How AI is changing the natural gas industry

0:00The demand on the grid is such that the grid needs to upgrade a lot of its network and that obviously takes a period of time. And in the data center sector, speed to market is all. Some would say now it's speed to power. So we can't afford to wait 5-10 years whilst the grid eventually builds the infrastructure and is able to supply the power. What makes our project unique and different from other LNG plants along the U.S. Gulf Coast is that we are intentionally employing carbon capture and sequestration in the design and the construction of our project. So we'll be the first in the U.S. to do that.

0:38It's easy to see a change and think that this is a pivot, but the gas industry in particular has been quite resilient. You could have thought the same thing after Russia invaded Ukraine, but has demonstrated a lot of resilience from a gas and energy perspective.

1:00Hello and welcome to The Energy Gang, a discussion show from Wood Mackenzie about the fast changing world of energy. I'm Ed Crooks, and this is the second of two special episodes that we're bringing you from Wood Mackenzie's conference on gas, LNG, and the future of energy. In this episode, we're going to be talking about surging electricity demand for AI and what that means for energy security and for emissions. To discuss that, I spoke first with Neil Kalita, who is the Senior Director for Global Power and Energy at NTT Global Data Centres, which is one of the world's leading data centre developers.

1:32And I asked him what he was seeing in terms of rising energy demand. What we're seeing in the data center sector as a whole is, in North America specifically with the AI, the onset of what we call gigawatt scale data centers, which are much, much bigger than we've ever built before. And as a result, need a lot more power. And the demand on the grid is such that the grid needs to upgrade a lot of its network. And that obviously takes a period of time. And in the data center sector, speed to market is all. Some would say now it's speed to power. So we can't afford to wait 5, 10 years whilst the grid eventually builds the infrastructure and is able to supply the power.

2:13So we've started looking at on-site generation and the only fuel that really provides that 24-7 firm capacity which data centres need is gas. So the various technology providers who sell gas turbines, gas engines are all pretty much sold out now and the sort of lead-in time for those types of technologies are a good couple of years as well. So I think where we're coming from from a gas supply is really interesting because our core business is building data centers and operating them, not building power plants. And from my perspective in energy, the commercial angle I'm looking at is actually, okay, what is the technology that we're going to get that we can get access to in time?

2:59And then who is taking that commodity risk? of the supply. We're kind of toing and froing at the moment internally about whether we should be the ones responsible for buying the gas supply, or we should really just give it to the IPP plant operator, let them have that as upside. But with the kind of flip side of the coin of that, we want price certainty for that term. Yeah, no, it's really interesting. So you're sort of, if you like, going further and further up the supply chain, right, that you first of all had to start thinking about power supply and where that was coming from, and now you're having to start thinking about the gas supply that comes to the power plants to provide the power that you're using.

3:38And also, from a site perspective, is there a gas connection nearby? You know that often. And the siting of those power solutions is a really interesting question. Do we actually want it within our red line boundary, which takes up space, means we can't build as much as we'd like to, or do we actually co-locate as some of our clients and competitors are doing or do we work with an IPP and come up with an alternative strategy for an off-site private wire type arrangement? Right, because I feel like probably, let's say, a couple of years ago when the AI boom was just sort of starting to gather momentum and people were excited about the technology and people knew there was going to be a need for a lot more power, At that point, still, there was a lot of scepticism about on-site power and people saying they didn't want to get into the power generation business if they were data centre operators.

4:34And people didn't want to take the risk of having stranded assets. If you build a power plant to support a data centre and then, you know, what happens to the data centre, maybe it doesn't need the power long term, maybe because it gets a grid connection in the future or maybe because the AI industry changes and people don't need that capacity. anymore. You're saying those fears have kind of fallen by the wayside. Is that right? Because as you say, speed to power is everything. People just need that connection as quickly as they can get it. I wouldn't say it's fallen away. I think we've just become, we've got a better understanding of the economics and the commercial realities of that from the energy sector.

