AI in NC? Only if there’s power.

“AI” is a strong contender as the buzzword of 2024.

In what seems like the blink of an eye, the most prominent technology companies (and some heretofore unheard of companies) have gone from dreamily talking about the future potential of artificial intelligence, to rolling out multiple AI platforms; from Large Language Models (LLMs) able to pass the Turing test, to image and video generation AIs so impressive they bend one’s sense of reality.

Much like the move to “cloud computing,” AI is a business of computing power, data storage, and economy-scale energy consumption. Seriously: By 2030, it’s estimated that data centers in the US could consume as much electricity as some major industrialized economies produce.

That makes energy — its cost, reliability, and availability — central to the budding AI industry.

If North Carolina is to capitalize on the AI boom, then, it must be able to power it. The Tar Heel State has been rated a top-tier state for business multiple years in a row, and, in addition to lower taxes and regulations, the state’s relatively low-cost energy rates are a big part of attracting industrial enterprise.

It’s a position the state’s leaders, and people, would like to maintain. After all, “Best for Business” means growth, jobs, and prosperity. Yet, staying on top requires policies that reinforce such an environment, and clearing impediments.

According to CBRE Research, nationwide, robust demand from cloud computing and artificial intelligence led to a significant increase in data-center construction activity in the first half of 2024 (10%). Year-over-year growth stood at 24%, and the pipeline is growing. Where new AI data-center installations are located, however, will depend on whether or not a given location can efficiently and confidently provide the requisite energy to power the server farms that undergird innovations like ChatGPT.

The largest grid operator in the US, PJM Interconnection, warned earlier this year that aggressive closure schedules for coal-fired power-plants, and a lack of timely new supply, threatens a major energy supply gap.

“The amount of supply resources in the auction decreased again this year, continuing the trend from recent auctions and underlining PJM’s stated concerns (PDF) about generation resources facing pressure to retire without replacement capacity being built quickly enough to replace them,” stated PJM in a July 30 press release. “These reliability concerns associated with reducing supply and increasing demand are not limited to PJM; the North American Electric Reliability Corporation has identified elevated risk to the reliability of the electrical grid for much of the country outside of PJM.”

As part of Duke Energy’s Carbon Plan, in response to state energy policy legislation in 2021, North Carolina is retiring coal plants and doubling down on problematic renewables to meet emission-reduction goals. At the same time, projected energy demand in the state is soaring as electricity-hungry development and substantial population growth portend a massive need for power.

The result has been higher prices and undermined reliability; eating away at North Carolina’s comparative advantage in courting AI data centers and other energy-intensive businesses.

Understanding why North Carolina’s energy costs are below the national average, and still low compared to surrounding states, requires an acknowledgment of good policy decisions along the way. Decades ago, a firm priority placed on the provision of plentiful, reliable energy at affordable prices led the state to leverage an energy source that hit the mark — nuclear.

What followed was the development of one of the most resilient and reliable energy markets in the country. A far cry from the “Rip Van Winkle” conditions in which the Old North State has found itself previously.

As recently as 2022, North Carolina was ranked among the five states producing the most electricity from nuclear power, accounting for 32% of the state’s net generation.

But despite nuclear’s record of success, current plans to introduce new Small Modular Reactor (SMR) technologies that could meet the demands of new residents and new industries are still a decade out; far too long a wait for the site selection executives looking to break ground on new data centers as soon as possible to service the AI boom.

Whether traditional factory managers, or data-center directors, each prefer their power source plentiful, there when they need it, and not deleterious to the bottom line. Such an energy infrastructure is a must in order to become an AI superpower.

Industry experts note the following requirements of the AI data-center industry:

• Inexpensive electricity costs given the immense amount of power to be consumed on an ongoing basis, particularly since inference needs will only compound over time.
• Stability and robustness of energy supply chain against geopolitical and weather disturbances to decrease likelihood of energy price volatility, as well as the ability to quickly ramp up fuel production and thus rapidly provision power generation at great scale.
• Power generation with a low carbon intensity power mix overall, suitable to stand up massive quantities of renewable energy that can produce at reasonable economics.

SMRs fit the bill. They also enjoy the “green” status, with zero emissions, to satisfy the carbon-neutral targets currently complicating energy policy in the state.

According to Jon Sanders, director of the Center for Food, Power, and Life at the John Locke Foundation, a public push to accelerate the timeline for deploying SMRs will go a long way in reducing uncertainty, and thereby increasing opportunities for prosperous growth.

“The longer we go with the uncertainty of how we will replace baseload coal power in this state — not to mention if we do so with reliable resources like natural gas leading to nuclear or lean heavily on unreliable, intermittent sources like solar and wind — the more difficult it will be to expand economic opportunities, especially in rural communities,” says Sanders. “The current energy policy of replacing working coal plants with other, lower-emissions power facilities is going to make power bills more expensive. That said, adding reliable, clean-energy plants like SMRs would mitigate the worst of the price increases and provide reliable power rather than cause residents to pay a lot more for power they can no longer depend upon.”

Leveraging such technologies might even be the only way data-center capacity can be expanded to meet AI demand.

In congressional testimony earlier this year, CEO of one of the nation’s largest utilities pointed out the true step-change in demand AI poses. He said, while a large manufacturing facility might need 100 megawatts of electricity — equivalent to about 100,000 homes — it is now increasingly common for a single data center to need anywhere from three to 15 times that amount of power.

In an earnings call last month, Constellation Energy CEO Joe Dominguez told investors that hooking data centers directly to nuclear reactors is the fastest and most cost effective solution.

“The notion that you could accumulate enough power somewhere on the grid to power a gigawatt data center is frankly laughable to me,” Dominguez said to investors in August.

SMRs would be a great asset to meet the challenge, former state lawmaker Andy Wells told the Carolina Journal last month, but he laments the timelines utilities have for deploying advanced nuclear as far too long.

“Kennedy said in 1962 : ‘We choose to go to the moon in this decade…’,” Wells noted in an email to Carolina Journal on the topic of SMRs. “It took seven years. Now we hear it takes 10 years to replicate something on land that the US Navy has done successfully for 75 years.”

His conclusion? The onus is on local communities to seek out partnerships with private companies to electrify opportunities for 21st century prosperity.