How power players can navigate energy shortages for U.S. data centers

Invest Insight
9 followers
1 Video View·Dec 10, 2024
Stephen Byrd, head of Morgan Stanley global sustainability research, joins CNBC's 'The Exchange' to discuss the scale of the power issues generative AI is facing, energy companies he sees opportunity in to help address the energy shortage for U.S. data centers, and more.
Let's bring in Stephen Bird.
He's head of Morgan Stanley's Global Sustainability Research, a topic, Stephen, that has everyone's
attention.
So do you think we literally need to change these crypto mining sites into data centers
so we don't have an electricity shortage?
It would be a big part of the solution to get enough power to these AI data centers.
The problem is it's a very high class problem, but these Bitcoin mining companies have two
very attractive decisions.
They can use their power to mine Bitcoin at record profits, or they can get paid very
handsomely to convert those essentially to data centers.
So it's frankly a little challenging to figure out which direction they'll go.
So far to date, the companies in the Bitcoin world that have converted, like Core Scientific,
have been richly rewarded.
You can take a look at their stock price chart.
Core Scientific is a good example of what we think could happen.
But the new price of Bitcoin makes things more complicated, and without the power from
all of those Bitcoin sites, we most certainly will have a shortage of power for data centers
in the U.S.
Yeah, how much of a shortage do we actually face or a bottleneck are we looking at?
Yeah, I'm a power nerd, so I'm going to try to turn the numbers into something that's
relatable.
The total shortage we see through 2028, absent some interesting solutions, and there are
some in the stocks around that we like a lot, but the shortage is over 30 gigawatts.
Now, to put that in context, the greater Philadelphia area would be three gigawatts.
So we're talking about a huge amount of an issue.
And I think many companies in the space are aware of this, and now they're looking for
deep bottlenecking solutions, solutions that will get power to these data centers more
quickly than the standard sort of power grid interconnection process.
An example of that is Bloom Energy's fuel cells, as well as converting crypto sites.
Yeah, so what so OK, I'm a company that needs a lot of power next year.
If what you're describing is going to take place, shouldn't should prices already be
surging for for electricity for everybody?
You know, the price of power is going to rise.
The utilities are very aware of that risk, that the price of power might rise for everybody
else.
And so those utilities are quite careful about that.
In some ways, that does slow down the pace of these data centers getting added to the
grid because you want to make absolutely sure that when you do so, you're not hurting other
customer class.
So that's part of the slowdown is making sure that everybody is protected for those data
center companies or the hyperscalers, the Microsoft's of the world.
Those companies will be willing to pay a large premium for the price of power to get connected
earlier. We've already seen some signs of that.
Some of those deals have been at multi hundred percent premium for the price of power.
I think we'll see more of those.
Yeah, I think you and I both know that the day that consumer bills start increasing because
of data center demand is the day that there is a massive political problem on our hands
and a problem that would need to be solved urgently, that to your point, you can't really
bring some of these assets online that that that quickly.
So what are some of the other options and how let's say they say, OK, I'm going to turn
to Bloom Energy.
I mean, is that really can they scale enough to meet that potential demand?
They can, though even Bloom is not enough to fully meet the needs.
So going back to this greater than 30 gigawatt shortfall, Bloom Energy, we think you get
up to two gigawatts a year fairly quickly.
So a big part of the solution and it is pretty valuable.
So, for example, if you were not to use something like Lumi just went to an average U.S. grid,
it used to be two years, three years to connect to the grid.
Now, in many cases, it's four or five, seven years, sometimes 10 years to connect.
So you need something faster.
But if I tally up all the sort of debunking solutions, Bloom could be seven, eight gigawatts
over the next several years, which is would be tremendous for that stock.
The crypto world, if we saw a lot of conversions, could be above 10 gigawatts.
But I doubt we'll see that many because the price of Bitcoin is so high.
We'll see quite a few nuclear deals where you build data centers at nuclear plants and
you also agree to build new nuclear as well.
So nuclear is going to be a big part of the solution.
And we will see some natural gas solutions.
So, for example, in West Texas, where there's a lot of gas, you could see new gas turbines
powering very large data centers.
Absolutely.
