California voters heading to the polls this week will encounter Proposition 45, a measure whose bland title—"Essential Infrastructure Streamlining Act"—belies what environmental advocates call a potentially massive carve-out for the artificial intelligence industry's most resource-intensive assets.
The proposition promises to shorten timelines for permitting "critical infrastructure" projects, a category its drafters left conspicuously elastic. Hospitals, water treatment plants, and electrical grid upgrades all qualify under the measure's explicit language. But buried in the definitional fine print is a clause encompassing "digital infrastructure essential to economic competitiveness"—language that data center developers have already signaled they intend to exploit.
The quiet lobbying campaign
Prop 45 arrived on the ballot through a signature-gathering effort funded primarily by construction unions and a coalition of housing advocates frustrated by California's notoriously glacial permitting process. What received less attention was the late-stage involvement of several technology trade associations, which contributed significant sums to the campaign in its final months. Their interest was not altruistic.
Data centers—the sprawling, energy-devouring facilities that power everything from cloud computing to large language model training—have become increasingly difficult to site in California. The state's environmental review process, governed by the California Environmental Quality Act, routinely adds years to project timelines. For an industry racing to deploy AI infrastructure before rivals, those delays translate directly into competitive disadvantage.
What the loophole enables
If Prop 45 passes and courts interpret "digital infrastructure" broadly, data center developers could bypass the most onerous CEQA requirements, including full environmental impact reports. Projects that currently face three-to-five-year review cycles might instead proceed under expedited categorical exemptions designed for genuinely urgent public works.
The implications extend beyond permitting speed. CEQA reviews often surface concerns about water usage, grid strain, and local air quality—all areas where data centers have drawn scrutiny. A facility training frontier AI models can consume as much electricity as a small city and require millions of gallons of water annually for cooling. Shortcutting review means shortcutting public disclosure of those impacts.
Our take
California's permitting dysfunction is real, and the state's housing crisis alone justifies reform. But Prop 45's drafters either failed to anticipate or deliberately enabled a loophole that could let the AI industry's most environmentally consequential projects slip through with minimal scrutiny. Voters deserve to know that a "yes" vote on infrastructure streamlining may also be a "yes" on fast-tracking facilities that strain the grid and drain aquifers. The measure's backers have not been forthcoming about that trade-off, and the silence is telling.
