This is Tom Rowland Seriice and you're listening to Switched on the B and EF podcast.

AI Data centers are big news these days.

They're also big users of electricity, and with a whole fleet of data centers set to be rolled out across the US, energy sources need to be found to power these behemoths and fast.

The US does have a bit of an energy superpower.

However, it's a wash with natural gas, and that gas is cheap, and the new Trump administration is throwing its weight behind production of the fossil fuel.

So might gas be a domestic resource for powering the new data center load?

BNF estimates that the current data center pipeline could add sixty nine gigawatts of average power demand to the US grid.

That's equivalent of about eight point four percent of average domestic gas consumption in the US last year.

On a regional level, though, the story is far more nuanced, with data center build up plans disproportionately affecting different electricity grids, and given the global shortage of gas turbines, could the required plants even be built in the first instance.

Today, I'm joined by B andF gas market analyst Henry Eaton, and we discussed five from his note fueling the Cloud data Centers bond with US gas, which BNF clients can find a BNF go on the Bloomberg terminal or on BNF dot com.

All right, let's get to talking about US gas and data centers with Henry.

Henry, welcome to the show.

Thank you for having me.

So we're talking about gas and data centers today, and we've been doing a lot on data centers either directly or indirectly in the podcast this year.

But for those listeners who haven't listened religiously to everything we talk about, could you just maybe frame the scale of the challenge slash opportunity as it relates to AI and data centers and the US energy system.

Sure.

So, I think a lot of people have heard and you reading the news a lot about how massive this AI opportunity is, both from like a workflow and techside, but also the infrastructure needs and the money that's pouring into infrastructure.

The information that.

We've been able to see or the information that we use in our research, shows that there's about one hundred and twenty eight gigawatts of data center capacity coming online.

And I just want to frame that this to really show the speed and scale of how much that's growing.

That number from the time that we put out this research, or really at the end of the second quarter this year, that number has already grown to one hundred and eighty four gigallotts.

So the scale and speed that this is growing, we almost can't keep up.

And I think that really just goes to show how much of a.

Pertinent topic this is.

You know, we have twenty two gigawotts or so of live data center capacity, and so one hundred and eighty four in theory is you know, almost almost ten times, right.

That's a massive number.

I mean, the listeners who aren't familiar with you know, how big is a gigawatt, and a nuclear power plant typically has output between depending on its size, like one and two gigawatts.

So we're talking of the scale of like more than one hundred nuclear power plants exactly, which is just an unfathomable amount of energy in a system that where demand hasn't really been growing that much till now.

We also, I mean, and we've spoken about this on the podcast before, but again it's useful for us to sort of recap just for the people who haven't listened to everything that we've done on this.

Where is this pipeline likely to be realized?

Right, Yeah, that's a good question.

So the you know, data centers, you see growth across the entire country.

The scale of that growth varies quite vastly between various regions.

In our research, where we split the data centers up in our data set into five regions, which is essentially the East, which is the eastern seabord of the US, South Central which is Louisiana, Texas, Oklahoma, the Midwest, the Mountain region, which is essentially the Rockies in some of the states in the plains, and then the.

Pacific Coast as a Pacific region.

To elaborate more on where we see that growth.

The Eastern region, I think the Eastern Seaboard encompasses Virginia and all the states that border the Atlantic.

That region houses something a little under fifty percent of all data center capacity that's live today and in that US in the US, Yeah, and that's going to grow to about sixty percent given the pipeline that we can see.

And so that is a region that really really has already dominates the space and will as from what we can see will continue to do so going into the future.

There are a couple other like.

Very interesting dynamics going on where the Texas, Louisiana, Oklahoma area right now, excluding the bitcoin mining data centers, is the smallest market for live capacity live data center capacity in the US.

However it's the quickest growing.

It's growing the quickest, so that's going to become the second largest market for data centers.

And I think the last interesting thing here is the Pacific region, which is Washington, Oregon, California, that was the second largest market live today but will actually become the smallest market in the future.

And so what I like to call is it's a mature market.

It is going to see growth.

It's going to essentially double, but the other regions in the US are basically going to four or five x at the very least, whereas the West.

Coast is really just going to double one.

I think there are a couple of challenges there that's so interesting.

