S&P Global Inc. (SPGI) Earnings Call Transcript & Summary

Hello, everyone, and welcome to today's webinar. My name is Mike Berko, and I'm part of the customer success team here at S&P Global Market Intelligence. I'm pleased to moderate today's webinar: Talk to the Specialists - An AI-fueled resurgence story - U.S. datacenters and nuclear energy outlook, a Q&A session from our main webinar that took place on March 6. If you missed it, you can find the replay link in the Related Content widget. [Operator Instructions] Under the Related Content widget, you can find links to our thought leadership resources. You'll also find our webinar replay portal to access this session and other webinars on demand. This webinar provides closed captioning in English. So please click on the CC icon in the media player to activate it. And please note that the activities of S&P Global Market Intelligence are independent and separate from S&P Global Ratings. S&P Global Ratings maintains a separation of analytical and commercial activities. Thank you for joining us. I want to introduce my colleagues, Dan Thompson, a principal research analyst who focuses on data center analysis from 451 Research, a group within S&P Global Market Intelligence. Joining Dan on the panel will be Tony Lenoir, a senior analyst focused on energy research at S&P Global Commodity Insights. We also have Katherine Matthews, a research analyst from the Metals & Mining team within Commodity Insights. Thank you, Dan, Tony and Katherine for joining us today. So very quickly, we have a question for the audience. Please select one of the following questions if you're interested in learning about more of the products featured in this webinar. We'll leave this on the screen for just a moment.

All right. So off to some Q&A. So Tony, we're going to start it off with you. We've got a question from the audience asking what are the main bottlenecks in nuclear energy development?

Yes. Sure. So thank you, Michael, and good morning, good afternoon, good evening, everyone. So here, a couple of bottlenecks definitely come to mind. And I'm talking about permitting and financing. On the permitting front, the Nuclear Regulatory Commission in the U.S., the NRC, is the primary authority. All nuclear power plant application are subject to an environmental review, a safety review and an antitrust review. And it's all under the purview of the NRC. The reviews include looking into the characteristics of the proposed site, geology, seismology, population surroundings, et cetera. And it also includes studying the design of the proposed project, including how the plant will be operated, emergency plans for it, anticipated responses for hypothetical accidents and so on and so forth. And as part of all this, there are public meetings that are held near the proposed site, and it's all to familiarize the public with the proposed project. So of course, all of this takes time, right? Historically, the time line has ranged from 3 to 5 years. And when it comes to permitting, we should also take into consideration the time that a proposed project would spend in interconnection queues. The last time we looked at interconnection queues in the spring of last year, the average duration from when a project is dropped in a queue to its proposed online date range from about 3 years in PJM to just about 6 years in SPP. So even with an optimistic view of these time lines, 3 years for the NRC reviews and 3 years in interconnection queues, we are at a minimum of 6 years of dealing with permitting processes, even though sometimes there's a little bit of overlap. So -- and the other bottleneck that I would point to is investing. Historically, it's been challenging to finance nuclear projects because of the capital intensity required, the technical complexities of building nuclear reactors and also the relatively long lead times, which overall include the permitting process we just discussed. Now that isn't to say that all of this can change. In our webinar presentation about a couple of weeks ago, we alluded to the renewed sense of urgency around developing additional nuclear capacity, right, for those who were on that presentation. We talked about electrification, decarbonization, and we emphasized the critical energy security angle of nuclear energy. And I think we can all agree that things are moving pretty fast these days. If I can just give a couple of examples. In the interval since our presentation 2 weeks ago, the U.S. Secretary of Energy, Chris Wright, gave a keynote address at CERAWeek last week and publicly reiterated his support and more broadly, the support of the current U.S. administration for more nuclear energy. Also just a few days ago, also at CERAWeek actually, a group of high-profile public companies signed a pledge to help triple global nuclear capacity by 2050. And this includes a few hyperscalers, by the way, since we were talking about data centers and nuclear energy. Amazon, Google and Meta signed the pledge. So to recap, nuclear has the backing of the Trump administration, and it is increasingly receiving the support of corporate giants with their deep pockets. So overall, there are reasons to be optimistic, I would say, about those bottlenecks, perhaps cautiously optimistic, but optimistic overall that all of those could be seriously addressed.

Amazing. Thank you, Tony. Dan, a question for you. What did the load profiles for these new non-crypto AI data centers look like? How much does demand fluctuate on an hourly, daily and seasonal basis? And is nuclear power able to meet those variations?

