AI and the Energy Transition: When Megatrends Collide

Two generational megatrends are colliding and, without bipartisan collaboration and partnerships among public and private entities, the fallout could have catastrophic economic and social implications. 

The transition to clean energy is one of our century’s great societal megatrends. Global investment in clean energy infrastructure is on track to hit $2 trillion this year, approximately twice the worldwide investment in fossil fuels but still short of the $4.5 trillion per year needed to limit global warming to 1.5 degrees Celsius and avoid the most extreme effects of climate change. 

A second generational megatrend is complicating the energy transition: The sudden and meteoric rise of artificial intelligence has led to a surge in demand for electricity. Grid operators are expecting electricity demand to increase by 40% to 100% by 2030, largely driven by data centers associated with AI, and this degree of growth is unprecedented since power grids were established nearly 150 years ago.  

The intersection of energy and AI will reshape both industries, but, currently, no single technology can fill the substantial gap between demand and supply. Solutions will require collaboration across the Federal Energy Regulatory Commission, the Department of Justice, utilities, grid operators, state and local agencies, as well as the private sector.  

Related:Uniting IT, Finance, and Sustainability Through the Integrated Profit and Loss

Power-Hungry AI 

A ChatGPT AI query uses about 10 times more energy than a Google search. The force behind AI -- the hyperscalers -- have aggressive decarbonization targets. However, most AI computing requires consistent power that naturally intermittent renewable power generation cannot provide.  

Without enough consistent power generation to meet data center demand, the US risks falling behind in the global AI race. That would be catastrophic on several fronts: from ceding economic growth and productivity gains; to jeopardizing national security; to conceding the leadership position that would allow a nation to shape global standards and steer the overall direction of AI’s development and deployment. 

Pressure on the Grid  

With unprecedented levels of new demand, grid operators are increasingly concerned about their ability to provide the power generation that data centers need while also maintaining grid reliability, especially while decarbonizing the broader energy mix. 

Consider Loudoun County, Virginia, where over 35% of hyperscaler data centers are located -- and where Dominion Energy has seen a 500% increase in power demand from data centers from 2013 to 2022. Dominion’s solar capacity has grown by more than 630% since 2015, but its 2023 Integrated Resource Plan would add up to seven new gas-powered plants, more than doubling the company’s gas fleet in Virginia to maintain reliability. The plan also proposes delaying by more than a decade the closure of two existing coal-fired power plants.  

Related:Purpose Versus Profit: Should Tech Companies Have Both?

Even with these adjustments, Dominion has been telling large purchasers that the utility won’t be able to provide sufficient new power to meet demand for at least five years due to grid constraints. Keep in mind that a data center takes only about one year to build. 

Multiple Solutions  

Analysts estimate that the US will need to spend $665 billion through 2030 on generating capacity alone, not to mention spending on transmission and grid upgrades. Europe’s grid, which is older, could need on the order of $1 trillion in investment.  

While energy storage is on a long-term trajectory to solve renewables’ intermittency problem, deployment remains nascent as technical performance and safety track records improve and costs decline. Small modular nuclear reactors offer another potential solution, but substantial improvements are required which many experts estimate will take 10 or more years to achieve.  

Related:Why CIOs Must Lead the Charge on Sustainable Technology

To meet the insatiable interim appetite for carbon-free power, several previously uneconomic, decommissioned nuclear power plants -- including Three Mile Island, scene of the worst commercial nuclear accident in US history -- are slated to reopen and they still would not provide enough new power. Parties are even starting to discuss possibly building new nuclear reactors.  

Perhaps even more vital and beyond technological solutions, legislation must create uniform permitting processes for such large amounts of new generation and transmission to be built. Currently, these differ by state and county, which creates enormous inefficiencies and results in moving targets that are prone to change based on partisan politics.  

Interconnection reform is also required to enable new power generation to dispatch onto the power grid. According to POWERGRID International, interconnect timelines from making an initial request to having an operational plant have increased from less than two years for projects built in 2000-2007, to more than four years for those built in 2018-2023.  

Unless we advance collaborative and holistic solutions, the energy transition and AI are two megatrends that will continue to collide and intersect in ways that create considerable challenges.