Bay Area houses may soon be doing more than keeping the lights on. A trio of energy and tech firms is pitching a plan to turn ordinary homes into dispatchable power plants for the data centers fueling the artificial intelligence boom. In their proposal, rooftop solar, home batteries, smart thermostats and EV chargers would be choreographed to trim peak demand and free up electricity for compute-heavy facilities, which they argue can happen faster and more cheaply than waiting years for new transmission lines or large generators. For Bay Area homeowners, that could translate into small monthly payments and some new equipment mounted on the side of the house.
The companies, which include Tesla, Sunrun and a device-aggregation firm, say the strategy would pool millions of household devices and dispatch that capacity to data center operators during crunch hours. They estimate the freed-up power could match the needs of about 17 large data centers and amount to roughly 16 times as much electrical capacity as California's program across the country, as reported by The New York Times.
Why the grid is under strain
AI-focused data centers are already pushing U.S. electricity use toward new records and changing how facilities are designed and cooled, a shift that is ramping up near-term stress on local grids and transmission systems, according to reporting by the Los Angeles Times. That growing strain is a key reason companies are pushing distributed, behind-the-meter fixes instead of relying only on multi-year utility buildouts.
Sunrun's numbers and the home-to-grid pitch
Sunrun says its distributed power plant programs grew rapidly in 2025, operating 17 programs and enrolling more than 106,000 participants while dispatching nearly 18 gigawatt-hours of battery energy with a peak output comparable to many peaker plants. The company also reported installing more than 237,000 solar-plus-storage systems by the end of 2025, giving it a sizable fleet of residential batteries and panels that can be coordinated for grid services. According to Sunrun's prepared remarks, that scale is central to meeting demand spikes tied to AI.
How regulators and utilities are responding
Regulators are treating the AI buildout as a grid-wide problem. The federal energy regulator recently ordered regional grid operators to justify or rewrite rules for connecting very large energy users such as data centers, a move intended to speed up or reshape how those hookups occur. Utilities point out that building new lines and power plants usually takes years and can cost billions, which is why some see residential aggregation as a short-term patch.
Mini data centers, in your backyard?
Another version of the concept puts small, liquid-cooled compute nodes on or next to homes so they can run inference or other AI workloads when household demand is low. Startups such as Span, working with Nvidia and homebuilder PulteGroup, are piloting so-called XFRA nodes that combine a smart electrical panel, battery backup and compact servers. Fortune has described the model as a way to add compute capacity close to users without waiting for major grid upgrades, and reports that builders and chipmakers are testing the economics in new homes.
Households already receive payments to participate in some aggregation programs. One aggregator told investors its customers earned money for dispatches, and The New York Times reports that Sunrun and other aggregators together paid about $67 million to participating households last year. That cash is a central part of the recruitment pitch, although payments and eligibility vary by market and by which devices are installed in a given home.
What this means for neighbors and next steps
Not everyone is eager to plug into the AI surge. Communities and local planners have been pausing or tightening data center approvals while they study water, noise and grid impacts, and some counties have imposed moratoria while they decide whether and how to allow large compute projects. Local examples of that pushback are emerging in places like Lake County, which has paused new AI data-center approvals as officials consider changes. Hoodline coverage underscores that the rollout will be as much political and regulatory as technical, and that pilots and policy shifts will determine whether homes become a routine part of AI's power stack.