Fantastic question. Actually, I had a slide just so that we can change things up on screen here. Let me see if I can do this here. Hopefully, everyone can see this slide. So this is a slide that we used back in the presentation a week or so ago, which sort of outlines what the load profiles of various data centers look like. This is sort of -- this may be missing a little bit of context from that conversation. But in any case, this gives you a bit of a snapshot here. So generally speaking, when we think of data centers, other than cryptocurrency, the load profile is fairly flat and consistent. It will go up over time when a brand-new facility comes online. The reasoning for that, you would imagine when a brand-new building opens, it takes a while for all the gear to get put in, all the gear to get turned on and fully utilized. So there will be -- you can think of from a brand-new perspective, kind of this ramp-up period. The interesting thing about AI data centers, though, is that the demand can come and go pretty quickly, and so particularly the training data centers. So this only really pertains to the training data centers. These can see load swings, massive load swings actually because think of AI training in terms of like a job, these jobs can be batched. And so the hope is that they follow one right after another. And then the GPUs essentially never spin down or at least don't spin down much. But there can be sort of sags in processing, which means sags in energy consumption. And so this has been one of the key concerns of the utilities is, for example, like it could be a thing that 50 megawatts of demand appears on your grid and then subsequently disappears maybe hours later. So that is a bit of a challenge. I would say, though, just as a point of encouragement, the AI workloads and the GPUs, the hyperscalers, Microsoft, Amazon, Google and so on, they are making money when these GPUs are being utilized. So it behooves them to keep them utilized all the time. And so their desire would be that these things are running full tilt basically all the time because that means they're making money on them all the time. So just kind of keep that in mind. They are financially incentivized to keep these things running. And so these dips, while they could be dramatic like in theory, it is perhaps the real-world application of this may look a bit different. It may be fairly flat or at least a level of consistency that's not as bursty as what I'm describing. But I do just want that out there that it is possible that these hyperscale, particularly AI training facilities, will be bursty. Now one more thing to consider is up until this point, there are very few dedicated AI training data centers. There are data halls within much bigger data centers which house AI. But then they also -- these data centers also house other things, which will help essentially flatten that curve. I do -- there are full data centers which are just AI training. But I just want to kind of put it out there that like those are more of the exception necessarily than the rule. So most of these data centers, they are low profile, is going to look fairly flat.

All right. Thank you so much, Dan. Kat, a question for you. Could thorium replace uranium as the main nuclear fuel?

Thanks, Michael, and hello to everyone. From the outset, I must admit that I had to do a bit of research around this topic as I'm just a geologist. We don't cover thorium as a commodity in our team, and I'm definitely not a nuclear scientist. So a lot at the fueling side of the question I'll be -- that I'll be relaying here is primarily sourced from the WNA and the International Atomic Energy Agency. So on a geological and mining front, thorium is more abundant in the Earth's crust than uranium. In comparison, it is inert. So it's common in deposits such as pluses, whilst uranium can be found in soluble complexes, which hence its amenability to in situ leach mining. It's most commonly found in a mineral called monazite, which is a type of rare earth phosphate. So true global resources aren't known, but it's estimated by the WNA. It's around 16 million tons with the greatest concentration currently around the eastern and southern coast of India. Other known resources exist in Brazil and Australia, and monazite is currently extracted in Brazil and India. However, thorium is allegedly not economically recoverable at present. So the research indicates that thorium recovery is most viable via byproduct methods, being the primary rare earth element mining of the monazite, for example. Now on a nuclear fuel front, which I'm certainly less familiar with, thorium isn't directly usable in a reactor as it's not fissile, but it's considered fertile. Now what this means is it needs to be given a neutron in order to trigger decay and conversion to uranium-233, which is a fuel that's ideal for reactors. It can be done, but it also essentially means a fissile material such as plutonium or uranium is needed to provide this extra neutron and be incorporated into the system to maintain the chain reaction. There are a couple of extra steps in this, not very scientific overview at all. But there are a number of reactors that can handle thorium, including the heavy water reactors such as CANDU, pressurized water reactors and the boiling water reactors. There are also molten salt reactors, which are still at project stage, but these have been touted as potential reactor types that could leverage thorium fuel. Now thorium as a nuclear fuel is considered cleaner and more efficient than conventional ones. However, a drawback is the current state of R&D. So uranium has been at the forefront of nuclear science. There's been comparatively research conducted into thorium applications, and the nuances and technicalities around processing of spent fuel is costly at present and challenging and the additional steps required to create the fuel in comparison to uranium are also expensive. And this is apparently due to some of the daughter isotopes being heavy -- like heavily gamma emitting. So whilst there is potential, we have the mining issue that until it takes off as a primary nuclear fuel source, thorium is it going to be economically viable to extract, and at least for the foreseeable will remain as a byproduct of your rare earth mining, for example? And even currently, that source isn't largely exploited. We also have the prevalence of uranium fuel. And whilst that is easier to produce, more abundant in terms of global upgrading and conversion capacity, it's better understood, has had far more testing and applications in day-to-day energy supply, we see it will be a while before thorium might find its corner in the energy market.

