Will AI data centers raise California electricity rates?
California’s energy regulator reports all data centers — not only AI facilities — at about 1,000 MW, or 2% of California ISO peak demand in early 2026, forecast to reach 9% by 2040. No published statewide analysis isolates a residential-bill increase attributable specifically to AI. What the data does show is where the cost risk sits, and it is not where most coverage points.
Quick answer
Short answer: not measurably, not yet — but the risk is real and it is about who pays for the grid, not about how much power the servers use. As of August 2026 no published statewide analysis has measured an AI-specific increase in California residential bills, and the CPUC does not name data centers among the drivers of the past decade’s rate increases. Upward pressure appears when grid investment made for large new loads ends up spread across existing customers — above all when a project is delayed or never gets built. That is a question of cost allocation, and California has not finished settling it.
- The CEC figure covers all data centers, not AI specifically. No published dataset separates AI workloads from the rest.
- About 1,000 MW, or 2% of California ISO peak demand, in early 2026, forecast to 4,500 MW / 9% by 2040.
- Data centers are not named among the primary drivers of the decade-long rate increases. The CPUC names wildfire costs, transmission and distribution investment, and rooftop solar incentives.
- That is not the same as zero effect. A share of peak demand is not a share of costs, and no published analysis quantifies the current bill impact either way.
- Two distinct cost risks exist: infrastructure built for projects that never materialize, and genuine new demand requiring generation and network investment.
The question, answered line by line
Each row states what the published record supports as of August 2026.
| Question | What the published record shows | Answer |
|---|---|---|
| Are AI data centers raising California bills right now? | No statewide analysis isolates an AI-specific bill effect; the CPUC names wildfire costs, transmission and distribution investment, and rooftop solar incentives as the drivers of recent increases | Not measurably |
| Could they push rates up in future? | New large loads can require generation and network investment; if that cost is spread across all customers, existing ratepayers carry part of it | Yes — conditionally |
| Which risk is larger? | Infrastructure built for projects that are later delayed or cancelled leaves the cost without the offsetting load — a bigger exposure than the electricity the servers actually consume | Stranded cost, not consumption |
| How much of California’s peak demand is this? | All data centers together: about 1,000 MW, roughly 2% of California ISO peak demand in early 2026; forecast to 4,500 MW and about 9% by 2040 | 2% today, 9% forecast |
| Does the law already make developers pay their own way? | Measures assigning long-term grid costs to data center developers have been advancing through the legislature, but the framework is not settled | Not yet fully |
| Can a household do anything about it? | Rooftop generation reduces the volume of grid electricity a household buys, which limits exposure to whatever the volumetric rate becomes | Partly — by using less grid supply |
What data centers actually draw today
Start with the number, because most coverage skips it.
The California Energy Commission puts current data-center load at about 1,000 megawatts, or 2 percent of the California Independent System Operator’s peak electricity demand. The CEC began folding data-center growth into its demand forecasts in 2024, using the applications developers actually submit to utilities, and the figures feed the Integrated Energy Policy Report.
| Period | Load | Share of CAISO peak |
|---|---|---|
| Early 2026 | ~1,000 MW | 2% |
| Forecast, 2040 | 4,500 MW | 9% |
Source: California Energy Commission — Data Centers, figures from the CEC Planning Forecast. Verified August 17, 2026.
Set those numbers against the bill history carefully. Residential rates at the three large investor-owned utilities rose 69% at PG&E, 97% at SDG&E and 101% at SCE between January 2016 and June 2026. The CPUC names the primary statewide drivers as wildfire mitigation and liability costs, transmission and distribution investment, and rooftop solar incentives. Data centers are not on that list.
But two cautions belong here, and most coverage skips both. A share of peak demand is not a share of costs — megawatts at peak do not translate directly into kilowatt-hours, revenue requirements or bill impact. And the CEC figure is for all data centers, a category that long predates the current AI build-out.
The honest position: published statewide data does not isolate a residential-bill increase attributable specifically to AI, and data centers are not identified as a primary driver of the decade-long increase — but their current bill effect has not been quantified either. Anyone asserting a precise number in either direction is going beyond the evidence.
Why one facility matters more than 2% suggests
The share is small; the individual units are not. The CPUC’s Public Advocates Office notes that developers in California increasingly plan facilities with loads exceeding 50 to 100 megawatts — roughly the residential energy use of a city the size of Santa Rosa or Huntington Beach.
A single customer request the size of a mid-sized city does not slot into an existing grid. It triggers infrastructure:
- Individual interconnection projects run from a few million dollars to more than $100 million.
- In the South Bay, CAISO approved transmission upgrades costing more than $2 billion, largely to serve 2.5 gigawatts of concentrated data-center and electrification load growth between 2026 and 2039.
- CAISO has identified billions of dollars in transmission upgrades needed over the next decade.
Source: Public Advocates Office at the CPUC — How Will Data Center Growth Impact California Ratepayers? (October 27, 2025). Verified August 17, 2026.
Two cost risks, not one
Stranded infrastructure is the one that gets least attention. It is not the only one.
Here is the mechanism, in the Public Advocates Office’s own words. Under current FERC rules, transmission costs are “socialized across all ratepayers within the CAISO territory.” If data-center projects are cancelled or fail to reach the load they promised, “the significant costs that utilities incurred to serve these data centers will be passed on to existing ratepayers.”
