113 entrieslast append amendments 0register open

dcentralmind

A public register on decentralization: why power should move from the few to the many
← register
№ 43 · appended

Google is paying for PG&E's new virtual power plant, which enrols nearly 21,000 devices customers already own

A virtual power plant’s politics are decided by who pays for it and who is owed the value it creates. Most Californian VPPs of any size have been tariffed programmes, which means the terms on which a household lends out its battery are set in a CPUC proceeding that any customer, ratepayer advocate or community group can intervene in. This one is bankrolled by a hyperscaler, so there is no docket, and the reason there is no docket is precisely that no ratepayer money is involved. The capacity still moves to the edge; the argument about its price moves the other way, into a private contract the customer cannot argue with. If buying the edge directly proves faster than approving a programme, participation stops being a public question and becomes a procurement decision made by whoever has the balance sheet. The test is 2028: renewal makes this a template other load-heavy companies copy, and non-renewal tells 21,000 households that what they were paid for their flexibility was a pilot budget.

A hyperscaler, not a tariff, is paying households in two Bay Area counties to lend out the batteries and thermostats they already own. That changes who the household is a party to, and it does it without a regulatory docket.

On 3 September 2026 PG&E, Rewiring America and Google announced a virtual power plant called SHARE, for Smart Home Assets for Reliability and Efficiency, covering Santa Clara and Alameda counties in California. PG&E's release says the programme will draw on "nearly 21,000 existing flexible energy devices" its customers already own, including Tesla and Sunrun batteries and smart thermostats operated by Renew Home. Rewiring America adds that nearly 500 further eligible households will be offered new equipment, including battery-enabled heat pump systems supplied through Carrier and whole-home batteries.

The funding is the part worth reading twice. PG&E's release states that the programme "will be fully funded by Google to support customer incentives, technology deployment, and program delivery". Rewiring America puts it more plainly, saying Google is covering the cost of the upgrades and of enrolment so that nothing is passed on to ratepayers. The programme is expected to begin supporting the grid in autumn 2026 and to run through 2027, with initial findings due in late 2026 or early 2027.

Operationally it remains a utility programme. PG&E builds and operates SHARE, issuing dispatch instructions through its distributed energy resource management system and measuring performance, with Demand Side Analytics handling orchestration. Google's contribution is the balance sheet.

The numbers are not in the releases

Neither release states a capacity target. pv magazine USA reports that PG&E and Rewiring America expect the programme to reach approximately 12 MW by 2028, counting the new battery-enabled heat pumps alongside the existing devices; Latitude Media carries the same figure. Twelve megawatts against 21,000 devices is a little over half a kilowatt each, which is a realistic number for thermostats and partially committed home batteries rather than a headline one, and it makes clear that the point of this pilot is not the megawatts.

The incentives are similarly small in aggregate and specific in structure. pv magazine reports Google-funded discounts of at least $5,000 on a new Carrier heat pump system, rising to $10,000 for the first 25 households only, a detail that matters because the $10,000 figure has been reported elsewhere as though it were the general offer. Owners of devices they already have are to receive ongoing incentives paid separately from their electricity bills, with amounts varying by technology; for scale, pv magazine notes that Powerwall owners in PG&E's existing programme are paid $2.00 for each additional kilowatt-hour dispatched, to a ceiling of roughly $350 a year.

Why it matters

A virtual power plant's politics are decided by who pays for it and who is owed the value it creates. Most VPPs of any size in California have been tariffed programmes: the utility proposes, the California Public Utilities Commission approves, the cost sits in rates, and any customer, ratepayer advocate or community organisation can intervene in the proceeding and argue about the terms. SHARE routes around that entirely. Neither release mentions the CPUC, and the reason it can avoid mentioning it is precisely that no ratepayer money is involved. A bilateral arrangement between a utility and a company with a large balance sheet does not need a docket.

This is not automatically worse. Ratepayer-funded programmes are slow, and slowness is a large part of why flexible capacity sitting in tens of thousands of garages and hallways has stayed mostly idle. But it changes what enrolment is. When the payment comes from a tariff, the household's terms are a public document and the household is structurally a party to how they were set. When the payment comes from a sponsor, the terms are a commercial contract and the household is a supplier. The capacity still moves to the edge. The argument about its price moves in the opposite direction, into a room the customer cannot enter.

The second thing to watch is 2028. Google's commitment runs through 2027. If it renews, direct corporate procurement of household flexibility becomes a template other load-heavy companies can copy, and the CPUC's role in pricing demand-side capacity quietly narrows by however much of the market moves outside it. If it does not renew, 21,000 households learn that what they were being paid for their flexibility was a pilot budget rather than a market.

One inference is tempting and should be resisted on the current evidence. Latitude Media reports that capacity from the pilot is not currently tied to any existing or planned data centre in the San Jose region. Google's stated framing, from its global head of energy and power Amanda Peterson Corio, is about optimising the existing grid as demand scales. The company has not linked SHARE to its own load, and no source consulted does so either.

What is still unknown

  • Where the value settles. PG&E issues the dispatch instructions, but neither release says who captures the wholesale energy and resource adequacy value the devices produce, or whether Google receives anything contractual in return for funding the programme. [NEEDS DATA: the value stack and the counterparty]
  • How much of the capacity is incremental. Many of the 21,000 devices are presumably already enrolled in existing PG&E, Sunrun or Tesla demand response programmes. No source states how much of the 12 MW is new. [NEEDS DATA: currently enrolled DER capacity in Santa Clara and Alameda counties]
  • Whether SHARE required any CPUC authorisation, or whether using no ratepayer money places it outside the tariffed-programme process altogether. [UNVERIFIED: no source consulted addresses the question either way.]
  • The incentive schedule by device type for existing devices, and whether payments are per event or periodic.
  • The two announcements do not list the same partners. PG&E names encoord; Rewiring America names Olivine. [UNVERIFIED: whether both firms are involved and each release is partial, or one is in error.]
  • What happens to enrolled customers and their incentives when the funding commitment ends after 2027.

Sources

Recomputes the content hash in your browser and compares it with the stamp.

Public comments

Loading…