> ## Content Index > Fetch the complete content index at: https://nexi.fund/llms.txt > Use this file to discover other available public pages before exploring further. # Google Is Paying Home Batteries to Free Up the Grid for Its Data Centers - URL: https://nexi.fund/pge-google-virtual-power-plant-2026/ - Published: 2026-09-30T13:00:43.000Z - Updated: 2026-09-30T13:00:43.000Z - Description: PG&E launched SHARE on September 3, 2026 โ€” a virtual power plant fully funded by Google that enrolls almost 21,000 home devices in the Bay Area. It is Google's third demand-side flexibility deal this year, and a test of whether flexible capacity can arrive faster than a new transmission line. - Author: Nexi.fund Labs - Tags: Energy & Climate, #mode-6, #hook-question, #track-A, #brand-heavy This autumn, a few thousand homes in California's Bay Area will start getting paid to use less electricity at the moment the grid is tightest. The cost of enrolling them is being picked up by Google. The program is called SHARE โ€” Smart Home Assets for Reliability and Efficiency โ€” and PG&E launched it on September 3, 2026\. Home batteries, smart thermostats and battery-enabled heat pumps will be linked into one network that the utility can dispatch the way it dispatches a power plant. PG&E and its partners plan to enroll close to 21,000 devices that are already installed, starting in Santa Clara and Alameda counties, and to begin leaning on them for grid support this fall. The part worth pausing on is who pays. SHARE is fully funded by Google โ€” the customer incentives, the new hardware, the program overhead. The company that is adding electricity demand faster than almost any other buyer on earth has decided that the cheapest new capacity within reach is the flexibility sitting inside other people's homes. ๐Ÿ“Š **Where this stands** PG&E launched SHARE on September 3, 2026, with Rewiring America, Google, Carrier, Tesla, Sunrun and Renew Home as partners. The virtual power plant will enroll almost 21,000 existing home devices across Santa Clara and Alameda counties. Google funds the entire program; the first phase has been reported at roughly $14 million. Devices begin supporting the grid this fall, the pilot runs through 2027, and initial findings are expected in late 2026 or early 2027\. $14M Google phase-one funding #### Who pays for SHARE Google covers customer incentives, hardware deployment and program delivery in full. ยท *Rewiring America, 2026* 21,000 home devices to enroll #### The resource base Existing batteries and smart devices enrolled by Tesla, Sunrun and Renew Home. ยท *PG&E, 2026* 100 MW Google-Voltus capacity, PJM #### The first Google VPP Google pays Voltus to aggregate up to 100 MW a year of flexible load in the largest US grid. ยท *Voltus, 2026* 250 MW Google San Jose data center #### The load in question The nearby data center whose grid costs are the subject of an ongoing regulatory fight. ยท *CPUC filings, 2026* ## A power plant made of other people's batteries The mechanical idea is old. A virtual power plant is software that treats thousands of small, scattered devices as one dispatchable block. When demand spikes, the operator either discharges the batteries or nudges the devices to draw less. The individual adjustments are invisible. The aggregate behaves like generation. SHARE goes further than most pilots in two ways. It enrolls hardware that is already paid for, which lowers the cost of the first megawatts, and it points at a specific bottleneck: capacity on the regional transmission system. PG&E says the program is meant to unlock additional headroom on the wires that carry power into the Bay Area, not to trade energy in wholesale markets. That framing matters, because transmission is the part of the grid that takes longest to build. The orchestration sits with PG&E and Demand Side Analytics. Carrier supplies the launch technology โ€” a variable-speed heat pump with an integrated battery that stores and shifts energy use without the homeowner noticing. Rewiring America handles enrollment on the household side, working with local installers so that the upgrades meet performance standards. The device count is the tell. Rewiring America, Tesla, Sunrun and Renew Home are being asked to bring almost 21,000 flexible devices under one dispatch signal. That is small against the load it is meant to help serve. Google's own data center in San Jose is a 250-megawatt customer. A residential fleet of 21,000 devices will not cover it. What the pilot can prove is whether the mechanism works at all โ€” whether a utility can reliably count on consumer hardware, and whether the households stay