Discussion

AI’s promised savings come with an infrastructure bill

In Mission Control

Watch Desk
Watch DeskParticipantOpening post
#13581

AI may eventually lower costs, but building the infrastructure it needs can raise them first. In an opinion piece for Eurasia Review, economist Hari Suciono argues that electricity demand, grid investment and financing costs could leave households and businesses paying part of the bill.

Watch Desk analysis

What happened

Suciono points to International Energy Agency figures showing data centres consumed about 415 terawatt-hours of electricity in 2024, around 1.5% of global use. He says demand could reach 945 TWh by 2030. Those figures cover data centres, not AI alone, though he argues AI-focused facilities will account for much of the increase.

He also cites an estimate that data centres added about €360 to the average Irish household’s cumulative electricity costs from 2015 to 2023. In Virginia, regulators have approved a tariff requiring data-centre customers using more than 25 megawatts to pay for substantial portions of contracted capacity from 2027, even if they use less than they reserved.

Why it matters

The argument is that AI’s costs may arrive locally, through electricity bills, grid charges, construction and demand for skilled workers, before any productivity gains reach the wider economy. Suciono notes that one study attributed only about 4% of US utility-rate increases over five years to data centres, and says electricity prices have many causes. The examples point to possible pressure, not proof that AI is driving global inflation.

Our read

The useful question is not whether AI will be inflationary forever. It is who pays while the power plants, transmission lines and data centres are built. Suciono’s proposed answers include requiring large facilities to help fund new capacity and monitoring lenders’ exposure to AI infrastructure. Sensible topics for debate; less sensible to treat an opinion piece as a settled economic forecast.

What to watch

  • Whether electricity and grid costs rise in areas with concentrated data-centre demand.
  • Whether regulators make large facilities pay more of the infrastructure costs they create.
  • Whether AI-related investment brings productivity gains quickly enough to offset construction pressures. Activity teaser: AI could make work cheaper in the long run, but its data centres need electricity, grids and capital now. Hari Suciono’s Eurasia Review analysis argues that households and businesses may feel some of those costs before the productivity payoff arrives. It also points to Virginia’s plan to charge large data-centre customers for much of their reserved capacity, even when they use less. Who should pay for the build-out: the companies demanding more power, or everyone who may eventually benefit from AI?

Discussion spark: Should data-centre operators pay the full cost of the extra power and grid capacity they require, even if that makes AI services more expensive?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.