I mean, I think just because you know, we all think of the West Coast and we think of tech, and that's where it's going on, and you know that might still be where a lot of the I suppose the human intellect will be behind those industries, but maybe not the computer hardware.

I think that I think it's just worth pointing out here as well, particularly to people who are outside of the US.

You know, if this what we're talking about sounds like a very local US thing.

I think today about fifty percent of global data center capacities in the US, and I think that ratio stays roughly the same in our projections.

So you know, when you're talking about maybe half of it being on the Eastern Seaboard in the US, that then means that's a core sort of the data centers in the world.

So this is not just something that is a interesting question from a US perspective.

This is a kind of a thing that is a global trend happening in the US.

That's how I would describe it.

Kind of last question before we kind of dive more specifically into your research and your specialization.

But another framing is is how big are these data centers that we're talking about.

Yeah, that's a great question to start off with some numbers.

I think the average size of a data center today is just under twenty megawatts.

And you know, we started the conversation off by talking about how the pipeline is one hundred and eighty four gigawatts, right, The average size of these data centers, particularly in various stages, the ones that are announced but not built yet, so the early stage data centers, they grow to about over two hundred and twenty or something megawatts.

So not only is our pipeline almost going to ten x, but the size or the average size of our data center is going to really increase about the same it's going to They're they're going to mass for me, I.

Mean a two hundred and twenty men data center, that's about a very normal size for like a gas power plant.

Yeah.

Yeah, So I mean that segues nicely into your research because you're from our US Gas team.

Our US Power team has already done a bunch of research on data centers and the implications on the grid, and we had Helen co and Natalie Leamandebrata from that team on the podcast earlier this year.

You're looking at specifically from a gas angle, So as a gas analyst, what's interesting about this?

Looking at the whole data center boom from a gas perspective, I'd say there are two things.

One, the developers and operators are looking to gas to power their data centers, and that's for reliability reasons, base load reasons, right, the dispatchable characteristics of gas that make it favorable to power.

These data centers.

So there's interest in natural gas as a topic from that side.

There's also a lot of interest from the natural gas market participants, the midstream developers, the power utilities, the people who are trading gas every day, who want to know how much impact is this revolutionary technology going to have on my business and where are the opportunities and so naturally, the amount of power that these data centers require is also going to call quite a bit on natural gas, not just because it calls on the power system in general, but also because gas.

Is favorable for empowering these data centers.

I question I have is it makes sense that gas is well positioned to meet this demand, and the US produces a lot of natural gas.

Is some of the cheapest natural gas in the world.

So I can see that that is putting a lot of advantages in the US's hands when it's thinking about its data center development strategy.

And so obviously, if these data centers do depend someone on gas, that will mean more gas demand to what extent is this because I know your team looks at the gas balances.

You know how much is coming in and out of storage, how much is coming out of the ground, how much is being consumed.

So obviously this is going to affect that.

But how much of it is going to affect the infrastructure, you know, the pipelines, the ability to deliver I suppose what I'm really saying is how much of this is just a question.

Yeah, there'll be more gas consumed versus than needing to be investment in the infrastructure around gas because of this.

Yeah, that's a That's an interesting question and a tough one to give a direct answer to.

I'd say, without a doubt, there.

Is no question that it's going to increase natural gas demand.

I think that is a foregone conclusion and is known by people that are that are looking at it and are that have interested in it.

The infrastructure side of things is a more interesting one.

I would say, right now, the US has enough infrastructure.

We certainly have the gas production to get the gas out of the ground, right, So it's a matter of can we get it to where the data centers are and do the data centers have for example, a turbine that can turn that natural gas into electricity.

I think what you'll begin to see.

And part of the reason why I brought the difference between the pipeline that we looked at the time we put out this research in the pipeline now has grown another fifty percent, is to show that I think there will need to be infrastructure upgrades no matter what, especially if all of this data center demand does come to fruition and it's not just companies announcing that they're going to build them and then they don't get built.

So which is which is possible?

Which is one hundred percent possible?

That we've seen pipelines before for different things, and yeah, we don't always get one hundred percent of what's in the pipeline.

And I think the interconnection que problem exacerbates that problem as well, because really the number one priority for these developers to get online as quickly as they can, right, So they're going to submit a data center to every interconnection possibility that they can just on the off chance that they get it.

It gives them an opportunity to build that.

So there might be the amount of capacity that we see could be inflated.

