The huge demand for AI is creating a huge demand for energy, which could strain the U.S. energy grid.
“When it comes to AI deployment, we are likely to face an energy bottleneck, requiring us to either slow down data center buildout and (possibly) GPU consumption or significantly accelerate deployment of the energy resources to power them,” Canaccord Genuity managing director George Gianarikas wrote in a recent note.
Gianalikas said there are plans to add more energy capacity with solar and, in the long term, nuclear, naming NuScale (SMR) as a potential beneficiary. Gianalikas also believes companies like Ameresco (AMRC) and Fluence Energy (FLNC) could benefit from increased energy demand. He's also keeping an eye on Generac (GNRC) in case more outages drive up demand for generators.
For more expert insights and the latest market trends, click here to watch the full episode of “Ask the Trends.”
This post was written by Stephanie Mikulich.
Video Transcript
Well, power grids across the country are feeling the pressures of the AI boom.
The next guest says that in order to gain more, you have to sacrifice something.
We now bring in George Generica, Managing Director of Canon Court Ingenuity.
George, thank you so much for coming.
Well, this is something we've talked about a lot, right?
This huge demand for AI, and with that comes a huge demand for electricity.
But of course, there's also a lot of discussion about the security and resilience of the national power grid and whether we can keep up.
So, in your view, where is the problem?
Well, thank you for having me.
So we wrote a memo last week, inspired by a report from the Electric Reliability Council of Texas.
That's the organization that runs the power grid there.
Well, like other guests, we've probably written quite a bit about the potential strain that ID centers could put on the energy grid.
But the reason we value Elcott's study is that it's one of the few studies that we've seen that actually tries to quantify in detail what the impacts will be.
As you may know, this study was prompted by a bill in the Texas Legislature that called for an analysis of the impact that data centers and other such facilities have on the reliability of the power grid.
The 2021 winter storms significantly impacted the power grid and caused deaths, and the study found that the curve margin, the buffer they are trying to create to be able to handle peak loads, will turn negative in both winter and summer in 2028, 2029 and 2030.
And that's clearly not what they want.
So we either need to add a lot of energy and generating assets to the grid or the demand for all of the AI-related hardware that people are buying goes down, and as I wrote in my memo, something has to be sacrificed.
Well, the latter isn't necessarily going to happen, is it?
At least not in the short term.
Are there any plans in the works?
I mean, I think for years now we've been talking about R CT, specifically the issues with the power grid in Texas due to other winter storms, heat waves from winter weather.
So are there any concessions, any compromises on that side?
So, over the longer term, we have plans to add significant energy generation assets, whether that be solar or battery storage technology.
And we're very excited about the potential of nuclear technology, particularly small modular reactors, to help generate baseload clean energy across the country.
And we have a company called NuScale (SMR), which is hoping that over the next few years it'll start winning some contracts to build energy infrastructure for data centers in the U.S. And this seems like a regional issue.
I spoke yesterday with another analyst who covers the utilities industry.
And she said, well, it depends on where you're talking about.
Maybe places like Texas will have issues, but other places have plenty of capacity.
I guess the question is, will data center operators want to locate in places where there is more capacity?
Yes, that's a very valid point.
So Texas has a lot of cheap green energy.
So companies are putting their data centers there, you know, in the traditional Mid-Atlantic areas, including some locations in Ohio.
Well, in a way, it is geographically specific.
But eventually, as we add more data centers, as we add more electric vehicles to the power grid, it will start to put a strain on the power grid.
And I found this particular study alarming because it shows that potential negative reserve margin scenarios could be very dangerous for the grid and the country.
And as I said, we followed up on a note that we wrote citing a post from yesterday or the day before yesterday where Elon Musk and someone else were going back and forth on X and basically said that these companies are buying a lot of AI-related hardware.
I don't know where to put it, so that's also causing problems.
This means that simply burying energy-generating assets in the ground is not enough.
It's also important to figure out where to put all the AI-related hardware you're buying.
And it seems like there could be air pockets as ambitious plans to build AI data centers run up against the realities of building them, buying energy generation assets, getting permits and securing the workforce.
These are all real-world things that can slow down AI-related work, right?
Needless to say, even though it is a regulated industry, this could have an impact on electricity prices, and we are already seeing increases.
George, as you said, you're looking at it from the perspective of who can benefit from all of this, and as you just mentioned, not just where the bottlenecks are going to be, but you mentioned one idea, one company that could benefit, but are there others that you're looking at here, like grid resiliency operators or alternative power operators?
Certainly, we're talking about a company called Maresco, which is essentially a clean technology integrator that should benefit. They haven't seen the demand from the data center yet, but we're starting to see signs of that with a company called Fluence that we're talking about from the same perspective. They haven't seen the demand from the data center yet, but we're starting to see signs of that.
Also, if you are buying a generator for your home, make sure it is safe if there are problems with the power grid.
Buy Genack, GNRC.
They manufacture generators that essentially act as backup power sources for residential and commercial industrial uses.
George, it was good meeting you.
Hopefully something will be announced, possibly in part 3.
Um, next note, next update on this.
This is definitely a topic that's not going away.
So we keep talking about it.
Thank you for inviting me.
