Gigawatt Paradox: Finding the Right Balance Between AI and Energy Capacity

Contrary to revisionist history: reshaping an age and how the world progresses is rarely consensual, except in hindsight. The loudest voices in those times often were laggards eventually swept away by the irresistible force of reality.

Consider the start of the automobile age, with old-timers unfavorably comparing the loud, unreliable early engine to the trusty horse and carriage. People were literally killed over the rise of the printing press and electricity was often feared as a supernatural force.

So, people are often slow on the uptake and do get it wrong. Technological innovation, however, is relentless and nearly always wins.

Some of the same debates are happening now over the rise of artificial intelligence, which ironically is the technological Johnny-come-lately begging the same age-old questions. Is AI a hyped bubble which will prove fake in the annals of the future? Or is it a life-changing economic and technological force which, like crop machinery in hesitant agrarian societies slow to see the benefits, will eventually convince those who first opposed it?

 Hint hint! It might be a good idea to keep an open mind because most companies out there already see the value in AI and are exploring generative AI in many of their processes. This is driving revenues for companies which develop AI training or produce AI-enabled tools, while also raising energy demand exponentially.

AI may goose the market here and there in historically bubbly fashion , but the so-called hype shows every indication of coming true as it is infused into nearly every economic sector.

And sometimes truth strikes fear in the heart of the status quo.

“Well, I think it is unprecedented what we face,” Mark Christie, a longtime utility regulator and former chair of the U.S. Federal Energy Regulatory Commission, said during a keynote session at the Schneider Electric Innovation Summit last month in Las Vegas. “We are sleeping on a volcano.”

Christie was quoting author Alexis de Toqueville, the writer who was describing forces driving the French Revolution. In this case, Christie is referencing the reality of power-hungry AI, which is driving individual data racks eventually to 1 MW and higher as well as creating data centers  demanding 1 GW plus at a time when the power generation sector is not ready.

Without adequate power generation capacity and the financial wherewithal to build it, electricity affordability could be in peril in a way that makes a historic power outage look tame. Supply and demand imbalance don’t lie.

“We cannot forget that what retail customers are paying is going to be a huge part of this,” Christie added. “If we don't, the volcano could blow up, and it could be bad.”

Christie was not disputing the importance of AI but was focusing on the feared energy imbalance it could bring. There are plenty of AI haters out there, from skeptics who don’t believe it will dominate digital infrastructure to humanists worried about job losses and stealing intellectual property.

A prominent AI critic is television legend Vince Gilligan, the creator behind “Breaking Bad” and “Better Call Saul.” In a recent interview about this new show, “Pluribus,” Gilligan had some choice words about AI, so wording so choice that EnergyTech cannot run his quote verbatim.

“I hate AI,” Gilligan said in a chat with Variety. “AI is the world’s most expensive and energy-intentive plagiarism machine. . . . It’s basically a bunch of centibillionaires whose greatest life goal is to become the world’s first trillionaires. I think they’re selling a bag of vapor.”

Hmm, one can understand a creative mind lamenting the potential drain of human invention to drama, comedy and music. But calling AI a bag of gas may not only miss the bigger picture, but also exclude what’s driving economic engines today.

As Schneider Electric CEO Olivier Blum pointed out during the Innovation Summit, some 80-plus percent of the world’s companies already are implementing AI technologies in everyday and long-term operations. These could include anything from predictive maintenance or digital twinning to generative AI for marketing purposes.

“AI is touching every aspect of our lives,” Blum pointed out.

He was backed up quickly by FedEx CEO Raj Subramaniam, only the second CEO ever of a logistics giant which moves 17 million packages a day. Data is key to AI, and no one entity sits on more transportational data than FedEx.

One sure way for a company to get run off the competitive superhighway is to ignore that.

“I didn’t know the AI revolution was coming, but it did come,” Subramaniam said. “The way I think about AI is in improving efficiency and back-office function. But (now) more important to me is how to leverage AI to create new value for our customers, especially our industrial customers.”

And herein comes the gigawatt paradox. “There is no AI without computing,” Schneider Electric CEO Blum noted. “And there is no compute without energy.”

Many observers, from Goldman Sachs to McKinsey to Bloomberg, are anticipating the power grid will require an additional 150 to 200 GW of new power generation capacity to meet the demands of future AI-enabled and cloud-based computing installation. Two hundred gigawatts are the equivalent of 200 new nuclear power reactors.

It’s going to take an “all of the above” approach to match 2030s AI demand with commensurate power generation resources. More nuclear, more gas-fired power, more renewables, increasingly. More and more of more.

Power utility capital expenditure for generation is growing after years of flatlining, now expanding building plans by as much as 70% in the past three years. Too slow, some say.

“That investment is happening but that investment by itself is not enough,” said James Lee, partner at asset management and private equity giant BlackRock Global Infrastructure Partners, itself a huge investor in the power sector.

“It requires all of the above solutions on baseload power, with other technologies needed,” Lee added. “Gas plants with carbon capture, next-gen geothermal, and all forms of nuclear. We need all of these to be able to meet the task at hand.”

Indeed, hyperscalers such as Google, Microsoft, Meta and Amazon are out in front on this, either signing long-term power purchase agreements around reviving the current U.S. nuclear generation fleet or building next-gen small modular reactors and perhaps even nuclear fusion at some futuristic point.