Public Utility Regulation

This white paper calls for building "vertically-integrated public utilities" that provide full-stack AI services comparable to OpenAI or Anthropic, but operated as public infrastructure. Their framework centers on three features: public access (universally available for free or at-cost), public accountability (ultimate control rests with the public through governance mechanisms), and permanent public goods (sustainably funded to prevent private capture). They propose that public AI utilities would "set a dependable price and quality floor" that stabilizes economic activity by ensuring users can safely depend on AI services without fear of price fluctuations, service degradations, or outages that could cause cascading failures throughout society. The paper emphasizes that public utilities can tackle problems underserved by markets driven by quarterly returns, such as moonshot efforts in healthcare, education, housing, and clean energy, while creating meaningful public sector jobs and training a generation of technologists in serving the public interest rather than maximizing profitability.

This report examines how state Public Utility Commissions (PUCs) are emerging as influential AI regulators through their oversight of utilities serving data centers. This indirect regulatory pathway allows PUCs to shape where, how, and whether AI infrastructure gets built through decisions about power allocation, pricing, and infrastructure upgrades, with growing tension between AI companies seeking massive buildouts and advocacy groups concerned about grid stability, energy costs for other customers, and environmental impacts.

The authors argue that cloud computing infrastructure should be regulated as a public utility by state PUCs, just like electricity or water. Drawing parallels to 19th-century railroads that operated as price-discriminating monopolies before regulation, they document how the "Big Three" cloud providers (AWS, Microsoft Azure, Google Cloud) control two-thirds of the global market and engage in opaque pricing practices, including exclusive deals with AI labs like the Microsoft-OpenAI and Amazon-Anthropic partnerships that provide massive compute at steep discounts while raising barriers for competitors. They claim that PUCs enforce pricing transparency, non-discrimination requirements, and interoperability standards, arguing that unlike personalized services such as social media or search (which have failed legal tests for common carrier status), cloud compute is "highly quantifiable and meterable" and offers neutral, uniform service to all customers.