Nuclear And Solar Energy Could Converge To Solve The AI Energy Crunch

While many factors are converging to power the current AI boom, there are two major components leading the charge: data and energy. Without the trained datasets that power AI models, there would be no AI boom. But perhaps the energy needed to train these massive datasets is even of greater significance.

The energy required to train large language models, process massive datasets, and power always-on AI systems is simply staggering. For context, it took as much electricity as 120 American households consume in a year to train OpenAI’s GPT-3, which has 175 billion data parameters, according to The Association of Data Scientists. Per one report by the BBC, by 2027, projections suggest that AI-driven energy consumption could rival that of a mid-sized country like the Netherlands.

As these energy demands continue to skyrocket, how can governments and companies power AI’s future sustainably and reliably? Just a day after his inauguration ceremony, President Donald Trump announced the Stargate project, a $500 billion investment commitment that’s aimed at positioning the United States as the global leader in AI. While that’s a welcome development, one question that’s being asked about Stargate is where the energy to power its lofty AI infrastructure will come from.

Nuclear and solar energy, which Stargate also highlights, are increasingly emerging as viable answers to that question. How will these energy sources converge to address AI’s energy dilemma?

The AI Energy Crunch: Why Existing Systems Are Struggling

AI data centers are voracious energy consumers requiring massive amounts of power for computation and cooling — running 24/7 in many cases. Most data centers rely heavily on centralized grids, many of which are outdated and ill-equipped to handle the burgeoning load.

According to the International Energy Agency (IEA), over 3,000 GW of renewable energy projects were stuck in grid connection queues at the end of 2023, signaling significant bottlenecks in integrating new energy sources. There’s a dual challenge here: While it’s important to scale power generations, it’s equally important to ensure it doesn’t adversely impact the environment.

AI’s huge carbon footprint has become a hot topic since large language models went mainstream a little over two years ago. With global climate commitments now pushing for net-zero emissions by 2050, the stakes have become even higher.

Nuclear Energy

Nuclear energy is a stable, carbon-free source capable of delivering uninterrupted power — a stark contrast to the intermittency of energy sources like solar and wind. “Nuclear energy is providing massive and stable thermal energy to power rotating electricity generation machines,” Richard Schomberg, special envoy for smart electrification at the International Electrotechnical Commission told me. “Their significant mechanical inertia stabilizes the entire electric system,” he added.

However, nuclear energy isn’t without its challenges. High upfront costs, lengthy development timelines, and public safety concerns — all amplified by catastrophic nuclear incidents like the Chernobyl disaster of 1986 and Fukushima in 2011, particularly in the face of extreme weather and geological events — continue to hinder widespread adoption.

Solar Energy

Solar energy, on the other hand, is agile and scalable. “Solar is inherently versatile and generates power closer to where it’s consumed,” said Gil Kroyzer, CEO of solar technology provider Solargik.

As a result, solar energy production reduces reliance on centralized grids, minimizes transmission losses, and improves overall energy efficiency, added Kroyzer.

You can deploy solar on rooftops, in deserts or other underutilized or remote areas, and even alongside agriculture through agrivoltaics — a dual-use concept that uses land for both solar energy production and agriculture at the same time.

“This dual-use concept not only maximizes land efficiency, but also aligns with the growing trend of placing data centers in rural areas. That’s why solar production with advanced trackers is uniquely suited to support these evolving needs, offering versatile solutions tailored to each site’s conditions,” explained Kroyzer.

However, even with all the incredible benefits of solar energy, it faces a big limitation— intermittency: solar panels don’t generate energy at night or during cloudy weather. The good news is that by pairing solar with energy storage systems, solar’s intermittency problem can be mitigated.

Thankfully, some companies are already doing that, developing battery energy storage systems and hybrid models that better conserve solar. For example, Solargik’s hybrid solar project at a leading European automotive plant integrates rooftop installations with ground-mounted systems, demonstrating how diverse, distributed energy systems can meet high energy demands, while improving reliability. As Kroyzer noted, these systems help to ensure continuous, round-the-clock energy availability by seamlessly complementing advanced storage solutions.

The Convergence

Rather than view nuclear and solar as competing solutions, Schomberg noted that the future of energy is not about choosing one technology over another. “It’s about creating a balanced mix that leverages the strengths of each,” he added.

One way to look at it is how each of these two energy sources could address the other’s shortcomings. Nuclear could provide the stable baseload power that AI systems need to keep running 24/7, while solar could offer scalability and rapid deployment to meet peak demands and supplement grid capacity.

In another instance, nuclear energy can stabilize grids with high solar penetration by compensating for variability. Meanwhile, solar can reduce the burden on nuclear plants during daytime peaks, optimizing overall energy efficiency.

Hurdles On The Way

The energy industry is a regulation-heavy one. Developing nuclear facilities involves navigating stringent safety protocols, while solar projects often face local opposition due to land use concerns.

“The regulatory environment is evolving rapidly and often presents conflicting priorities,” explained Kroyzer. “Solar, in contrast to nuclear, is a good neighbor. It’s quiet, safe, renewable, and imposes no harmful demands on its surroundings.”

But, according to Gilles Thonet, IEC Deputy Secretary-General, international standards can play a vital role in harmonizing these regulatory frameworks which are always evolving and getting even stiffer.

“The IEC and ISO joint committee for AI (SC 42) is working on metrics to measure energy efficiency and sustainability, helping businesses navigate these challenges more effectively,” he said.

What’s Ahead?

AI’s energy demands will only intensify. While it’s necessary to meet these demands, players in the AI industry must do so in a sustainable manner. It’s great that big tech companies like Google and Microsoft are already setting the precedent by integrating nuclear and solar into their energy strategies — but more can be done.

At the current rate, AI’s voracious energy appetite only means that the global carbon footprint will continue to widen, further pushing the net-zero goal for 2050 out of reach. However, by combining the stability of nuclear energy with the flexibility of solar energy, companies can build a more energy infrastructure to power AI’s future sustainably.

For Thonet, international collaborations, consensus, and transparency are key to not just developing solutions to measure and reduce the impact of AI on our planet, but also ensuring it's safely, responsibly, and sustainably deployed.

In addition, Kroyzer noted that advanced solar technology has the potential to produce energy at the times that it’s most needed. “For energy-intensive users like data centers, this ensures reliable, sustainable power that can meet growing demands. By integrating seamlessly with advanced storage and grid solutions, flexible and future-ready systems offer a renewable path forward,” he concluded.