5:13And also, So, you know, with our own internal risk assessments worked out or formulating what we believe to be the most risk balanced strategy for us. Where this has come into play more so is the fact that there's been such a proliferation of gigawatt scale campuses. And as a result, the grid has got more taken, taking a lot longer to come online. The fact of the matter is the data center will always need the grid. There is some basis for the fears that the gas power plant might be stranded. And what we've invested a lot of time and thinking on is to actually say, OK, how can we develop an equitable structure with the IPP and make sure that the term that they need is aligned to what we need and gives them the flexibility, revenue stacking and changes in use case of the power plant over time as the data center evolves.

6:07And the fact of the matter is, you know, in the conference yesterday, there were some concerns about whether the power will actually still be required. And I think as the grid gets built out, the power will still be required because the GPUs are increasing massively in what we call density. So the amount of power one of these chips needs is, I wouldn't say exceeding Moore's law, but it's getting there. So that requirement is going to be driving the power need. So that's really interesting. So in other words, sometimes you hear talk about on-site generation being a kind of a bridge solution. In other words, as you say, a grid connection is ideal.

6:46The grid has all kinds of advantages. It's more reliable. It's probably lower cost, may have lower emissions. For all of those reasons, in the long term, every data center operator we talk to always says grid connection, that's what we want. As you say, that's not always possible, not always possible on the timescale that people want it. So as a second best option, you go to on-site power, and then you can have that kind of bridge where you run the on-site power for a while as your main power supply until you get the grid connection. But what you're saying, what, is that then when you do finally get that grid connection, you're basically going to be using both.

7:23So you'll keep using the on-site power, and you'll have the grid connection as well because your power demand is going to be so great. Yeah, exactly. I think, you know, the first point I pick up is that you made was really interesting is which is the more reliable network, gas or the grid, right? And I think, yeah, with the increasing power density, the energy consumption that's done by the data centers is largely driven by how the clients use the space. And that we're all beholden to, you know, the foibles of AI and the use cases around that. I think with the gas, what's really interesting in having on-site generation is it does give you flexibility as an operator to, as you say, initially during the construction and development phase, bring the data capacity online quicker.

8:12And then as the grid gets built out, we can then continue to use that to facilitate the gas to facilitate the build out of the rest of the campus. and then once the grid is fully operational and the client is fully installed then depending on what the gas is doing what the what the AI client is doing we may see scenarios where grid price the cost of electricity from the grid is prohibitively more expensive and so we might then actually say well actually the on-site generation we might flip to that and an energy mix of a data center is something that we're really having to start to consider a lot more.

8:52Right, because you talked about gas as essentially being the only solution for 24-7 power, which is what the data center users need. You hear talk about, well, there's maybe a mix of storage plus renewables, some combination of wind, solar and storage that could provide 24-7 power. Do you think those kind of solutions don't really work? So, I mean, renewals for storage eventually could, but I think there's two factors to that. One is, can you build a solar farm and wind farm big enough to meet the demands of the day center? And then can you get a battery long enough, long enough duration to meet that requirement for that 24-7?

9:36Because the wind doesn't always blow and the sun doesn't always shine, obviously. I think, you know, there will be a space for renewables and storage, mainly, I think, through corporate PPAs rather than actually having a solar plus storage asset next to a data center. And those corporate PPAs, power purchase agreements, that's a very well-established model, right? There's been a lot of that. Those big tech companies in particular, they've been the big buyers of energy, renewable energy under those PPAs. They have. We all have. And, you know, that's an important message to get across is the data center sector, wherever we go, we are almost a catalyst for investment in renewable energy because we have that demand.

10:14And with the forthcoming changes in the greenhouse grass protocol, where cross-border PPAs might not be considered as favorably, I think we are going to see, you know, a greater data center demand in countries where, you know, some of my, I mean, for example, NTT, I actually buy all the renewables. energy PPAs for GDC at the moment, we've always advocated this one principle of in-country for country. So we won't buy, for example, Spanish solar to decarbonize German consumption, although we could. So yeah, we don't have to unwind our positions from that at the moment. And you say you could do that at the moment, but under the changes in the greenhouse gas protocol, which is what essentially the framework for emissions reporting, you wouldn't be able to do that in future?

