by Kyle Johnson
“Data is the new oil.” British mathematician Clive Humby coined this phrase to express that data, like oil, isn’t very useful in its raw state; it must be refined. Humby made this observation in 2006. It’s an idea whose time has come.
Data analytics is currently a $200 billion-per-year market. Recent advancements in AI look guaranteed to cause rapid and significant growth. That is, assuming data companies can get enough energy.
Current energy consumption and projections are staggering.
It’s estimated that data centers, AI, and the cryptocurrency industry used 460,000 gigawatt hours of electric power in 2022. The International Energy Agency (IEA) projects those sectors could use 1,000,000 gigawatt hours per year by 2026. For reference, the US currently has 94 nuclear reactors that generate about 778,037 gigawatts of electric power per year.
Obviously, data centers will strain power grids. Pablo Vegas, president and CEO of the organization in charge of Texas’s electrical grid, claims that meeting energy demand from crypto miners and AI data centers will require the state to expand its grid capacity from 85 gigawatts to 150 gigawatts by 2030.
Should investors trust these kinds of projections?
Probably not… and that’s a good thing.
A recent piece in the Wall Street Journal noted that in the last year, analysts doubled their five-year projections for US electricity demand.
You may not like AI. Feel free to snicker about it producing images of six-fingered hands. I chuckle at the ways schoolkids have been using it to cheat on their assignments. Thus far, AI is a bit gimmicky, and there are serious problems to address. But ask yourself: what are the odds that everyone will suddenly give up on AI?
I cannot envision such a scenario. If you see AI being around for the next few years, then it’s reasonable to conclude that energy demand will grow… even at rates that many would dismiss as unreasonable.
So where are private companies looking to get their energy needs met?
Nuclear reactors.
Data-center developers in Northern Virginia recently asked the local utility company for as much power as possible from nuclear reactors.
Amazon Web Services (a subsidiary of Amazon) purchased a nuclear-powered data center from Talen Energy for $650 million. The station generates 2.5 gigawatts. As Talen President Mark McFarland explained, “data centers are at the heart” of increasing energy demand worldwide.
This spike in energy demand has been paired with a boost in public perception. Every day, nuclear energy becomes increasingly palatable to people inside and outside the Beltway (that’s Washington, DC, for our non-American readers). Bill Gates recently invested more than $1 billion into TerraPower, a nuclear-energy startup. Gates estimates the world needs to build 100 such reactors to make a significant contribution to the climate. He loves to wag his finger at the average person’s carbon footprint, and I’m sure he has investments in so-called renewables—but he understands that they cannot meet baseload power.
A Panasonic subsidiary is building a $4 billion electric-vehicle battery factory in De Soto, Kansas. When fully operational, the factory will require between just 0.2 and 0.25 gigawatts of electricity… roughly the same demand as a small city. The factory’s energy demand helped delay the switch from coal to natural gas at the nearby power plant. Meanwhile, Zuckerberg fantasizes about running a one-gigawatt single training cluster for AI.
Old hippies and young voters may prefer solar panels and windmills, but the laws of physics don’t care. Nobody smart enough to make advancements in AI is foolish enough to hitch their wagon to unreliable “renewables.”
Fossil fuels or nuclear. With respect to AI, these are the only viable choices.
I have no doubt that some climate alarmists and ideologues would insist on renewables even if it meant radically reducing everyone’s quality of life. But political jurisdictions will not act in unison. Whether they realize it or not, governments are currently in a race to attract data-processing centers and intellectual talent.
Edwin L. Drake is largely credited for starting the Pennsylvania oil rush in 1859. But unlike oil, AI is not tied to certain patches of land. Anyone can stake their claim—if they have enough energy. I’m not sure who deserves the credit, but the AI rush has already begun.
It’s somewhat ironic that Dubai, of all places, is making strides to attract data centers. It’s a few nuclear reactors away from having enough power to crunch data and prevent the GPUs from melting despite the desert sun.
Yes, it takes years to permit and build a nuclear facility. I don’t recommend cutting any corners. But smarter governments will learn to expedite the permitting process as data centers flock to cheap, reliable energy hubs.
Fortunately, it’s not too late for speculators. There’s plenty of work to do before more construction crews break ground. During the nuclear renaissance of the past few years, we’ve seen contracts signed long before new reactors come online. Reactors slated to be decommissioned have been kept online. And though it’s a long and painful process, dormant reactors have been restarted.
Now add AI into the mix.
You might have been upset about not fully capitalizing on uranium’s breakout last year. I think more profits are baked into the cake. AI may become the single greatest source of energy demand in human history.
Play your cards right, and there’s still a lot of money to be made.
KJ
P.S. Lobo went heavy into uranium before it became fashionable. There’s only place to get his latest thoughts on the markets, uranium or otherwise: his free, no-hype, no-spam Speculator’s Digest.