It could not be right, right, right right.

I think, so we know there's going to be more you know, how much is a question, and in a way, some of the constraints on the system, and you mentioned the interconnection queue, but you're also alluding to the need to invest in more gas infrastructure, and if that doesn't happen, then some of this pipeline might not materialize.

But we'd be wrong to say like, oh, well, the infrastructure was fine because they never built as much as they were going to.

But you will say, well, that was because the infrastructure was exactly Yeah, given what we've just said, maybe this is a dumb question.

Expecting to be able to answer this.

It's hard to say exactly how much is going to get built.

But you mentioned that they'll be more gas consumed in the US because of this.

A question I have is, do you mean there'll be more gas consumed compared to if there hadn't been the data centers, which seems fairly uncontroversial because it's an additional source of demand, or do you mean there'll be more gas full stop?

Because if we look out far enough from the future, currently you know gas is displacing coal, increases gas demand.

Let's leave exports to the side, because that is an increasing source of demand for gas.

But so we're just looking at us domestic consumption because the other thing that's happening is renewables are getting built, which which is displacing coal, and eventually we'll start displacing gas.

So in the long run that downward pressure on gas demand, I suppose I'm saying, do we think that data centers more than offset that to the point where domestic gas demand is growing, It just means that it's going to be more than would have been.

That's an interesting question.

I don't have the perfect answer for you.

I think that's going to come down to whether we build out the power infrastructure to connect these data centers to the grid.

So I think if they're connected to the grid, I think that natural progression that you explain, where gas replaces coal and then renewables and batteries start to eat into that gas share, I'd say that you'd see gas's share of the fuel mix begin to fall far out into the future.

Right now, what we see is I think in twenty twenty four, gas represented about forty three percent of.

The fuel mix in the power sector.

In the power sector, right and that has grown in the past couple of years.

So renewables are eating part of that share, but gas is gaining more from replacing coal.

Without this data center demand.

Where gas is growing.

Gas is growing either way, right, and so that's an important thing to note.

I'd also say, like back on your infrastructure question, if data centers can't connect to the grid and they decide to go for gas generation, they're not disconnecting from the grid, They're just connecting to a different grid, which is the gas grid.

So you're now talking not about data centers creating more demand for power from the grid and some of that will be met by gas plants connected to the grid.

You're talking about data centers having their own gas power supplies.

Is that what you're getting at?

That is what I'm getting at, if that's feasible for these developers.

And I don't want to get into the question yet of the whole turbine unavailability, because that's a whole separate story, right, So if they let's say that.

Some listeners don't know about that yet, but there's a whole other story about turbine availability.

Yes, I'd say that because speed is so important and interconnection times in certain regions are so long, that there is a question that these operator and developers are asking themselves, which is, can we use gas to get online quicker?

And it may not be like a permanent solution, Maybe it's just a temporary solution, but it is a question that they're asking themselves for sure.

I mean it's a really interesting thought though, because you know, we have these two parallel energy delivery systems in the US.

You could say, which is the power grid and the gas network?

And I know they're not the only two energy delivery systems, but let's just talk about these two things.

They interact with each other, but they are also independent from each other to some extent.

The picture you're painting is if that first energy delivery mechanism, the grid, is too much of a constraint, then data centers might go straight to the second one and kind of rely on that.

So then that leads me to the question, how does that influence the location of these data centers?

Because I imagine that this the gas grid has more capacity in some parts than others.

Yeah, that's absolutely correct.

We've seen things get more constrained in the gas grid, generally, specifically coming out of some the large production regions as those producers try to meet.

Look a f natural gas to man.

You'd see in certain regions, for example, in northern Virginia, or data center ally, as it's often coined, where a vast, vast majority of capacity is going.

I think I said half of future capacity is going to the east.

About half of that eastern capacity is going to Virginia.

So key, lucky Virginia.

If you do another half a half, right, let's call it a half of us capacity.

Another half is in Virginia, so let's call it a quarter of us.

And then I don't know, an eighth of the world capacity, eighth.

Of eighth of the world date cent capacity is in one little stay.

In one state.

So I mean pipelines there like we see that if let's say all future just for just theoretic all future data center capacity says, Okay, the wait time's too long.

We can get a gas turbine and we're going to connect to the gas grid.

I don't know if that's.

