California Signed Seven Data Center Laws in One Day. The Price Tag Isn't Written Yet.

On September 21st, Governor Gavin Newsom signed seven data center bills in a single sitting — three shifting grid-infrastructure costs onto operators, two forcing water-use disclosure, one narrowing environmental-review exemptions, and one adding energy-reporting requirements to the California Energy Commission. His office called the package "the most comprehensive data center laws in the nation," and by sheer bill count, that's a defensible claim. Look past the count, though, and most of what actually changed on September 21st is who gets to write the real numbers, and on what calendar — not what those numbers are.

The backdrop explains the timing. A PPIC survey published in July found only about a quarter of California voters favor building new data centers in their state at all, and a separate poll of 850 likely November 2026 voters put opposition to a data center in one's own community at 70 percent. The same poll found that number nearly inverts once conditions are attached: 63 percent support new data centers if they come with guardrails, and majorities as high as 87 percent want operators to pay their own grid-connection costs, cover cancellation risk, and buy or build their own clean power. That's not a population that wants data centers gone. It's one that wants the bill sent to a different address, and Sacramento spent 2026 writing legislation aimed at exactly that address — over sustained lobbying against it, including more than half a million dollars Amazon alone spent pushing back on data center-related bills in the first half of the year.

The reversal on water

The water provisions are worth pausing on because Newsom didn't arrive at them in a straight line. In October 2025, he vetoed AB 93, a narrower bill from Assemblymember Diane Papan that would have simply required data centers to disclose water consumption as part of local business licensing. His veto message called it premature — he was, he wrote, "reluctant to impose rigid reporting requirements about operational details on this sector without understanding the full impact on businesses and the consumers of their technology." Eleven months later he signed two bills that go considerably further than the one he rejected. AB 2469, also from Papan, now bars a city or county from approving a data center permit at all unless the applicant files a water scarcity plan and a water supply assessment, and the applicant — not the water utility, not other ratepayers — eats the full cost of any conveyance, treatment, storage, or distribution upgrade the project requires. AB 2619 requires every data center in the state to file water-use estimates and, later, actual annual use, under penalty of perjury; the largest operators, defined as more than 10,000 servers or more than 25 megawatts of draw, have to hand their water utility a good-faith use estimate before they can even apply for a business license.

What changed in eleven months wasn't the underlying facts about data center water use. It was the polling, the volume of local opposition, and the fact that neighboring states had stopped waiting. Voters didn't need to be told cooling systems use water; they'd already noticed, and enough of them said so that the disclosure bill Newsom called premature in 2025 looks, in hindsight, like the floor rather than the ceiling.

Who pays for the grid — eventually

The cost-allocation bills are the ones with real financial weight, and they're also the ones where the September 21st signing did the least immediate work. SB 886 directs the California Public Utilities Commission to adopt new data center tariffs — covering interconnection, transmission, distribution, and generation — by January 1, 2028, with developers on the hook for transmission facility upgrades, including shared network upgrades where federal law allows it, and a ten-year early-termination fee if a project walks away from its power contract. AB 2383 requires data centers to meet the state's clean-energy procurement standards and pay the incremental cost of any generation built to serve them, for at least ten years. SB 1168 sends the CPUC back to the same table to write rate structures ensuring data centers, not the broader ratepayer base, absorb the cost of the transmission and distribution upgrades their load growth triggers. The number that got these bills moving in the first place: CAISO data cited by The Utility Reform Network put potential transmission-upgrade costs from data center demand growth in PG&E's territory alone at up to $1.8 billion, a bill that, absent these laws, would have landed on every residential and commercial ratepayer on the system.

None of those three bills sets a dollar figure. They set a deadline for the CPUC to set a dollar figure — January 1, 2028 for the tariff structure itself, with the new framework applying to transmission interconnection agreements signed on or after January 1, 2027. That's a full year where the statute is live but the number a specific developer would actually owe doesn't exist yet, and a further year after that before the tariff structure has to be finalized. The lobbying that couldn't stop these bills in the legislature doesn't disappear on signing — it moves into a CPUC rulemaking docket, a much quieter room, for the next two years.

The one bill that bites immediately

SB 887 is the exception, and it's the bill developers should actually be tracking for near-term timeline risk. It creates a statewide legal definition of "data center" and makes most private data center projects ineligible for CEQA's categorical exemptions — the fast-track status that lets projects presumed to have no significant environmental effect skip a full environmental review. It also opens a path to expedited judicial review for projects that meet additional conditions, a partial offset but not a full one. Unlike the cost bills, SB 887 doesn't wait on a rulemaking. It takes effect January 1, 2027 and immediately changes how long environmental review takes for a project that previously could have claimed an exemption — which, for anyone underwriting a construction schedule in California, is a bigger near-term variable than a tariff that won't have a number attached for another year past that.

Reading it from outside California

GridPermit works Colorado, not California, so none of these seven statutes directly touch a site we're evaluating. The pattern behind them is the one worth watching from anywhere. In a Kill Test, we grade every claim on an evidence ladder specifically so a state passing a law that promises future cost allocation doesn't get treated with the same confidence as a utility's actual tariff filing with an actual number in it. California just gave a clean, current example of why that distinction matters: as of September 21st, a developer can accurately say California now requires data centers to pay their own grid-upgrade costs, and just as accurately say nobody, including the CPUC, can yet tell them what that payment will be. Both statements are true, and only one of them is usable for underwriting. It also puts California alongside Texas's refundable interconnection deposit and the still-stalled federal Ratepayer Protection Act as a third, separate answer to the same question — who pays for a data center's grid footprint — arrived at on three different timelines and enforced, if at all, by three different bodies.

Colorado hasn't passed anything like SB 886 or SB 887, but Xcel Energy's own large-load tariff filings already carry collateral and minimum-bill terms that do some of the same work by contract rather than by statute, and the CEQA exemption fight is a preview of the kind of environmental-review carve-out states with their own version of CEQA-style review could eventually take up. The lesson isn't that California's laws will show up in Colorado verbatim. It's that a signed statute promising a future number is a different evidentiary rung than the number itself, and the gap between the two — in California's case, a year or more — is exactly the gap a diligence process needs to name before a client prices a site as if the number already exists.