10:59It would be. Yeah, I don't think it'd be considered very favorably. I want to go back to a question you raised earlier and try and get your answer to it. You said there's an interesting question about what's more reliable, the electricity grid or getting gas supply from the gas pipeline network. What's the answer to that? My personal view, and this isn't NTT or anything, but my personal view is I think gas is more reliable. I would say that in a conference today, but no, I think so. Some of the data that we've seen for particular countries bears that out. And also, you know, logically also makes sense because, you know, with the electrification of everything, the low carbon transition, et cetera, the pressures on grid are going to increase.

11:39That obviously relates to two four parts. One is the megawatts available and then also the price of those. And I think that's where we're going to see the challenge and why on-site generation isn't just a bridge for the short term. I firmly believe that data center operate the smart ones. We'll actually see it as creating an energy mix to make the energy cost competitive. And also to take into account the vagaries of what's going to happen with the grid. Like, for example, climate change. If temperatures get too high, you know, overhead lines will start to fail. Blackouts, et cetera, which aren't great for data centers.

12:12So having both our on-site resilience, which is normally gensets, and then our own power plant, access to our own power plant, makes a lot of sense. And one of the key service level agreements that colo operators like us operate under our clients is our availability. And being available 99.999%, five nines, we can only be off for 20 minutes a year. If the grid starts to start getting more flaky, having that access to the gas is really useful. Right. And that's a really interesting point, I think, which is worth focusing on, which is this question of the availability of the data center. And as you say, this five nines reliability, because there's a lot of talk about data centers having the ability to be flexible load.

13:05and governments and regulators around the world, certainly in the United States, are very much focused on this idea that if data centers can be flexible, they can get connected more quickly, they can be an asset to the grid because they can reduce their load at times when the grid's under strain. You're saying that actually that's not very easy to do, is that right? And actually you do need to be essentially a flat load and a very steady, consistent power supply? Yeah. Pre-AI, I would have said yes. But I think the whole flexibility question is really, really interesting. I think because fundamentally there are now two uses for the data centers, cloud, which has always been the classic 24-7 high and flat load, and then the AI loads, which are incredibly volatile.

13:54And that's particularly for AI model training, is it? So if you have a training run, and then when that training run ends, kind of suddenly your power demand goes right down. That's one use case that we've seen. And that's actually the really fascinating challenge of the AI data centers and the energy uses. As more use cases of AI become more popular, let's say, we don't know how the AI is going to use the energy. So the point I was making was that with the AI loads, there are periods where the data centers are at a lower level of utilization, just because the model doesn't need it or the use case doesn't need it, whereas in the cloud capacity, we're always on.

14:36So when we respond to an in-North America entity, we've entered into what's called the Voltus program, which is the demand-side response. So I think it was last year we actually supported the Virginia grid by coming off grid, running on our diesel generators for a couple of minutes, a couple of hours, to allow the grid stabilize and keep the lights on for the community and the county. I think in the rest of the world, it's a lot more challenging. And that's largely because of, in Europe, the permitting. What we're trying to do, work with regulators at the moment is to say, look, if we are turning up for the grid, i.e.

15:15we're coming off grid and running on our diesel generators, you can't then penalize us for violating our environmental permit because we're actually doing a service to the community. The flip side of that coin is that, okay, so you're doing that, but you're also meeting our SLAs by providing a stable grid and by coming off. So, yeah, it swings and roundabouts with all these things. I think the decarbonization piece is really interesting as well because running on diesel gensets isn't exactly sustainable. Yeah, good point. So how do you do that then? A lot of these companies, the big hyperscalers in particular, still have these ambitious goals for reducing emissions.

15:52How do you reconcile that with increasing gas consumption? Yeah, so one is the specification of the gensets, so scrubbers and all the like. And then the other is alternative fuels. So rather than diesel, maybe HVO. And then, you know, going back to the on-site generation, if we do have a gas power plant there, then there are schools of thought that are saying, well, why are we even installing the diesel gensets when the gas power plant could do that for us? So how is this going to play out, do you think, over the next few years? It seems to be an industry that is characterized by an enormous amount of uncertainty.

