy".

Jevons

How Could Efficiency Improvements Backfire?

Davies warned that framing data center expansion as something that can be made more efficient could be a dangerous precedent. If repurposing data center waste heat becomes easier and cheaper, companies may simply build more data centers, knowing they can offset some environmental costs. This creates a perverse incentive: the more efficient the system becomes, the more aggressively companies expand, ultimately consuming more total resources than if the efficiency improvements had never occurred.

The Heata spokesperson countered this concern by arguing that the company deals with substitution rather than increased efficiency, since homes must be heated whether or not a server is present. However, Davies remains skeptical that the demand for computing power can be met through waste heat repurposing alone, and he fears that continued expansion will further burden environmental capacity.

"The strategy that I see the sector applying here is seductive. It seems useful, we want it to be useful. But in a whole systems analysis, it's really not," Davies concluded.

Robert Davies, Physics Professor at Utah State University

Steps to Evaluate Home Data Center Proposals in Your Community

• Request a Full Systems Analysis: Before approving any home data center installation program, ask local officials and companies to provide a comprehensive environmental impact assessment that accounts for total energy consumption, water usage, and long-term infrastructure expansion plans, not just efficiency gains.
• Examine Suitability Constraints: Understand that only a fraction of homes can actually support these systems due to internet stability requirements, electrical capacity limitations, and physical space constraints; ask what percentage of your neighborhood would realistically qualify.
• Question the Heating Claim: If a company claims waste heat will offset home heating costs, request independent verification that the seasonal heating demand in your region actually aligns with year-round computing loads.
• Investigate Expansion Plans: Ask companies directly whether efficiency improvements will lead to additional data center construction in your area, and request binding commitments to cap total infrastructure growth.

What Does This Mean for the Future of AI Infrastructure?

The emergence of home data centers reflects a genuine tension in the AI industry: demand for computing power is growing faster than the grid can support, and communities are increasingly resistant to massive warehouse-scale facilities. Distributed systems offer a politically attractive compromise, allowing companies to expand AI infrastructure while appearing to address environmental and community concerns.

However, Davies' analysis suggests that without addressing the fundamental question of whether all this computing capacity is necessary, efficiency improvements may simply enable faster expansion. The plastic piping market for data centers, which is projected to grow from $2.06 billion in 2026 to $6.53 billion by 2032 at a compound annual growth rate of 21.2 percent, reflects the scale of ongoing infrastructure investment across all data center types, including hyperscale facilities. This growth suggests that even as home-based systems come online, traditional large-scale data centers will continue expanding.

The real test of home data centers will not be whether they reduce energy consumption per unit of computing power, but whether they actually slow the total growth of data center infrastructure. If they simply enable faster expansion while appearing more environmentally friendly, they may ultimately accelerate the very problems they were designed to solve.