Tag: CPUC

  • Up to $1.8B in Grid Upgrades: California’s AI Data Center Deal Is About Who Pays First

    Up to $1.8B in Grid Upgrades: California’s AI Data Center Deal Is About Who Pays First

    TL;DR · 30-second read

    The Short Version

    California lawmakers have agreed on new rules for data centers, the giant warehouses of computers that run artificial intelligence and everyday online services.

    The central idea: companies building them would pay upfront for the power lines and grid upgrades their electricity demands, instead of spreading that cost across everyone’s utility bills. State grid planners estimate those upgrades in one major utility’s territory could cost up to $1.8 billion.

    The deal comes as nearly three in four Californians say they oppose data centers in their own communities, and several cities have banned or paused them.

    California lawmakers reached a compromise on Friday, Aug. 28, on legislation regulating the electricity use of data centers, the Los Angeles Times reported. The package, from Sen. Steve Padilla (D-Chula Vista) and Assemblymember Rick Chavez Zbur (D-Los Angeles), directs the California Public Utilities Commission (CPUC) to create special rates and updated rules for data centers, including how the cost of new power and infrastructure upgrades is charged.

    The final package includes the approach from Padilla’s SB 886, sponsored by The Utility Reform Network (TURN), requiring data centers to pay up front for broader grid upgrades needed to serve their demand. TURN cited a California grid operator transmission plan projecting up to $1.8 billion in upgrade costs, including transmission lines, driven by data center demand in PG&E’s service territory, where most current and proposed California data centers are concentrated.

    Executive Summary

    California has chosen a cost-allocation answer to data center growth rather than a moratorium. The deal does not cap how many facilities can be built. It changes who pays for the grid capacity they need, and when. Under the upfront-payment approach carried over from SB 886, a data center’s share of transmission and distribution upgrades becomes part of the developer’s capital budget instead of a cost recovered later from the general customer base.

    That matters because the numbers are no longer small. The California Energy Commission expects data center electricity use, now about 2% of statewide demand, to double within ten years, and the grid operator’s plan puts potential data-center-driven upgrades in PG&E territory at up to $1.8 billion. Much of the practical detail, including rate design, thresholds and timing, is left to the CPUC. Meanwhile, a provision allowing utilities to strike their own agreements with data centers before the CPUC finishes its rules creates an interim window that developers and critics will both watch closely.

    For AI infrastructure planners, the signal is mixed. The state has set a clear principle that growth pays for itself, which may reduce political risk over time. It has also added cost in a market where high power prices, scarce land and limits on gas-fired generators already hold most projects below 100 megawatts.

    The Fight Was Never About Whether, But Who Pays First

    The core dispute in Sacramento was not whether California should host data centers. It was how much they should pay for power and grid infrastructure, and whether the Legislature or the CPUC should decide. The two options on the table amount to two different financing models. When a utility builds new substations and transmission lines to serve a large new customer, the cost can either be added to the utility’s rate base and recovered over years from all customers, or charged to the customer that triggered it before the work proceeds. SB 886’s approach, which made it into the final package, leans toward the second model.

    The scale comes from the grid operator’s transmission plan, which projects up to $1.8 billion in upgrade costs from data center demand in PG&E’s territory. Moving even a meaningful portion of that onto project budgets changes a developer’s math in three ways. More capital is needed before a building is energized. Grid upgrades shift from a utility’s problem to a line item that competes with land and equipment. And a site’s value depends more heavily on how much grid headroom already exists nearby. TURN’s Matthew Freedman framed the result as keeping data center costs from “being foisted on other customers.” PG&E, which favored a less stringent approach, warned that SB 886 would “risk higher costs for customers and delay critical infrastructure,” which is a reminder that upfront payment also affects how quickly utilities can build ahead of demand.

    The groups most directly affected are developers targeting PG&E territory, which covers most of the state’s data center pipeline including the San José market, and household ratepayers, whose exposure to data-center-driven upgrades is what the deal is designed to limit. How much protection they actually get depends on rate design the CPUC has not yet written.

    A Market Already Built Small

    California enters this regime with an unusual profile. Its data centers are typically under 100 megawatts, well short of the 500-megawatt-plus AI campuses rising in other states, because of high electricity costs and state limits on gas-powered generators. The 49.5-megawatt facility under construction in Vernon is representative. For operators already constrained this way, an upfront grid charge adds to a cost stack that the Data Center Coalition’s Khara Boender says is uncompetitive: she noted that dozens of states offer data center exemptions and California does not, and called added regulation “another signal that the state is a more challenging place for data center development.”

    That argument deserves a fair hearing, and so does the counterpoint. The Coalition opposed both bills for “singling out” one type of power user. Yet the premise of cost-causation pricing is that the customer creating a large, identifiable load pays for it, and data centers are unusually concentrated loads. The strongest case on the industry’s side is not that the principle is wrong. It is that the details, such as how “broader” upgrades are apportioned among several projects on a shared line, could make early movers pay for capacity later arrivals use. That question now sits with the CPUC.

    The Interim Window and the Self-Supply Gap

    Two late amendments drew criticism from Monica Embrey of the Affordable Energy Campaign. The first allows a utility to reach its own agreement with a data center in the period before the state finalizes its rules. In practice, that creates a window in which terms are negotiated bilaterally rather than set by tariff, and projects that sign during it may lock in arrangements that differ from what the CPUC ultimately adopts. The second is the absence of clean energy requirements for data centers that generate their own power.

    Together these point to where AI campus developers are likely to focus next. If grid-supplied power carries upfront upgrade costs while self-supplied power carries no clean energy obligation, on-site or behind-the-meter supply becomes relatively more attractive, within the limits California already places on gas generators. Whether that tilt is intended, and how regulators will treat it, is one of the more consequential open questions in the package.

    Local Opposition Sets the Ceiling

    State rate rules do not settle the land-use fight playing out city by city. A Public Policy Institute of California poll in July found 73% of residents oppose data centers in their own communities. Monterey Park voters banned them outright in June, at least four other San Gabriel Valley cities have adopted moratoriums, Coachella imposed a permanent ban, and Tulare County adopted a moratorium in August. In San José, the state’s main development hub, residents have pushed for a pause while the city updates its standards.

    Water sits on a parallel track. Newsom vetoed a water-disclosure bill from Assemblymember Diane Papan last year, citing “rigid” reporting requirements, but lawmakers have since approved separate bills requiring energy and water use disclosure. For developers, the combined effect is that a California site now has to clear a cost-allocation test at the utility, a disclosure regime at the state level, and a local political test that some communities have already answered with a no.

    Background

    California has not experienced the wave of very large AI data centers seen in states such as Texas. High power prices, limited land and restrictions on gas-fired backup and primary generation have kept most facilities comparatively small, with San José and the wider PG&E service territory hosting most activity. The CPUC, whose board is appointed by the governor, regulates investor-owned utilities such as PG&E and sets the rates they charge.

    Data center growth has nonetheless become a political issue in the state. Community opposition has produced a local ban in Monterey Park, moratoriums across the San Gabriel Valley, the Coachella Valley and Tulare County, and pushback against a proposed 75-acre project in Imperial Valley. In 2025, Gov. Gavin Newsom vetoed a water-disclosure bill, citing concern about rigid requirements on critical digital infrastructure, which left energy cost allocation as the main legislative battleground in the 2026 session.

    Sources

    Source: California lawmakers reach deal in high-stakes fight over regulating data centers (Los Angeles Times), on the legislative compromise directing the CPUC to set data center electricity rates and upfront grid-upgrade charges.