A feasible solution for them, because pipelines there, they already fill up during times of seasonally high demand.

Your your cold peaks in the winter, you're super hot peaks in them.

So those pipelines are full and the gas flows generally north to south through Virginia through through you know, three major pipeline systems there.

Building more infrastructure there, particularly in east in these developed states where land is.

You know, once you get into the Washington, DC area, you can imagine it's probably not so easy to put a pipeline through it.

That's an extremely extremely difficult feat to achieve and takes years and years and years.

So this data center capacity, they're all trying to come online much quicker than you can build the infrastructure if you can get approved for it, right, So there are constraints there for sure, especially if these operators decide to go for like temporary gas solutions.

You know, you can bunch up a small bunch of small gas generators, I think enough.

Helen and Natalie's research that you mentioned earlier estimated and through surveys, found that the interconnection time is seven years, right, Virginia.

That's a long time to wait for a competitive industry like AI.

If you want to come online as quickly as possible, are you willing to wait seven years?

Yeah?

I mean it's not just a competitive industry between different companies, but it's competitive between different countries.

You know, in the US, AI is seen as a strategic industry, you know, with maybe even national security.

And the biggest competitor is China.

And China's grid is it growing and expanding has been for decades, So I mean very naively, and I'm sure someone on our China team will say this as an oversimplification, but you know, I was a little bit of extra demand when you're all already growing x percent a year compared to the US, where both for power and gas, is maybe like you're more at a standing start on this stuff.

Yeah, it's so interesting.

I mean, one of the things that in Helen and Natali's research they've spoken about is that energy ideal locations versus the locations that are ideal from everything else that data center developers might care about.

So connection to the well, obviously everywhere can connect to the internet pretty much, but you know, where are you going to get the bandwidth?

Where is the local skills?

You know, there's a kind of clustering effect, and that's why so many in Northern Virginia now from a power sector perspective, if you didn't care about any of that stuff, there are regions of the country where the wind resources are so great, power prices get solow.

There's a bunch of electricity that's just waiting to be kind of mopped up by a data center if someone would just build one there, And there's the reasons why they don't, but they might do one day when these areas like Virginia just become saturated.

So I'm curious to know what is the gas equivalent of the west of the Rocky Mountains, Because if you're saying that the kind of the gas delivery grid is constrained in many ways, then there must be regions where you have cheap gas that can't get out.

They're just waiting for someone to eat it up.

So, yeah, where is that in the US?

Yeah?

Well, for the gas, that's the Permeuan Basin, which is an oil production basin in West Texas and New Mexico.

Essentially, the drilling economics are favorable to drill oil and gas is a byproduct of that, and so they actually have all this excess gas that oftentimes they can't even get out of the premium basin in the They have to flare it and sometimes to sell it it's negative prices, so they're paying someone to take it away from them.

Right.

Wow, So if you could get paid to consume energy to power your data center, that's a pretty good deal.

Like you said, though, there are a lot more factors.

That these developers need to consider before they go to that route.

I'd say regardless, Texas is a pretty interesting market, probably pretty favorable for new builds right now.

I think that's why back to one of my first points was that Texas and Louisiana is essentially the fastest growing market in the US because you may not have negative gas prices if you're near Dallas, but they're going to be pretty cheap.

Right Building new infrastructure there as much easier.

They've been doing it for years.

They still do it.

We build a lot of infrastructure, particularly to meet new which look finnatural Estiman and so there's more pipeline capacity being built to carry that gas out of the Permian, and it has Internet exchange points.

So it's like maybe you don't build directly in the Permian, but you build trice even if.

You build close to it.

I think you get some of those benefit fits.

Anyways, so in a world where gas is the defining fuel for data censers, which we you know, we may well be entering into that kind of tips the scales in favor of Texas being where a lot of this stuff gets built.

Yeah, I think so.

I think for flexibility and look, I mean interconnection times and Texas are still quite long.

But if you can get your hands on a gas turbine, it's a pretty favorable place today.

If you can get your hands on a gas turbine.

Let's unpack that a little bit.

Why is that an if?

Well, it's an if because the suppliers, the builders of these gas turbines have backlog order sheets for years and I'm not an expert on this topic at all.

Our team is doing research with another BNF supply chain team to put out more in depth research on.

What's going on and what's going on there.

