Energy: Tech Constraint or Accelerant?

The technology industry is insular and self-absorbed. Its denizens and principal players often see themselves as singularly important, the ever-shining sun around which all other businesses and industries revolve. 

One could argue that there’s some justification for the tech industry’s bumptiousness. The Magnificent Seven have dominated stock-market gains during the past two years; the world’s leading technology companies flaunt market valuations that are the envy of companies worldwide. If you consult any list of the world’s richest billionaires, technology entrepreneurs and founders are well represented. Technology companies have generated tremendous wealth for their founders, shareholders, and employees; and the tech world’s global capital, Silicon Valley, continues to spin riches from a golden wheel of venture capital. 

Nonetheless, I am here to tell you that there is an industry that is more important than technology, certainly more strategic and more essential in all critical respects. I am alluding, of course, to the energy industry, which is currently shambling its way through a major, occasionally fitful, transition from fossil fuels to renewable energy. 

Think about it: Without abundant, secure, and reliable energy supplies, no other industry can endure, much less prosper. 

The energy industry confers great riches on those who own and control it. For example, the royal families of the Gulf States, according to some estimates, possess greater wealth than many of Silicon Valley’s most celebrated billionaires. The sovereign wealth funds of the Gulf States, which built their empires on oil and gas, are currently investing billions of dollars in everything from professional sports to artificial intelligence, electric vehicles, and life-extension technologies. 

Bloody and costly wars have been fought, and continue to be fought, over energy. As energy is the lifeblood of all other industries, it is imbued with undeniably strategic geopolitical value. 

Salience of Energy and Energy Efficiency

There is no question that the next phase of the technology industry’s growth and prosperity – represented by next-generation datacenters playing host to AI applications and services – will be determined by the abundance and continuous availability of energy supplies and by the enhanced energy efficiency of IT infrastructure.

You might be familiar with the oft-repeated adage attributed to hockey star Wayne Gretzky (though his father, Walter Gretzky, might have been original source): “Go to where the puck is going, not where it has been." In hockey, as in life, achieving the outcome of that maxim is easier said than done. Even so, as an aspirational principle, the Gretzkys’ counsel has merit. 

As we shift our gaze toward the horizon and what’s next for the technology industry, we should ask: Where is the energy puck going? 

Well, for starters, we might need a bigger puck. For a while now, and confirmed resoundingly in the past few days, we’ve heard that the advent of generative AI (genAI) and the seemingly impending arrival of artificial general intelligence (AGI) will require vast amounts of energy to power the datacenter infrastructure that processes AI models and inferences. The drumbeat of energy warnings grows louder by the day. 

A Reuters article, published yesterday, reports that U.S. electric utilities predict a tidal wave of new demand from datacenters powering application such as generative AI. Some utilities are projecting electricity sales growth several times higher than estimates offered only months earlier.

Datacenters are driving capital expenditures and growth at leading U.S. electric utilities, according to the Reuters analysis. The following excerpt from the article provides further detail:

"The growth is going to kick in faster than it has in decades," said Jim Lydotes, head of equity income for Newton Investment Management, a BNY Mellon IM firm that is shifting its holdings in European electric utilities to U.S. companies . . .

Overall, power use from the thousands of giant computing warehouses that comprise data centers is expected to triple globally from less than 15 terawatt-hours (TWh) in 2023 to 46 TWh this year, according to Morgan Stanley research.

"The truth of the matter is these things (data centers) are pigs when it comes to energy use, and now they're the size of an elephant," said Eric Woodell, an expert who specializes in data center operations.

Going Crackers

Pigs, elephants? The surging energy requirements of AI-related datacenters have unnerved market watchers so much that they sound as though they’ve swallowed a box of animal crackers. 

You can understand why the energy cognoscenti are afflicted by double-edged emotions of anxiety and excitement. Where there is a crisis, opportunity waits furtively in the shadows for its cue to take the stage. For energy, that moment has arrived, but the energy required will increasingly have to assume renewable form. Nobody wants to be burning coal, oil, or gas in service to datacenter expansion. What’s the point of running for the AI rainbow if it ultimately leads to a scorched husk of environmental devastation? 

In case you think the electric utilities are overstating the business opportunity represented by the ravenous appetite of datacenters, the Reuters report includes reference to McKinsey estimates suggesting that longer-term power demand from IT equipment in U.S. datacenters is expected to reach more than 50 gigawatts by 2030, up from 21 gigawatts in 2023. About a year ago, McKinsey had set a significantly lower estimate for 2030, but recent evidence compelled an upward revision.  

Meanwhile, the Wall Street Journal quotes Rene Haas, CEO of semiconductor-design company Arm, saying AI models such as OpenAI’s ChatGPT “are just insatiable in terms of their thirst” for electricity. “The more information they (the models) gather,” Haas said, “the smarter they are, but the more information they gather to get smarter, the more power it takes.” 

Rather than underscoring the need for greater energy provision, which may be needed at any rate, Haas emphasized the requirement for greater energy efficiencies in the design and operation of artificial-intelligence applications and IT infrastructure. 

If greater energy efficiency is not forthcoming, according to Haas, “AI datacenters could consume as much as 20% to 25% of U.S. power requirements (by the end of the decade). Today that’s probably 4% or less. That’s hardly very sustainable, to be honest with you.”

At least in this context, Haas might be a master of understatement. 

The Wall Street Journal article also cites a report published in January by the International Energy Agency (IEA), indicating that a standard ChatGPT request requires ten times as much energy as a Google search. As genAI queries proliferate, the energy implications grow potentially ominous. 

Even Elon Musk, the enigmatic and intermittently lucid tech billionaire and celebrity provocateur, recently opined that artificial intelligence could become smarter than anyone on this planet by next year, contingent – yet again – on the availability of sufficient energy and AI chips. 

Referring to constraints on the blitzkrieg technological advances of artificial intelligence, Musk said, “Last year it was chip constrained. People could not get enough Nvidia chips. This year it’s transitioning to a voltage transformer supply. In a year or two, it’s just electricity supply.”

Masters of Understatement

“Just electricity supply”? Like Haas of Arm, Musk, against all previous evidence to the contrary, burnishes his credentials as a master of understatement. 

Since my ostensible retirement – though I’m not sure I’m adept at the conventional art of retirement – I have spent considerable time reading, learning, and tweaking my modest investment portfolio. I do what I can with my limited human intelligence, 

The fact is, there is only so much I will never know, despite my best efforts. I always thought, when I was younger and ingenuous, that there would be more than enough time for nearly everything in this life, including comprehensive learning. But that’s not true, dear reader. There are limits to everything.

Despite my all-too-human limitations, I have done my utmost to adhere to the Gretzky family’s ambitious precept of anticipating where the puck will be rather than where it has been. 

One of the hypotheses I’ve devised, based on extensive research, is that energy will remain the fount of all other forms of industrial activity and wealth generation. Part of the value proposition around energy involves its abundance, or lack thereof, but another notable piece of the puzzle involves the sort of energy that is needed. As we know, the age of fossil fuels dominance is reaching its end, and a major transition to renewables has begun.  

If you consider the burgeoning energy requirements of modern datacenters, renewable and cleaner energy will be required, including everything from small modular reactors (SMRs) to geothermal, solar, and hydrogen power, the latter preferably sourced from renewables. Also in the picture are new materials and technologies, hardware and software, that yield greater energy efficiency. Plenty of work has begun on that front, with more to follow. 

Further exploration of this scenario offers what is arguably a more intriguing and rewarding long-term investment gambit than trying to ascertain whether a richly valued Nvidia has reached its top.