As the AI industry’s computing power race intensifies, data centers are becoming the new “not-in-my-backyard” facilities. On September 21 local time, California Governor Gavin Newsom signed seven billswhose core provision is a single sentence:Electricity and water costs for AI data centers must not be passed on to ordinary residents.This is the first comprehensive legislative package in the United States to systematically constrain energy and water consumption by AI data centers—and a forceful move by California to tighten regulation amid federal deregulation.
What Exactly Does the New Regulation Cover
Key Provisions of This Bill Include:
- New Rate ClassificationRequires the California Public Utilities Commission (CPUC) to establish a separate rate category for data centers, preventing them from subsidizing low-cost residential electricity rates.
- Grid and Water System Upgrade FundingData centers must fund the expansion and upgrades of local grid and water systems, rather than shifting costs to all taxpayers.
- Mandatory DisclosureProposed data centers must disclose estimated water usage, energy efficiency metrics, and drought emergency response plans to local governments.
- Approval ThresholdsOnly projects meeting specified energy, water, and fuel consumption standards qualify for streamlined permitting.
In his statement, Newsom directly criticized the federal government: “As the Trump administration moves toward deregulation, communities bear the consequences—higher electricity demand, grid strain, water use, and pollution.” This message also aligns with his AI executive order signed last week. “Emergency Kill Switch”Through this executive order, California is attempting to establish a state-level AI governance standard in opposition to Washington’s approach.
Why “Energy and Water”: The Hidden Bill of Compute
Many people focus on AI model parameters, pricing, and benchmark scores, yet rarely consider the underlyingPhysical CostsWhile next-generation AI chips like those from Nvidia Vera Rubin Nvidia and Cursor deliver significantly improved energy efficiency, overall compute demand continues rising exponentially—Anthropic’s compute commitment has surged to $51.7 billionData centers operated by major vendors are expanding at a frenetic pace.
These massive facilities are veritable “electricity hogs” and “water guzzlers”: AI training and inference demand enormous amounts of electricity, while cooling systems consume vast quantities of water. When data centers cluster in specific regions, local power grids and water resources become strained—ultimately reflected in higher electricity bills and water shortages for ordinary households. California’s new regulations essentially reassign this “invisible bill” to its true beneficiaries: data center operators.
How large is this scale? A single large AI data center consumes as much electricity as tens of thousands of average households do in a year; training a cutting-edge large model routinely requires millions of gallons of water. Previously obscured by the glamour of the compute arms race, these figures are now being forced into plain sight by utility bills and legislative measures.
Regulatory ripple effects are spreading.
California is not the only jurisdiction tightening oversight. Just days ago, Virginia’s governor announced the formation of an AI task force and initiated efforts to restrict data center expansion—Virginia is one of the most densely concentrated data center regions in the U.S., dubbed “the heart of the global internet.” Meanwhile, data centers’e-waste problemis also surfacing: computing hardware turns over extremely rapidly, and discarded GPUs, servers, and thermal management equipment are forming a new pollution chain.
It can be said that AI’s breakneck acceleration is bringing three previously overlooked hidden costs—energy, water, and e-waste—onto regulators’ desks simultaneously. California’s seven new bills are merely the beginning; similar constraints are highly likely to emerge across more U.S. states and countries.
What this means for the industry
In the short term, this willraise construction and operational costs for data centers in California, potentially shifting some projects to states with looser regulatory regimes. But in the long term, it establishes a critical principle: AI’s costs cannot be borne solely by the public while profits accrue exclusively to model companies. This stands in sharp contrast toJensen Huang’s assertion that “AI safety is an engineering problem—not requiring new regulation”—while the industry remains divided on whether regulation is needed, California has already spoken through utility bills.
For China’s AI industry, this serves as a mirror: as domestic intelligent computing centers rapidly scale up, energy and water constraints will inevitably move to center stage. In fact, numerous newly planned intelligent computing centers have already relocated due to electricity quota limits or water scarcity. Those who first calculate this “physical bill” accurately will advance more steadily and sustainably in the next phase of the compute race.
Frequently Asked Questions (FAQ)
Why is California suddenly tightening AI data center regulation?
AI data centers consume massive amounts of electricity and water, with part of those costs previously shifted onto ordinary residents and businesses. The new rules require data centers to bear the costs of grid and water system upgrades themselves, preventing public cost-shifting.
Will AI data centers raise residential electricity rates?
Yes. Data centers consume enormous amounts of electricity; without a separate rate classification, their demand may strain the grid and drive up residential electricity prices. California’s new rules specifically introduce such a separate rate to prevent this cost-shifting.
What Impact Do These 7 Bills Have on AI Companies?
In the short term, they will raise the cost of building and operating data centers in California, potentially driving some projects offshore; in the long term, they require AI companies to internalize external costs such as energy and water usage.
Are AI Data Centers Really Water-Intensive?
Yes—cooling systems demand substantial water resources. This is why the new regulations mandate data centers to publicly disclose estimated water consumption and drought contingency plans.
Will China Introduce Similar AI Data Center Regulations?
Domestic intelligent computing centers already face constraints on power allocation and water resources; some projects have relocated due to energy shortages, making it highly likely that similar policies governing energy consumption and water usage will emerge.
Want to Discover More Useful AI Tools? Explore Our AI Model Library and Tool Comparison Engineor continue reading:Nvidia Vera Rubin Energy Efficiency Analysis · Anthropic’s Compute Bet · All Reviews.