Amazing. Thank you, Katherine. Tony, we've got a new one for you. Do you believe IRA tax credits will survive for nukes or any other renewables or not at all?

Thanks, Mike. Yes. So as we just said, the -- everything has been evolving very, very quickly in the last couple of months. So honestly, I don't know if anyone's -- I don't know about the reliability of my own crystal ball on the ultimate fate of the IRA. But -- so what do we know about the current situation? We know that President Trump suspended all IRA funding disbursement soon after he was inaugurated in January. We talked about that a couple of weeks ago. But we also know that it would take an Act of Congress to fully repeal the law, right? But I think that for nuclear, looking at the latest signals from DC, it gives us a pretty good sense of where this might be headed. First, as we said, the new administration has voiced its support for nuclear. They've been pretty loud about it actually. We were talking about Chris Wright's remarks at CERAWeek last week. But also soon after he was appointed, actually, he issued his first order for the DOE. And the document includes a paragraph named -- is it named unleash commercial nuclear power in the U.S.? So -- and it specifically says that the DOE will work diligently to deploy nuclear technology rapidly. So I think there's no ambiguity there, right? And given the importance of nuclear in the department's priority road map, I think it's difficult to imagine that the IRA tax credits for nuclear could disappear. And in the event that only parts of the IRA survived, I think the probability that whatever is salvaged would include the tax credits for nuclear, it's probably relatively high. And we have to keep in mind that many of the locations that make nuclear development attractive based on the IRA, and here, I'm talking about energy communities, for instance, are in areas that generally showed strong support for President Trump in the last election. This includes the U.S. Southeast, large parts of Texas, Appalachia, but also many parts of the Rust Belt. And along those lines, a little over a week ago, a group of Republican lawmakers, I think it was around 20, wrote a letter to the Chairman of the Ways and Means Committee to express their concern about repealing the IRA credits. They mentioned that a lot of development and capital allocation plans were devised around those credits, those incentives. So of course, there's the possibility that all of this could be phased out if the incentives from the IRA are repealed. And again, along those lines, some studies have shown that a significant amount of the clean energy investments that have been announced since the IRA, so it's 2.5 years ago, are to be located in Republican districts. So again, difficult to make predictions as to what will happen for sure with the IRA. But just like for the bottlenecks, it seems to me that if we take into considerations what I just said, certainly including the Trump administration's push to unleash nuclear energy, I think there are probably good reasons to be optimistic.

Definitely. Thank you, Tony. Dan, so these data centers typically operate at 100% capacity and have priority power purchase agreements, right? There have been some research describing how the current U.S. power grid has enough energy to support the current need but it would require smart meters and demand response. Any comments on that?

Yes. It was a fun science project from, I believe, it was Duke University that was sort of describing this. The problem with that theory and this idea of essentially using stranded power in different parts of the country and in cases, curtailed power, it sort of necessitates these data centers to turn themselves off during certain periods of time or run on generators, which I think we would all agree is even less desirable. But the challenge here is that the hyperscale data centers, even enterprise data centers, they need to be up all the time. In fact, that's the key feature of a data center is uptime. And so they actually build in systems of redundancy to keep them online. So this notion of coming off the grid while, again, they do have systems of redundancy to protect against that, every time the grid goes down, it introduces an opportunity for failure. And so their preference would be that the utility never go down, even though they have all these backup systems because it just introduces more risk. So I think while a brilliant idea, and I'm sure the cryptocurrency folks would be very interested in this idea because they can turn their workloads on and off, at this point, infrastructure can't be just turned on and off. I mean it could be in theory, but it is in practice not turned on and off. Training could be done really at any time. But again, like these companies have spent like literally, at this point, hundreds of billions of dollars on this gear. And as I was describing earlier, it behooves them, they're actually financially incentivized to for them to be running all the time. And so this notion of them turning off or somehow passing workloads from here or there or another, again, in practice, it's just different from what it looks like in theory. This is all in theory possible, but no one is doing this, except for the crypto folks. So again, I don't mean to sound negative on the research. I mean it is interesting research, but in practice, the world just doesn't work that way, unfortunately.