16:25It seems sort of unprecedentedly uncertain in terms of, at the one end, there are people saying we are on the verge of AGI, artificial general intelligence, something which is sort of an unprecedented innovation in terms of human history. On the other side, there are people who say, well, now it's emerging that the commercial case for AI is absolutely unproven. It's not at all clear that the returns are going to be there. We may be in a period of overinvestment. And this is essentially a bubble that is going to burst at some point. How do you cope with that degree of uncertainty? I think that uncertainty is a result of the fundamental changes.

17:12It's a revolution, as some people are calling industrial revolution. And revolutions, you know, things break and get recreated. And I think, I can't remember who quoted it, but before you can create, you first have to destroy, right? So, and I'm not suggesting that in any shape or form. But what I am suggesting is, as you've alluded to, the transition is a step change. It is challenging, made more so, I'd say, by the time frame, the compressed time frames that our sector operates under. And I think that's a good challenge for everybody to get used to because, you know, as I was just saying in there earlier about the just in time versus just in case scenarios that we're dealing with the volatility from geopolitics.

17:54similarly you know we need everybody to moving into a load transition we need you know to embrace ai um for for all of the potential applications and benefits it may bring uh and then obviously be more very cognizant of the impact that uh could arise from that as well from that unprecedented change and and and some would say uh not me or entity but some would say um reckless speed of change. I can't possibly speak to that, but it is coming. And bottom line, as you say, volatility is just something we're going to have to get used to. Yeah. And then one thing I picked up today on the conference as well, which I really found interesting was good volatility and bad volatility.

18:35That's a really new one on me, but I really enjoyed that just from a kind of conceptual piece. It's like, oh yeah, that actually makes a lot of sense. And that for me is the juice of coming to these sorts of conferences, particularly when it's commodity-based, is you pick up on that and take it back to the DC sector. And the fact of the matter is power and energy in data centers is becoming more complex, it's becoming more diverse, has to be decarbonized, and it's under massive amounts of scrutiny, not just from the regulators, but from the communities that we operate in. And at NTT, we're very, very clear about our commitment to the communities that we make when when we start building these things.

19:15Well, I hope you'll come back again on the Energy Gang very soon to talk about how this volatile world is changing, how this very dynamic sector is playing out. And how we did it as well. I can see now, though, that lunch has started, so we should break off for now. But it's been great talking to you. Neil Kalita, thank you very much indeed. Thanks for having me, Ed. I spoke next to Keith Shoemaker, who is the Chief Commercial Officer at Coastal Bend LNG, and I asked him first of all to describe what it is that his company does. We are a project developer, and we plan to build a LNG liquefaction plant in the Corpus Christi area of Texas on the Gulf Coast.

19:52And what makes our project unique and different from other LNG plants along the U.S. Gulf Coast is that we are intentionally employing carbon capture and sequestration in the design and the construction of our project. So we'll be the first in the U.S. to do that. And then in addition to that, I plan on procuring low methane emissions upstream supply to deliver what should be the lowest carbon intensity LNG in the world. So what's the commercial case for that then? Because, I mean, obviously, in theory, you might be able to get a premium price for LNG that has low associated emissions. In practice, it seems quite difficult to do.

20:41Yeah, no person who is a consumer of NatGas, either for residential reasons or for industrial or power generation, wants to pay a premium for the environmental attributes. So that's what makes it difficult to include it in your project. So we have a location which is not a pure greenfield project, but because we're partnering with Enbridge, who has the Ingleside Energy Center, they already have the dock facilities there and the piers that they have agreed through an arrangement to let us use. It'll be a combination greenfield for the liquefaction train and the carbon capture and the storage, but it'll be brownfield for the maritime facilities.

21:37So it will keep our cost lower, so we'll be able to put the carbon capture equipment on the project and not charge a price or a premium over other LNG offtake volumes. Oh, that's very interesting. Right. So it's not that, as you say, you don't need a premium price. You can match the price of your competitors, but with lower emissions. Exactly. And we will benefit from the IRS 45Q tax credit program. So that will cover the cost of the OPEX for the transportation and the sequestration of the carbon. So what kind of customers do you think will be interested in buying this gas with lower associated emissions?