It was when we were creating this current research that's trying to estimate guests demand and power demand and where they're going in location with infrastructure, that we were like, well, maybe they can't even get them.

It's so interesting because there's all these different ways that data centers in the future could be powered, and you know, and it could be that some things meet the demand in the near term and other things meet them in the long term, But it seems like every possibility has it's like an obstacle course between say, you know, renewables and gas.

So you have two people racing on an obstacle course, but they face different obstacles.

So renewables, it's like, grid constraints are a bigger issue.

There's the intermittency that's that's a challenge that needs to be resolved.

With gas, you don't have you know, if you're going on site, you don't have grid constraint issues.

There's can the gas network support it?

You know, there's a location, right, I guess that replies renewables as well.

But then there's like can you get the turbine?

Yeah, So it's like there's this race between these options, and there's unresolved issues in both pathways, and maybe which issues get resolved quickest might determine what the outcome is.

I think that's exactly what's going to determine what the outcome is.

What I tell people if I'm talking to friends or family about work and they ask me about it, I always say, we can only build data centers and therefore develop AI as quickly as we can power these things.

Yeah, and so it's very much a power constrained industry.

And really in a time where you don't want.

It to be constrained, especially if we go back to it's a competition between companies, but it's also competition between countries.

And all of a sudden, you're years and years old.

Infrastructure is what's preventing you from being able to get a leg up.

Maybe it's an interesting point.

I mean, in our world as benf analysts, so many of the questions have changed in the last few years, from you know, what's going to decarbonize the energy system to now, what's going to meet all this new demand?

And the cliched answer which I always found frustrating when people would talk about decombonization and say, oh, we're in favor of all of the above, which for me, if someone said that a conference, I was like, that's code for you don't have an answer to the question, because there's surely more you can say than all of the above.

But I find myself thinking about, you know, how we power data centers and you know this pressing need to meet demand is maybe it's all of the above.

But it's interesting that the US government currently is dialing back support for certain types of energy without necessarily turning the dial up for others as much as they might like you to think, or like might like the fossil fuel industry to think.

So the foot is not necessarily on the accelerator for the energy system as much as maybe one might expect, given how much central importance this as as a topic.

Yeah, I want to talk about one other kind of strategic topic relating to gas, which is liquefied natural gas energy.

You already alluded to it a little bit, but that is something that the current administration, and to be honest with you, I mean, the previous administration kind of flirted and chopped and changed on where it stood on energy a little bit.

But certainly at the moment in government has a consensus that that you know, the US destiny is to be an exporter of LNG and that will be not just lucrative for the kind of but it will also serve the country's strategic interests.

And so that has kind of been a source of growth for US gas demand as the export terminals have developed.

So how much does this potential growth in demand from data centers compare to the potential growth in gas demand from LNG.

Yeah, that's a good question.

I'll start with staying the stage for how big LNG is in the US right now.

As of September twenty twenty five, LNG exports are about sixteen billion cubic feet per day.

To put that into perspective, it's around I think sixteen percent of average domestic gas production or just average.

Demand in the US.

And it's growing.

Our estimates is essentially going to double by twenty thirty, that's four or five years from now.

For wow, double twenty thirty.

For the next five years, it's going to double to thirty, which is going to make it the second largest and that's you know, thirty bcfd at the peak in the winter in twenty thirty is still going.

To be up there right right.

It's going to become our second largest demand source in the US.

And just to remind you, I mean, even like a little more than ten years ago, we were importing energy right right, and now we're the world's largest exporter of LERG.

So's it's going to double, right from fifteen sixteen bcfd do around thirty by twenty thirty.

We have all sorts of new energy terminals coming online, and they've been planning in the works no matter what administration you look at, and there is a push to use it as a bargaining chip for trade deals and whatnot.

So that's our estimate for look a financial gas demand.

It is the greatest demand driver in the US for the US gas markets and really puts the most pressure for new production as well.

If we look at data centers, our estimate based on the pipeline that we could see two quarters ago is almost seven billion cubic feed per day.

So we're talking about LNG growing fifteen bcfd doing cubic feed per day in the next five years.

And I'm saying data centers, as big as they are, are going to be seven billion cubic for you per day, and that's probably later than the.

Next five years.

It takes a while to build.

These things, so it's not as much, but it is in the same ballpark.

