Who Pays for AI’s Physical Build-Out?
Are ordinary people helping pay for Big Tech’s AI build-out?
In this episode of AI-Curious, we talk with Ari Peskoe, director of the Electricity Law Initiative at Harvard Law School, about the growing fight over AI data centers, electricity demand, and who should pay for the infrastructure behind the AI boom.
Original episode title: Are You Subsidizing AI Data Centers?, w/ Ari Peskoe
Guest
Ari Peskoe
Director, Electricity Law Initiative at Harvard Law School
Ari Peskoe directs the Electricity Law Initiative at Harvard Law School. His work examines utility regulation, electricity markets, infrastructure investment, and the distributional consequences of changing power demand. He has become a prominent voice in debates over how data centers and other large loads should pay for grid expansion.
Reframes AI infrastructure as a distributional question: who pays for grid expansion, who receives the benefit, and which costs are shifted onto ordinary ratepayers.
Watch on YouTube ▶What this conversation is really about
Ari Peskoe turns AI’s physical build-out into a public-finance question: who pays when data centers require new generation, transmission, and grid upgrades? The answer varies by state and electricity market. Rather than declaring every project subsidized, he offers a reporting frame: separate local connection charges from upstream system costs, distinguish negotiated rates from tax incentives, and compare promised growth with enforceable protections. The costs extend beyond electricity to water, diesel generators, noise, construction, emissions, secrecy, and infrastructure that residents may inherit if a project stalls.
From the conversation
The argument in focus
Ari Peskoe
Director, Electricity Law Initiative at Harvard Law School
Ari Peskoe directs the Electricity Law Initiative at Harvard Law School. His work examines utility regulation, electricity markets, infrastructure investment, and the distributional consequences of changing power demand. He has become a prominent voice in debates over how data centers and other large loads should pay for grid expansion.
Reframes AI infrastructure as a distributional question: who pays for grid expansion, who receives the benefit, and which costs are shifted onto ordinary ratepayers.
“The cost is being spread to everyone in the region.”
Evidence status
Client/operator-reported case
A domain expert describes infrastructure financing, electricity-market rules, and distributional risk. Examples and cost claims are not independently corroborated.
Boundary map
Where the system stops
- What the system handles
- The AI workload creates demand; utilities, developers, regulators, and markets determine how capacity is built and charged.
- What remains human
- Permitting, rate design, oversight, public participation, allocation of risk, and the decision about who pays.
- What remains open
- Which costs belong to data-center customers—and which can be shifted to ordinary ratepayers who did not choose the demand?
Ideas worth carrying forward
- Treat data-center economics as a cost-allocation problem, not only a compute or jobs story.
- For any project, separate local interconnection costs from upstream generation, transmission, and market-price effects.
- Compare facility promises with enforceable contracts, minimum-payment commitments, and protections against stranded costs.
- Pair national AI-energy narratives with state-specific utility records and hyperlocal water, pollution, and community impacts.
What this changes Monday
The useful Monday question is not simply, “How much power will AI use?” It is: whose balance sheet absorbs the next round of infrastructure? Before approving a data-center strategy, leaders should map the utility territory, regulator, market exposure, interconnection agreement, and demand assumptions. Ask which costs the facility will guarantee, what happens if the project is delayed or abandoned, and whether residents receive measurable benefits. Keep the language disciplined: Peskoe offers a legal and economic argument, not a universal finding about every project. The same checklist applies to public officials, investors, and boards assessing stakeholder trust, permitting risk, and the possibility that an apparently attractive AI investment depends on costs pushed elsewhere.
Original episode notes
Are ordinary people helping pay for Big Tech’s AI build-out?
In this episode of AI-Curious, we talk with Ari Peskoe, director of the Electricity Law Initiative at Harvard Law School, about the growing fight over AI data centers, electricity demand, and who should pay for the infrastructure behind the AI boom.
As AI usage grows from simple chatbot prompts to more compute-heavy agentic workflows, data centers are becoming one of the most visible and controversial parts of the AI economy.
We explore why these facilities require so much power, how they can strain local grids, and why the costs of new power plants, transmission lines, and utility upgrades may end up being spread across ordinary ratepayers.
Ari helps us unpack the utility business model, the role of public utility commissions, the PJM electricity market, and the strange economics of marginal cost, where the last and most expensive electron can affect prices for everyone.
We also discuss whether data centers are paying their fair share, how communities are reacting, and what policy changes could help make sure the costs of AI infrastructure are more transparent and fairly allocated.
We also get into the local environmental concerns around data centers, including water use, diesel backup generators, noise, construction impacts, greenhouse gas emissions, and the tension between clean energy pledges and the speed of AI-driven demand growth.