FUTURE VISION v1.0:
AI AS PUBLIC UTILITY
Rethinking AI as civic infrastructure. Moving from a private, profit-driven AI ecosystem to a future where AI is an inclusive infrastructure akin to public education, libraries, or utilities like water and electricity.
What is public utility?
Public utilities are essential services provided to the general public that are necessary for everyday life and economic activity. These services are typically characterised by their fundamental importance to society and are often subject to government regulation or direct government provision.
Think of:
Electricity - Power generation and distribution to homes and businesses
Water and Sewer Services - Clean water supply and wastewater management
Natural Gas - Gas distribution for heating and cooking
Telecommunications - Telephone and internet services
Public Transportation - Bus, rail, and subway systems
Because public utilities often are subject to government regulation. Regulatory bodies oversee pricing, service quality, and accessibility to protect consumers from exploitation and ensure equitable access. In some cases, utilities are owned and operated directly by government entities, while in others, they are private companies operating under strict regulatory frameworks.
The goal of public utility regulation is to balance the need for reliable, affordable service with the financial sustainability of the utility provider, ensuring that these vital services remain accessible to all members of society.
Why AI Fits This Model
AI fits perfectly into the public-utility model for three reasons:
It’s essential.
AI systems will make decisions that profoundly affect individuals' lives. A public utility framework could guarantee that AI's benefits reach underserved communities, smaller businesses, researchers, and nonprofit organisations.It’s resource intensive.
Building data centres and training large models require billions in upfront investment, and it draws from natural resources.It’s highly concentrated.
Only a few companies (Microsoft, Google, Amazon, Meta) have enough resources to build and operate large-scale AI systems.
Smaller players, startups, and universities are locked out — not for lack of ideas, but because they can’t afford the compute.It driven by market forces.
Currently, AI compute largely flows toward whatever generates maximum short-term revenue. Commercial applications often prioritise entertainment, advertising optimization, and consumer conveniences over pressing social needs. Strong regulation of compute could mandate that a percentage of commercial AI infrastructure be reserved for non-profit research, or build dedicated public computing facilities for priority applications.
Learning from history
As cities grew in the 1800s, private companies built utilities—gas works, water systems, and electrical grids. Edison's Pearl Street Station (1882) launched commercial electric power.
By the early 20th century, monopolistic practices and poor service sparked Progressive Era reforms and regulatory oversight. The principle emerged that essential services couldn't be left purely to markets. The U.S. Rural Electrification Act (1936) funded local cooperatives, while Britain nationalised electricity in 1947.This pattern repeated with water, phones, and broadband: societies recognised that universal access drives innovation.
Today, AI faces the same crossroads electricity once did—powerful but concentrated, promising but unequal.
Systemic Perspective
From a systemic design perspective, declaring AI compute a public utility is a goal-level intervention - the highest leverage point in Donella Meadows' hierarchy. It fundamentally redefines what the system optimises for.
The current AI system optimises for shareholder value maximisation, making compute a public utility shifts the goal to collective wellbeing and equitable access.
Additionally, the self-reinforcing growth loop currently operates without bounds. Regulating compute supply introduces purpose-based constraints.
Making AI compute a public utility creates cascading systemic effects: economically, it lowers entry barriers and neutralises cash-burning advantages; politically, civic compute cooperatives counterbalance corporate lobbying; socially, AI dividends link automation gains to human wellbeing; and ecologically, energy and water consumption become actively managed through distributed, resource-efficient infrastructure.
Tiered pricing for compute
Access is allocated and priced by societal value and risk. Projects are classified into tiers using clear criteria – Low prices for public benefit, Standard for SMEs and Citizens, High prices for pure profit plays such as optimised advertising. Revenues from higher-friction tiers help fund AI dividends, worker transition support, and upgrades to public and edge infrastructure.
AI Dividends
A mechanism to redistribute the immense wealth generated by AI. Revenues from high prices of compute on AI systems that replace human jobs will fund AI Dividend paid out to the public or to affected workers and communities.
Compute Passports
A Compute Passport is a permit that any major AI project must obtain, indicating that it has met certain standards – like disclosing the compute resources it will use, the expected energy consumption, and the measures taken to mitigate environmental impact or social risk.
Distributed Edge Data Centres
Instead of relying on a few giant data farms run by global corporations, a network of smaller, distributed “edge data centres.” Each owned by cities, regional utilities, or cooperatives. These distributed data centres, located in housing districts will also re-use heat.
Liam stared at the email. "Redundancy." "Restructuring." After ten years as a software engineer at OmniCorp, he was out—replaced by AI-driven development tools. Anxiety tightened in his stomach. Rent. Student loans. How?
Then he saw the last paragraph: "As per the Public Utility AI Act of 2030, all employees impacted by AI-driven displacement are eligible for the Universal AI Dividend, payable monthly." The AI Dividend—funded by taxes on corporate AI compute usage. The more AI companies integrated, the larger the pool for redistribution.
Days later, a notification appeared. His first dividend payment had arrived. Combined with income insurance, it covered his basic expenses for months. It bought him time—to reskill, explore new passions, volunteer at the community garden without crushing financial pressure. The sting of unemployment remained, but now, a flicker of hope ignited.
Dr. Anya Sharma scrolled through her queue, a small smile forming. A decade ago, during the "Great Healthcare Crunch," the NHS in Neo-London had nearly collapsed. Specialists were overworked, diagnostic scans took weeks to read, patients languished on waiting lists. Anya remembered the crushing fatigue, the impossible ethical dilemmas.
Now, things were different. Thanks to the Public Utility AI Act, every primary care clinic had advanced diagnostic AI. Her tablet displayed a colour-coded list: red for critical, amber for urgent, green for routine. Each entry linked to analysis from "MediScan," the national diagnostic AI that could process an MRI in minutes and suggest treatment pathways with 98% accuracy.
"Next patient, Mrs. Albright," Anya murmured. The AI had flagged her lung scan as amber—suspected early-stage nodule. Before AI, that scan might have sat unread for weeks. Now, Anya had a detailed report before Mrs. Albright even sat down.
The system wasn't perfect. Critical cases still demanded immediate human oversight, and routine check-ups could take a day or two. But agonizing weeks-long waits for crucial diagnoses were gone. Anya could focus on the human connection—explaining findings, offering comfort and reassurance. She was no longer battling overwhelming demand; she was orchestrating a symphony of swift, intelligent care.
Einar Vance closed his laptop. "I'm sorry. We can't invest in AdSense AI."
The young founder's face fell. "But the projections—"
She showed him the red banner: Societal Value Added Score — Negative. Compute Cost Multiplier: ×9.8.
"Under the Public Utility AI Act, manipulative algorithms pay premium compute rates. Nearly ten times standard, plus mandatory audits and harm insurance."
"So it's too expensive to run advertising?"
"It's too expensive to manipulate people. That's the point."
After he left, a note blinked: Public Compute Grid: 92% allocated to Education, Health, and Climate Services.
He exhaled. Somewhere, algorithms were saving lives, reaching students, modeling climate solutions. The math finally felt right.
Lina stared at her heating bill.
Energy Cost: €127. Edge Heat Credit: –€89. Total Due: €38.
The municipal data center down the street ran AI jobs overnight—hospital diagnostics, university research, city planning. Instead of venting the heat, it pumped warmth directly into the neighbourhood's heating pipes.
Her radiator kicked on. Properly hot, not the lukewarm trickle from previous winters.
In the lobby, a small plaque read: "Heated by Block 7 Edge Pod—Computing for Good."
Strange how the future arrived. Her home stayed warm because someone needed medical images processed.
Intelligence had a useful side effect.
