Mini Data Centers in San Diego Backyards?
A major homebuilder wants to host AI compute on the side of new homes. Here are the questions buyers and sellers should be asking.
PulteGroup, Nvidia, and Span are piloting a program to host AI compute nodes on new homes. The idea is clever. The implications for resale value, HOAs, and insurance are not yet priced in.
This week CNBC reported that PulteGroup, the third largest homebuilder in the country, is testing a program with smart panel maker Span and chipmaker Nvidia to install small AI compute nodes on the outside of newly built homes. The first pilot covers about 100 homes, likely in Nevada or Arizona, and is scheduled for the third quarter of this year.
The pitch to homeowners is straightforward. You let Span mount a liquid cooled box on the side of your home, roughly the size of an HVAC condenser, packed with sixteen Nvidia GPUs and four server processors. Span uses your home's spare electrical capacity to run AI workloads on behalf of cloud customers. In exchange, you get a discount on your electricity bill and on your internet.
It is a creative answer to a real problem. The traditional data center model is hitting walls everywhere. Counties are saying no. Power grids are saturated. Land is expensive. Distributing the compute across thousands of homes that already have power, an internet connection, and a roof over them sidesteps most of those fights.
I find the engineering elegant. I am also paid to ask what could go wrong with elegant ideas. Here are some questions I have.
Question 01What does this do to resale value?
A new home with an XFRA node on the side wall is a different product than a new home without one. Whether that difference is a positive or a negative for the next buyer depends on how the discount is structured, whether it transfers, what the maintenance obligations look like, and how the appraisal industry decides to treat it.
If the energy savings convey to the next owner cleanly, the node could read as a feature. If they do not, or if the contract creates an obligation the next buyer has to assume, it reads as an encumbrance, and encumbrances tend to compress price.
No one has done a comparable sales study on this yet, because the comparable sales do not exist. That uncertainty is the thing buyers should price in, not the technology itself.
Question 02Who owns the equipment, and who insures it?
Public reporting suggests Span owns the node and the homeowner hosts it. That arrangement is common in solar leases and has been a recurring source of friction at resale, particularly when the next buyer is using a lender who does not like the lien structure.
There is also a basic insurance question. A box containing roughly $300,000 worth of advanced GPUs sitting outside a home is an attractive target. Whose policy covers theft. Whose policy covers a fire that starts in the equipment and spreads to the home. Whose policy covers a fire that starts in the home and damages the equipment. These are answerable questions, but they have not been answered in public yet.
Question 03What happens when 50 homes share a transformer?
The Span pitch leans heavily on using existing, unused electrical capacity. That math works at the level of one home. It is less obvious at the level of a cul de sac, a phase, or a substation, particularly in California where the grid is already constrained.
In a master planned community of 200 new homes, what percentage hosts a node before the local utility has to upgrade the feeder. Who pays for that upgrade. Does the HOA have a say in any of this. These are the questions a buyer should be asking the builder before signing, not after closing.
Question 04What does the HOA think?
Most newer HOAs in San Diego County have detailed rules about what equipment can sit visibly on the exterior of a home. Satellite dishes, solar arrays, and even certain HVAC configurations have all been litigated. A liquid cooled compute box mounted next to the air conditioner is going to be a new entry on that list.
For buyers in a community where these are being deployed, the relevant question is whether the HOA has signed off, whether the rules apply uniformly, and what happens if a future board changes its mind.
Three risks the marketing does not address
The four questions above are what a buyer should ask. The three issues below are what I think the program is most likely to be wrong about.
If Span requires a UCC 1 fixture filing on the equipment or a recorded easement for maintenance access, that directly impacts title. Fannie Mae and Freddie Mac have strict guidelines on third party liens on residential properties. If conventional lenders refuse to underwrite these homes because of the contract structure, the buyer pool evaporates and resale gets ugly fast. This is the single largest risk in the program, and it has not been addressed publicly.
A homeowner who receives compensation, whether as a utility discount or a future direct payment, is effectively leasing residential space to a commercial enterprise. That raises two questions the program glosses over. Does hosting a commercial compute node violate residential zoning ordinances in San Diego County. And does the compensation trigger taxable income, including potential self employment tax exposure depending on how the arrangement is structured. Neither answer is obvious.
The marketing emphasizes that liquid cooling eliminates noise. That deserves pushback. Liquid cooling moves heat away from the chips, but the heat still has to go somewhere, which means heat exchangers and pumps on the exterior of the home. If 200 homes in a master planned community are running inference workloads at peak capacity on a 95 degree August afternoon, the ambient heat and noise profile of the neighborhood will change. Buyers should not take the silent operation claim at face value until there is a deployed community they can stand in.
Weighing the tradeoffs
For the homeowner considering hosting one of these nodes, the case sits on both sides of the ledger.
| Pros | Cons |
|---|---|
| Lower utility costsHosts may see internet and electricity bills replaced by a flat fee in the range of $150, or potentially reduced to zero. | Visual impactAnother large white box mounted on the exterior, alongside the smart panel and battery, adds to the equipment cluttering the side of the home. |
| Infrastructure includedThe installation comes with a smart electrical panel and a backup battery at no cost to the homeowner. | Privacy and securityHosting high value server hardware on the property may attract theft and raises questions about data flowing through the home network. |
| A more sustainable growth pathDistributed compute reduces the pressure to build massive industrial data centers that consume large tracts of land and strain local water supplies. | Maintenance accessTechnicians may need periodic access to the side yard or backyard for repairs and hardware upgrades. |
| Income potentialFuture iterations may move beyond utility discounts and offer homeowners direct payments for hosting a node. | Resale uncertaintyFuture buyers may be wary of renting yard space to a tech company, or may have concerns about perceived heat and noise from the equipment. |
The honest answer is that we do not yet know how the market will weigh these against each other, because the program has not been live long enough to generate the data. The first wave of buyers in the pilot communities will be the ones telling us.
Where this could land in San Diego County
The pilot is targeting newer master planned communities with single family construction, modern electrical infrastructure, and HOAs accustomed to managing exterior equipment. Several San Diego County neighborhoods fit that profile cleanly.
In North County Coastal, Carmel Valley and Pacific Highlands Ranch are obvious candidates. Both are dominated by newer construction with the kind of side yards and electrical panels the program assumes, and both have HOAs with established design review processes. Del Sur and 4S Ranch in the inland North County corridor along the I-15 are similar profiles, with the added advantage of slightly larger lots in many of the phases.
A bit further north, Bressi Ranch in Carlsbad and the newer phases of San Elijo Hills in San Marcos have the same DNA: master planned, single family, newer build, HOA managed.
Heading south, Otay Ranch in Chula Vista is the largest master planned community in the county and an obvious scale opportunity for any builder running a pilot, especially given the active new construction across its eastern villages. Park Circle in Valley Center is the freshest entrant, with five builders delivering homes in the same community, which is exactly the kind of multi builder environment where a program like this could be tested across more than one product line.
None of these communities have announced anything. The point is not to predict which neighborhood gets picked first. The point is that if you are buying new construction in any of them this year, the question of whether the builder is in conversation with Span, with PulteGroup's pilot, or with a competing program is now a reasonable thing to ask in writing during the contract phase.
The takeaway
The technology is interesting. The contracts and the resale economics are what will determine whether this becomes a feature on the listing or a footnote in the disclosures. San Diego buyers and sellers do not need to react to the headlines. They need to understand the mechanics now, so when the program does arrive in our market, you are asking the right questions of the builder, the HOA, and your lender, rather than learning the answers the expensive way.