22:26Primarily right now is the European utilities that will be subject to EUMR. The methane emissions regulation, right? Correct. Yes. There are primary customers right now, as well as utility and LNG merchants from the Far East. your big LNG players there, because even though the countries of the Far East haven't come up with their own individual EUMR-type program, they're watching Europe very closely and have this clean program that they're trying to pattern, let's just say, around EUMR. So I think, you know, big energy importers of LNG who have a mandate to lower their carbon footprint would be our buyers.

23:14And what about hyperscalers, the big tech companies? There seems to be a lot of tension in that industry right now because those companies are growing very fast, they're investing a lot in new data center capacity, and therefore their energy consumption is rising very sharply. And a lot of that energy consumption is in natural gas because that's the way they can get 24-7 reliable power, but they also have very ambitious goals to reduce their greenhouse gas emissions. And reconciling those two objectives is not easy for them. Are you able to provide part of the solution for that? We could if they were in a international hyperscaler or with international facilities, and they wanted to use low-carbon natural gas to generate the electricity.

24:08So they would import it in the form of LNG. We would be certified low-carbon LNG, which they could source for their power generation, let's just say in Europe, in an individual country. So I think that they could be a customer directly or indirectly, much like in the U.S. where the hyperscalers are trying to lower their carbon footprint of their fuel stack that generates electricity for running their data centers. Have you had any conversations with hyperscalers about that? Do you know if they're interested? I do, no. I mean, you can't help but not read about how they are, you know, trying to procure, let's just say, low emission energy.

24:57I have not seen that yet. I mean, I expect it over time, especially if they have facilities in countries that depend on LNG. and if they carry on with what their intentions are in the US and on a global scale, then they'll only want to supply the power that generates the electricity for their locations around the world with low carbon LNG. So how do you think that market for gas with low associated emissions is going to evolve over time? Is it really going to depend crucially always on government regulation to compel customers to get it? Or is it going to be a thing where increasingly you will see maybe the tech companies, maybe other large consumers saying, we have our own emissions goals, we need to hit those goals, therefore we're going to have to look for low emissions gas as part of the fuel mix to achieve that?

25:53What do you think is really going to be important in the future? I mean, if you look at the hyperscalers as a separate segment of the demand picture, what their cost structure is, they have said that the energy is a small portion of their total cost. So they're comfortable with paying a premium for low carbon energy. But they're only a small segment of the demand. You have industrial energy users, let's just say in Europe, who don't want to be disadvantaged by a tax on the LNG based on its carbon or methane emissions levels. And so it's a hard solution to solve the problem if you don't want to add cost to the delivered product.

26:51So one of the ideas I had, if you could combine what is proposed under EUMR with what is proposed under CBAM and look at natural gas as an LNG in the form of LNG and its carbon avoidance and create a currency for that avoided carbon tied to the price of carbon in Europe. and then use that offset against the CBAM tax on cement, steel, fertilizer, and electricity, then you could have a carrot approach as opposed to the stick and create a market where you would encourage continuous improvement and bend the curve on the greenhouse gas problem. Very interesting. So that is then, for example, you mean if you were using LNG imported into Europe to replace coal for power generation, that would count as an emissions benefit.

27:57You would then get some kind of certificate or credit for that, which you could then use against, as you say, CBAM, carbon border adjustment mechanism, which is essentially Europe's kind of carbon duty on imports. Exactly that. You're creating an offset that could be used against attacks for other imported products. And so it wouldn't penalize or burden European industry with a higher cost of their imported energy, but rather not increase their cost and encourage lower carbon imports and use it as a currency to bend the GHG curve. Very interesting idea. Is that something the EU is receptive to at all, do you think?

28:41Well, we just had a lunch earlier where there were buyers from all over Europe in this meeting and sponsored by RMI and Wood Mackenzie. And we talked about this extensively. So we ran out of time. But I think a group of buyers as well as environmental stewards like RMI could approach the EU and as well as the member countries and to propose something like this. And I think that there's opportunity out there for a pragmatic solution that will not burden the industry and the economies of the countries within Europe. Keith Shoemaker, thank you very much indeed. Finally, I spoke to my Wood Mackenzie colleague, Christy Kramer.