A version of this story appeared in the daily Threat Status newsletter from The Washington Times. Click here to receive Threat Status delivered directly to your inbox each weekday.
OPINION:
The U.S. is locked in a global race to dominate artificial intelligence. Winning that race will depend on many factors: advanced semiconductors, computing power, software innovation and talent. One essential ingredient will trump all: electricity.
AI is extraordinarily energy-intensive. Operating hyperscale data centers requires vast computing power and enormous amounts of electricity. A single data center campus can demand as much power as a midsize city.
As AI adoption accelerates across the economy, electricity demand in the U.S. is rising at a pace not seen in decades.
That reality raises an important question for policymakers: Does the structure of America’s electric system support the power needed for the AI economy?
For most of the 20th century, the answer would have been straightforward. Electric utilities were vertically integrated, responsible for the production, transmission and delivery of electricity within their service territories.
In exchange for a guaranteed service territory, power companies accepted an obligation to serve, meaning they were responsible for ensuring there was enough power and that the poles and wires were in place to meet growing demand.
Over the past several decades, however, many regions of the country have adopted restructured (or retail choice) electricity markets that have separated power production from regulated utilities. Under this model, independent power producers, rather than electric utilities, are tasked with building new generation in response to market demand and related price signals.
Competitive markets should encourage efficient investment and lower electricity prices. In practice, wholesale power markets in several restructured regions have struggled to deliver either outcome. Wholesale prices have risen, and investment in the new generation has lagged, raising concerns about whether these market structures are delivering the affordability and reliability they were intended to provide.
The rapid growth in electricity demand driven by AI and advanced computing is testing whether these fragmented responsibilities can respond quickly enough without driving up costs for ratepayers.
Consider the mid-Atlantic region served by PJM Interconnection, the nation’s largest wholesale electricity market, spanning 13 states. Utilities in the region operate transmission and distribution networks but generally do not control generation decisions.
As Edison Electric Institute President Drew Maloney recently explained, utilities in these markets are essentially “wires companies” that depend on others to build the power plants customers rely on. When demand accelerates rapidly (as it is currently doing), that separation can slow efforts to bring new power online.
Hyperscale data centers supporting artificial intelligence are being announced across the PJM region at an unprecedented pace, with individual campuses often requiring hundreds of megawatts of power. Meeting this demand will require rapid investment in new generation and grid infrastructure.
Yet in markets where responsibility for building new power plants is separated from the utilities that serve customers, coordinating those investments can become significantly more complex.
The data center boom is already pushing electricity demand to record levels in many regions of the country.
Meeting this demand will require building power plants, transmission lines and grid infrastructure at a scale the United States has not seen in decades.
Just as important, the structure of the electric system affects who ultimately pays for this infrastructure. Vertically integrated utilities can plan investments over decades and recover costs through state-approved rates, spreading large capital expenditures across broad customer bases and long horizons.
Long-term planning can help ensure that the infrastructure needed to power economic growth is built while maintaining affordability for households and businesses.
Who will win the AI race? No one knows, but the country that can build the necessary infrastructure fastest will almost certainly have the advantage.
Vertically integrated utilities offer one important advantage in this environment: They combine long-term system planning with clear responsibility for delivering electricity to customers.
Because these utilities coordinate generation, transmission and distribution planning across an entire system, they can align investment decisions and build new infrastructure when demand increases.
Equally important, states expect these utilities to maintain reliable service. That obligation provides strong incentives to ensure that new generation and transmission capacity are built before shortages emerge.
None of this suggests that markets cannot play a role in the electricity sector. Market mechanisms can encourage innovation and efficient operation of power plants. Yet as the scale of electricity demand growth accelerates, policymakers may need to revisit how responsibility for system adequacy is structured.
If the United States intends to maintain its global lead in AI, then it must ensure its electric system can deliver the power needed to sustain it.
• Joy Ditto is CEO of Joy Ditto Consulting and former CEO of the American Public Power Association and the Utilities Technology Council. Paul Griffin is a senior adviser at Joy Ditto Consulting and served as executive director of the Washington Rural Electric Cooperative Association.
Please read our comment policy before commenting.