AI-induced energy supply-demand imbalance endangers economic growth

Over the last two years, generative artificial intelligence (Gen AI) has taken the country by storm. Seemingly you can’t turn on the TV or read an article without some mention of the promising technology. But what is less understood is the burgeoning supply-demand imbalance these new hyperscale AI data centers are causing in the energy markets, and the risk this poses for the country’s economic growth prospects. To put Gen AI’s energy consumption into perspective, consider that one ChatGPT inquiry consumes 10x as much energy as one Google search. Couple that with the notion that the energy complex has yet to feel the true impact of widespread Gen AI deployments, and you have a recipe for increased economic risks.

Economic growth is anchored in two core pillars: a growing workforce and technology that makes the workforce more productive. Growing the workforce will be a challenge over the coming years given the structural headwinds facing the U.S. workforce (declining birth rates, aging population, early COVID-related retirements, lack of immigration reform, etc.). This makes the adoption of future Gen AI applications and the operational efficiencies they deliver critically important to economic growth. But here is the rub: Many market participants believe that the forecasted explosive growth means we are 12-24 months away from not having enough electricity to run these power-hungry Gen AI data centers. Should this happen, energy supply will become a bottleneck for data center growth and potentially the broader economy.

Today, data centers represent approximately 4% of the country’s energy demand, but that number is expected to grow exponentially over the coming years. Consider this: From an application perspective, ChatGPT – which was launched two years ago – is largely responsible for the current surge in Gen AI data center deployments. Given that it takes about three years to standup a new data center, the energy complex has yet to feel the full effects of Gen AI.

Historically, data center operators prioritized access to deeply integrated fiber networks and proximity to Fortune 500 companies as the main criteria for new data center locations. Today it is all about power – a clear path to power for a new data center is far and above the number one criterion. The situation has gotten so dire in Georgia that the state recently increased its new long-term energy needs by over 38% in just one year, which has firmly planted the state in uncharted waters and is overwhelming the Public Service Commission. Hence the move to rural markets where access to renewables and excess power is available, and the energy market’s increased attention during the presidential election.

We are at a point where a monumental type of effort is needed at the federal level. Issues related to generation and transmission, permitting and the enormous levels of capital needed to build new infrastructure are not being addressed quickly enough to prevent a long lasting and very problematic data center bottleneck.

Right now, there is a terrific opportunity for the utility industry to be intentional in how data center demand is met, setting an important precedent for handling future load growth in a way that better supports the grid.

There is a unique opportunity to invest in the broad array of solutions starting with consumer efficiency and demand management and working upstream to provide ample, reliable and secure power in an age of rising electricity demand. It is important to realize that even if the demand growth for data centers flattens or even reverses, it will set utilities on the right path for preparing for the subsequent wave of electricity growth driven by strong industrial policies and electrification of everything. That said, establishing priorities will remain critical, requiring the collective grid – from the consumer to the distribution system and beyond – to collaborate on solutions to avoid unintended consequences and optimally capture long-term benefits.

Jeff Johnston is Lead Economist, Digital Infrastructure for CoBank. This report first appeared in The Quarterly, prepared by CoBank’s Knowledge Exchange Division (October 2024). Greenwood Village-based CoBank is a cooperative bank serving vital industries across rural America.