California’s attempt to make AI data centers pay more of their true power costs collapsed into a weak law that only orders a CPUC study due in 2027, well after the next big build decisions.
Earlier versions would have set separate power rates, required big batteries, and pushed data centers toward 100% carbon-free power by 2030, but Big Tech lobbying and Newsom’s competitiveness concerns stripped all real teeth.
Meanwhile, developers have requested 18.7 GW of capacity for California data centers, more than enough to power every household in the state, raising major questions about who funds grid upgrades and whether residential and small-business ratepayers end up subsidizing AI loads.
Industry groups argue that tougher rules will send hyperscale and GPU-heavy builds to cheaper, laxer states like Texas, but critics note California’s AI talent base is decoupled from where the actual facilities land.
The veto of water-usage disclosure and the stalling of electricity-use transparency bills show that, for now, hyperscalers and cloud vendors have wide latitude to expand GPU farms without detailed public accountability on power and water.
Lawmakers plan to try again with proposals on long-term grid cost allocation and mandatory energy disclosure, potentially trading faster permitting for data centers shouldering more of the bill.
The piece is worth a read to understand how energy, regulation, and Big Tech lobbying are shaping where and how the next wave of AI infrastructure gets built in California.
Source: Big Tech blocks California data center rules, leaving only a study requirement | LAist