It's not a rounding era.

It's not insignificant.

Seven bcfd is not insignificant at all.

That's a number that people have to pay attention to.

And again, the pipeline's growing, right.

I said it was one hundred and twenty eight when I looked at it, and now it's one hundred and eighty four.

So it's interesting because as I reflect on this, you know, I mentioned earlier that the US is at a standing start when it comes to sort of energy growth compared to China, which has been growing for a long while, and so you know, naively you might say, well, what's a little bit more growth, But when you look at it from that perspective of gas and LNG growth, actually the US is not at a standing start.

There's this huge trend of growth already happening, and then maybe there's something to be said for piggybacking off of that.

I mean, do you see a scenario where these pipelines and infrastructure that is being built to take gas from where it's being originated US to bring it to LNG export terminals that might be where also there's going to be a new kind of clustering of data centers, because yeah, it's just taking it, you know, taking advantage of where the highways are.

Yeah.

One of the things I miss when talking about LNG for those who maybe aren't aware, is all of that demand.

All these export terminals are on the Gulf coast, right, They're in Louisiana and they're in Texas.

So to get that additional fifteen billion tw week feet per day of gas that they're going to need by twenty thirty.

We need to build a good amount of pipeline capacity.

And these are pipelines that have been planned for many years and are coming online now and we'll continue to come online incrementally through the next five years.

So we're building quite a good amount.

Of pipeline capacity in Texas, in Louisiana to get gas out of the Permium basin, which is favorable gas prices, to bring those to LNG terminals, the Hainesville production region, which is in eastern Texas, and Louisiana, which is more expensive gas that also feeds into.

A lot of those terminals there.

So we're building you know, essentially maybe just under but fifteen b cfd of pipeline capacity as well.

The supply has to get to the term end somehow, right, So I think it goes back to my point of Texas being a favorable place moving forward.

It's going to say, all right, lead to Texas, lead to Texas in the gas industry and.

Maybe in the gas powered data center yeah field as well.

I mean I realized, well, maybe you should say all pipelines lead to Texas.

Yeah, so even you know, because I don't want to oversimplify this and say that if gas wins out as the thing that's going to fuel this data center growth, then they're all going to be in Texas because we talked about how there's you know, growth potential on the on the East coast, et cetera.

But it does seem that there's a real opportunity there.

I think there's a real opportunity there.

They're not all going to go there because again, if they all go there, you're going to have the same thing.

That's you're going to have, right, they just have to spread out.

Really.

I think for a long time Texas was underdeveloped.

It was the smallest region.

Right now for a live capacity, again, it's going to be the second largest.

So it's kind of like there was this way of bitcoin mining facilities and maybe that was the foreshadowing.

Yeah, you know, bitcoin mining is a bit simpler than AI data centers.

You don't quite have many constraints.

Yeah, and bitcoin mining they have the luxury of not needing or necessarily caring as much about for example, wait and see it serve your application on your phone or someone's computer.

Right, They can go to the Permium basins actually are the areas with lots of solar and wind generation in West Texas and say we'll just take that power to mind bitcoin.

So there are more considerations for someone trying to develop, for example, a large language model.

This has been really interesting conversation and this picture is growing in my mind of this obstacle course being run in parallel, and you know, one of the tracks on that course is the gas pathway, and it's really fascinating to see, you know, what the future could look like and how that changes the dynamics.

And we are internally, we look at all of this collectively.

You know, it's not just Henry puts his head in the sand and thinks about gas and our power team just thinks about renewable as we're looking at it all together.

But I think there's some really interesting competitive dynamics and different hurdles they face that we're going to have to be keeping our eye on.

And so there's been a really illuminating conversation.

Henry, So thank you so much for joining today, Thank you for having me.

Today's episode of Switched On was produced by Cam Gray with production assistants from Kamala shelling.

Bloomberg NIF is a service provided by Bloomberg Finance LP and its affiliates.

This recording does not constitute, nor should it be construed, as investment advice, investment recommendations, or a recommendation as to an investment or other strategy.

Bloomberg ANIFF should not be considered as information sufficient upon which to base an investment decision.

Neither Bloomberg Finance Lp Nor any of its affiliates makes any representation or warranty as to the accuracy or completeness of the information contained in this recording, and any liability as a result of this recording is expressed.

Leek is quiet

MHM