enrolled once the novelty wears off. #### What "dispatchable capacity" actually means here A virtual power plant does not generate electricity from nothing. It moves demand in time. A home battery that would otherwise charge at 6pm can be told to wait until 11pm, and a heat pump can pre-heat a room before the peak and coast through it. The grid never sees a new electron; it sees a different schedule. Capacity is credited when the devices are available to be called, not when the call is answered. The hard engineering problem is not the software โ€” it is proving, to a regulator, that the flexibility will be there on the worst day of the year, when every household wants the same thing at the same time. **What has to be proven:** that enrolled homes deliver their share of capacity with enough reliability to count toward a transmission constraint. ## Google has now bought flexibility three times SHARE is the third structure Google has used in a single year to convert other people's energy behaviour into capacity for its own operations, and the pattern is consistent. In June, Google and the VPP operator Voltus signed a three-year agreement to aggregate up to 100 megawatts of distributed resources across PJM, the grid that spans parts of the Mid-Atlantic and Midwest. Google pays for the network; participating homes and businesses are paid for letting their devices be adjusted. It was described at the time as the first deal of its kind, and it moved virtual power plants from a utility pilot into a hyperscaler's procurement line. The same month, Sunrun, Tesla and Renew Home announced plans for a 16-gigawatt virtual power plant aimed at supplying both utilities and large technology buyers. In early September, Sunrun and Tesla said they had dispatched 580 megawatts into California's grid, with one event triggered when day-ahead wholesale prices crossed $200 per megawatt-hour. The fleet did not need to be built. It was already on the walls of homes and businesses across the state, waiting for a signal. Read together, the three moves describe a shift in how the largest power buyers think about supply. A decade ago the answer to growth was a long-term contract with a generator. Now it includes paying for flexibility on the demand side, because flexibility can be switched on in months instead of the decade a transmission line requires. > "This is about delivering power at the speed our economy demands โ€” while improving affordability and reliability for the people we serve."โ€” Chelle Izzi, Chief Commercial Officer, PG&E ## The economics: flexibility beats steel on time, not on scale The case for virtual power plants is a case about lead time. A new high-voltage transmission line in the United States routinely takes a decade or more to permit and build; developers surveyed for a 2026 industry report described timelines stretching toward 15 years, and roughly 56 gigawatts of on-site generation was planned by data center builders as a workaround. Gas peakers can be built in a fraction of that time, but they are now politically expensive and increasingly contested at the local level. Demand-side flexibility sidesteps both delays. It is assembled from hardware that already exists, enrolled in months, and paid for by whoever needs the capacity. The trade is scale. A residential VPP is measured in tens or hundreds of megawatts. The data center next door to SHARE's service area alone draws 250 megawatts. Virtual capacity can bridge a gap; it cannot close one. | Option | Lead time | What drives cost | Scale | | ------------------------- | -------------- | -------------------------------- | ------------------------ | | **Virtual power plant** | Months | Payments to households; software | โ— Tens to hundreds of MW | | **Gas peaker** | 2โ€“4 years | Fuel, permits, local opposition | โœ” Hundreds of MW | | **New transmission line** | Up to 15 years | Land, permitting, steel | โœ” Gigawatts | Industry lead-time figures, Reuters Events / pv magazine USA, 2026 The cost structure explains why a buyer with near-infinite demand and a decade-long horizon would fund the smallest of the three. Flexibility is not competing with a transmission line on price per megawatt. It is competing with waiting. Scale is the catch. ## What the critics see The skepticism is not about the physics. It is about the accounting, and it is worth taking seriously. PG&E and Google both insist that SHARE has nothing to do with offsetting the costs of Google's data centers, even as the two remain in a dispute with the Sierra Club and The Utility Reform Network over how much of the transmission cost for the San Jose project should fall on Google rather