Read that carefully, because it inverts the usual complaint. The danger is not a data center that runs at full tilt and buys a lot of electricity — that customer contributes revenue. The danger is the grid getting built for a facility that arrives late, arrives smaller, or never arrives at all. The steel is in the ground either way, and the bill for it is shared.
This is also why the 2% figure is not reassuring on its own: the cost lands when the infrastructure is authorized, not when the load shows up.
The second risk is the opposite case. A data center that does arrive and runs at full load is real demand — it can require new generation, transmission and distribution investment of its own. Both paths can raise costs; they simply fail in different ways. Presenting stranded cost as the only risk would be as one-sided as blaming AI for last decade’s bills.
What California has and has not decided
Not settled. The Public Advocates Office describes PG&E’s proposal to make large-load customers pay actual interconnection costs as “a step forward” that nonetheless “lacks other protections that would effectively mitigate financial risks.” The office has asked for minimum demand charges, early-termination fees on the models used in Ohio and Indiana, and proactive planning that reassesses transmission need when projects stall.
In the legislature. SB 1168 (McNerney), “Data centers: rate structures,” would direct the CPUC to assess rate structures so data centers pay a reasonable share of transmission and distribution costs and a proportionate share of load growth and procurement, while easing pressure on residential ratepayers.
Its status as of August 17, 2026: Active Bill — In Floor Process, sitting in the Assembly 3rd Reading File as item 242. It was ordered to third reading on August 13 and removed from the consent calendar. It is not law. Any page telling you California now requires data centers to pay their way is describing a bill, not a statute.
Bill status read directly from the Legislature’s official bill status page on August 17, 2026. Legislation moves; check the current status before relying on this.
What this means for your bill this year
Nobody has published a figure, and we are not going to invent one. Data centers are not named among the CPUC’s primary drivers of the past decade’s increases, and no statewide analysis isolates a residential-bill effect attributable to them — in either direction. That is the accurate answer, and it is less satisfying than the ones circulating.
What is worth watching is the next few years of transmission spending and who is made to carry it. That gets decided in CPUC proceedings and, possibly, in SB 1168 — not in headlines about AI. To see what is actually queued up in the two largest territories, our record of SCE and PG&E rate filings follows the filings quarter by quarter.
What this page cannot determine
- Whether any specific project will be built. Announced data centers are frequently revised, delayed or cancelled; that uncertainty is the entire point of the stranded-cost problem.
- Your future rate. Rates change through regulatory decisions. Forecasts here are the CEC’s load projections, not price predictions.
- How much of a future increase would be attributable to data centers. Transmission serves many loads at once; separating the share is a regulatory question, not an arithmetic one.
- The outcome of SB 1168 or any CPUC proceeding. We report status and date, nothing more.
Frequently asked
Are AI data centers raising my electricity bill in California right now?
No published statewide analysis isolates that. The CEC reports all data centers at about 1,000 MW, or 2% of California ISO peak demand, in early 2026 — a figure that does not separate AI workloads. Data centers are not named among the CPUC’s primary drivers of the 2016–2026 rate increases, which are wildfire costs, transmission and distribution investment, and rooftop solar incentives. But a share of peak demand is not a share of costs, so their current bill effect has not been quantified in either direction.
How much electricity do California data centers use?
About 1,000 megawatts, or 2 percent of California ISO peak demand, as of early 2026, per the California Energy Commission. The CEC forecasts growth to 4,500 MW, or 9 percent of peak, by 2040.
So what is the actual risk to ratepayers?
Two risks, not one. First, stranded cost: under current FERC rules transmission costs are socialized across all ratepayers in the CAISO territory, and the Public Advocates Office warns that if projects are cancelled or underperform, costs already incurred pass to existing ratepayers. Second, realized demand: a facility that does arrive can require new generation and network investment. Both can raise costs.
How big is one data center compared with a town?
California developers increasingly plan facilities exceeding 50 to 100 MW — roughly the residential energy use of a city the size of Santa Rosa or Huntington Beach, according to the Public Advocates Office. Individual interconnection projects range from a few million dollars to more than $100 million.
Has California passed a law making data centers pay their share?
No. SB 1168 would direct the CPUC to assess rate structures for that purpose, but as of August 17, 2026 it was an active bill in the Assembly 3rd Reading File, not a statute. Check the Legislature’s bill status page for the current position before relying on any summary, including this one.
Does rooftop solar reduce this risk?
Distributed generation and storage reduce load at the point of use, which is generally helpful for local distribution constraints. But the transmission investments discussed here are driven by large concentrated loads and regional planning, so it would be overstating it to present home solar as a solution to that specific problem.
Related reading
Sources & methodology
Figures on this page come from the primary sources below and, where noted, from Cali Energy calculations using the stated assumptions. Rates, incentives, and program terms change; each was verified August 17, 2026.
About this reference
Cali Energy is a licensed California contractor (CSLB #1032379) in Northridge. We publish this page as a reference and date it to the day each figure was checked against its source.
Prepared by Cali Energy, August 17, 2026. This article is for general educational purposes only and is not legal, tax, financial, engineering, or utility advice. Rates, incentives, codes, permit requirements, equipment specifications, prices, and program terms may change; figures and timelines are estimates, not guarantees. Confirm current requirements with the applicable utility, AHJ, program administrator, manufacturer, or a licensed professional. See our Content Disclaimer. Cali Energy · 19201 Parthenia St, Ste E, Northridge, CA 91324 · CSLB #1032379 (B, C-10, C-39)
