AI's Power Grab Is Ahead of Schedule
The Weekend Edition is pulled from the daily Stansberry Digest.
AI is using more energy than you might think...
Over the past few weeks, we've seen plenty of AI stories – from bubble concerns... to the "SaaSpocalypse"... to worries about AI replacing an unknown number of jobs.
Today, we're looking at the infrastructure needed to continue this story – data centers and the energy to power them.
When OpenAI's ChatGPT launched the AI boom back in November 2022, data centers only made up about 4% of total power demand in the U.S.
But since then, their energy share has increased substantially – hitting 7% last year, according to analysts at Goldman Sachs. Take a look...
The trend will only continue from here.
At the end of 2024, the Department of Energy projected that data centers could make up anywhere from 6.7% to 12% of total power demand by 2028.
Here we are, only about a year after that projection, and we're already in that range – two years sooner than expected. And those are just the headline numbers.
The story is even more pronounced in parts of the country with high data-center concentration, like "data-center alley" in northern Virginia. These digital hubs made up 26% of all power demand in the region in 2023.
The Power Grid Isn't Ready for AI
There's also a backlog for new data centers waiting to go on line...
On social media platform X, user Lukas Ekwueme shared the following chart based on recent survey data. As you can see, big tech companies (the green dots) believe their data centers can start accessing the grid next year. But utility companies (blue) are less optimistic...
The average difference between what hyperscalers are hoping for and what utilities can deliver is just about a year.
But in places like Texas and Ohio, utility companies are less optimistic. They don't think they can power new data centers for two years or more.
Energy companies' earnings are telling the same demand story...
Earlier this week, natural gas giant EQT (EQT) released its quarterly earnings results.
EQT produces and delivers natural gas across states like Pennsylvania, Ohio, and West Virginia.
In its earnings presentation, EQT noted that over the past 11 years, demand for gas-fired power "only" grew by about 14 billion cubic feet per day (bcf/d). But over the next five years, data centers will account for 10 bcf/d of power demand... potentially as much as 18 bcf/d.
In other words, data centers alone will demand the same amount of new power as we saw in total over the past 11 years.
The Fuel Behind the AI Power Boom
One energy source is the obvious solution in the short term...
We've written a lot about nuclear energy in previous Digests. While we're seeing tons of investment from Big Tech and the government, these new nuclear plants will be under construction for years... let alone up and running.
As Commodity Supercycles editor Whitney Tilson and his team explained in their January issue, there's only one energy source ready now to meet increased demand: natural gas.
It's the most reliable source to create electricity at scale. As they wrote...
There's no way to provide electricity to all these new data centers without building more power plants. And while solar, wind, and nuclear are working to make up some of that demand – and nuclear in particular is set to provide more power in the future – natural gas is the world's cleanest, most reliable energy source right now.
And much-needed natural gas capacity is already on the way...
Pipeline company Energy Transfer (ET), which runs natural gas pipelines throughout Texas, the Midwest, and the southwest, also reported earnings this week.
The company said it has started supplying one of Oracle's (ORCL) data centers with natural gas. It's part of a deal to supply three Oracle data centers with up to 900 million cubic feet of natural gas per day. That's enough to power at least 3.5 million American homes.
And we're seeing growing data-center demand in states like Arizona. According to the data-center-project tracker Cleanview, Arizona has 23 operational data centers, with another 29 planned. The planned data centers would increase the state's computing capacity sixfold – and bring a lot more demand for power.
In short, as the Commodity Supercycles team concluded in last month's issue, "with natural gas in such high demand, it's a great time to own companies in the space."
If you're not sure how to navigate AI's disruption in the market...
Here's a good place to start.
On Wednesday, our colleague Mike Barrett published a simple, effective framework for gauging the impacts of AI on your portfolio...
The rise of AI technology is introducing a new dynamic we must adapt to... The market has splintered. And it's helpful for us to place stocks in one of three categories...
Category 1: AI Winners (These stocks directly benefit from rising AI demand.)
Category 2: AI-Resistant Stocks (These stocks are unaffected by AI.)
Category 3: AI Victims (These stocks are at risk of disintermediation from AI.)
There's a lot of uncertainty around future revenue growth and the profitability of perceived AI victims. So investors are unloading them. They're redeploying their capital into AI winners and those immune to AI's impacts.
As UBS analyst Michael Brown recently noted, "We're in this moment where we don't really know what the next 12 months or 24 months [will bring for] companies [at risk from AI]."
The bottom line is, we want to own stocks that benefit from AI (Category 1), as well as AI-resistant stocks (Category 2). We also want to avoid stocks that have a high risk of being displaced by AI (Category 3).
So as AI shifts from a software story to a power story, there's a clear divide for investors...
Some companies will benefit directly from this surge in electricity demand. Others will feel pressure as AI reshapes their industry.
And for investors who focus on the businesses supplying the infrastructure for this boom, you won't just survive the AI boom... You'll profit from it.
All the best,
Corey McLaughlin
Editor's note: Marc Chaikin, founder of our corporate affiliate Chaikin Analytics, warns that a market "tug of war" will begin this month. That's why he went on camera earlier this week to unveil a powerful new tool designed to pinpoint the AI stocks most likely to deliver explosive earnings surprises this year. In his free presentation, Marc names one "double bearish" stock to avoid immediately... and the AI-driven health care leader to buy right now.