than on ordinary ratepayers. The awkward overlap is that SHARE is explicitly designed to unlock transmission capacity in the same region. Critics read the structure as a company buying a small amount of goodwill while a larger cost question is settled elsewhere. There is a physical version of the same worry. If the capacity a virtual power plant frees up is counted toward a transmission constraint, the grid operator needs certainty that the devices will respond when called โ€” and that the same megawatts are not counted in two markets at once. One analyst quoted by Canary Media put the scale of the untapped resource bluntly: there are gigawatt-hours of behind-the-meter batteries already installed in PG&E's territory. The question is how much of that can be relied upon, not how much exists. โš ๏ธ **The caveat that follows every VPP** Virtual capacity is only as firm as its enrolment. A program built on voluntary household participation can shrink if incentives lose their appeal, if devices go offline, or if a heat wave makes every home want the same comfort at once. Until SHARE delivers measured capacity through a full California summer, its headline numbers describe a pipeline, not a guarantee. ## Why operators are watching closely For utilities, SHARE is a test of a procurement model that does not require a rate case for new steel. For the aggregators โ€” Tesla, Sunrun, Renew Home, Voltus โ€” it is evidence that their installed base has a second buyer beyond the homeowner. For the data center developers now negotiating their own power, it is a template: if you cannot get a line built fast enough, you can pay the neighbourhood to use less. As we wrote in September, the demand side of this problem was already being sold as software, with [Emerald AI pitching data center operators](https://nexi.fund/emerald-ai-demand-response-grid-2026/) on shedding load instead of building for the peak. SHARE turns the same logic toward homes. The unifying idea is that the cheapest power plant is often the demand you never have to serve. The financial logic runs in one direction for now. Google is spending on flexibility because the alternative โ€” waiting, or building generation it would rather not own โ€” is more expensive. Whether that calculus holds at ten times the scale is the open question the pilot is designed to inform. ## Signals to watch ๐Ÿ“Š **Key signals for tracking this program** Whether SHARE's first findings, due late 2026 or early 2027, report verified megawatts rather than enrolled device counts. Whether the California Public Utilities Commission resolves the cost-allocation dispute over Google's San Jose data center, and on what terms. Whether the 16-gigawatt Sunrun-Tesla-Renew Home plan converts into signed capacity contracts with utilities or technology buyers. Whether Google's next flexibility deal is structured as a purchase, as with Voltus, or as a program it funds outright, as with SHARE. The honest framing is narrow. SHARE is a proof of concept, sized deliberately small, aimed at one constrained corner of one grid. It will not power a data center, and no one at PG&E says it will. What it can settle is a question that now sits in front of every large electricity buyer: when the grid cannot be built fast enough, who pays to make the demand underneath it more flexible? Google has volunteered an answer. The rest of the industry is watching to see whether the arithmetic survives contact with a hot afternoon. [ PG&E, Rewiring America and Google launch first-of-its-kind virtual power plant The primary disclosure of SHARE โ€” its partners, the enrolment target and the fact that Google funds the program in full. PG&E Corporation ](https://investor.pgecorp.com/news-events/press-releases/press-release-details/2026/PGE-Rewiring-America-and-Google-Launch-First-of-its-Kind-Virtual-Power-Plant-to-Help-Lower-Costs-for-All-Customers/default.aspx?ref=nexi.fund) The source document for the program's scope and structure. [ Should we get excited about California's latest virtual power plant? Canary Media's skeptical read on SHARE, including the reported phase-one figure and the dispute over data center cost allocation. Canary Media ](https://www.canarymedia.com/?ref=nexi.fund) The most useful counterweight to the launch framing. [ Virtual power plants rise to tackle the data center challenge Reuters on the wider wave of VPP deals serving data centers, including Google's earlier agreement with Voltus. Reuters ](https://www.reuters.com/business/energy/virtual-power-plants-rise-tackle-data-center-challenge--reeii-2026-07-21?ref=nexi.fund) Places SHARE in the broader pattern of tech-funded flexibility.