Yes. That makes sense, Dan. All right. Katherine, we've got another question for you. Will domestic uranium production ramp up quickly enough to supply nuclear plant development? Or will the U.S. need to rely on global supply chain for nuclear fuel?

Good question. So there is a reasonable volume of known resources and reserves of uranium in the U.S., placing it well within the top 20 globally. However, the problem isn't necessarily with the primary production sector. So there's lots of material out there as the material moves from upstream down the chain through to conversion and enrichment. So there are domestic upgrading facilities. However, the capacities at present are insufficient to meet requirements. We have Honeywell's Metropolis Works facility in Illinois, which is a conversion facility. So this takes U3O8 or yellowcake and converts it to uranium hexafluoride. And it's the only UF6 conversion facility within the U.S. From my understanding, it has a nameplate capacity of 17,600 million tons uranium as UF6. However, it was placed into a state of care and maintenance during 2017 due to poor market conditions. It did, however, opened its doors again back in July of 2023. And enrichment is a similar story. So domestically, there's only the Urenco USA or UUSA facility in New Mexico that's operated on a commercial basis. And that's a fairly young operation by all standards. It had commercial production or commercial operation approved in 2010 and has a current enrichment capacity of 4,300 tSW per annum, which is separative work, I also have to do a bit of research on that, and basically indicates the amount of separation done as part of the enrichment process. In comparison, Canada -- or take Canada for an example, it has a well-established chain, all based out of Ontario. So we've got the Blind River Refinery, takes primary yellowcake from mine sites and refines it into uranium trioxide. Conversion is carried out at Port Hope, and this upgrades the trioxide into uranium dioxide and UF6. The UF6 is exported to companies elsewhere around the world, mainly for use in light water reactors with further enrichment and fabrication, whilst the dioxide is then manufactured into fuel that's suitable for use in the CANDU reactors that are found around Canada. So both facilities are owned and operated by Cameco, and they've certainly caught up the market over that side of the world, especially with the sanctions surrounding Russian-sourced enriched uranium. So if you go back to the webinar the other week and check out the slides, you'll see Tony was reviewing the U.S. imports, and Canada imported well over 1/4 of the uranium needed to power the U.S. civilian nuclear fleet in 2023, with the top 4 countries being Canada, Kazakhstan, Australia and Russia totaling over 82% of the requirement. So from this perspective, the U.S. will certainly be relying on external sources and imports into the country for its nuclear fuel requirements, at least for the foreseeable.

That's great, Katherine. All right. Tony, any thoughts or any new news you've heard on public perception and acceptance of nuclear energy?

Interesting question. Well, I suppose past accidents, Chernobyl, Fukushima are still very much present in people's consciousness, right? I certainly remember the newscast around Chernobyl very well. I was of age at that time. But I would venture to say that public sentiment is shifting. Certainly, it's not what it was 10 years ago. It's not even what it was 5 years ago, even just 2 years ago. It's evolving very quickly, again, something that is evolving very quickly. Again, for those who attended our webinar 2 weeks ago, you might remember an infographic we had at the end on what we believe -- an infographic representing what we believe are the primary pillars supporting the energy transition, with those 3 pillars being decarbonization, electrification and energy security. We talked about the energy security pillar being reinforced by the shift in international relations, the situation in Ukraine, et cetera. And I think it's probably safe to say that in such an environment, reviving nuclear is an easier sell for the public, especially when the need to bolster energy security coincides with a period of higher energy prices and everyone looking at their high energy bills every month and asking for solutions. And I think other ongoing developments could help when it comes to the perception and acceptance of nuclear energy. If you think about it, electrification, for instance, in a world where more and more people adopt electric vehicles and install heat pumps to keep their homes warm, I would think that the public would start paying more attention and become more invested in the reliability of our electric infrastructure. I suppose it's easy to dismiss a power outage when all it does is impact your ability to turn on a light at your house or power an appliance. It's annoying, yes, but in the grand scheme of things, we can manage. It becomes much more problematic when the power outage may prevent you from going to work or picking up your kids at school or lead you to freeze at your house in the middle of winter. And so if we bring back the conversation to the topic we're discussing today, which is really about powering data centers, well, this in essence, is the predicament that data centers are facing, right? If AI is a matter of national security, as we've made the case, well, you cannot not have reliable power 24/7. Anyway, just also say that, it appears, again, here that several developments are converging to get the public onboard for a nuclear revival.

I think just to kind of tag on with what Tony said in there, having just gotten back from CERAWeek myself, there was a fair amount of publicity at the event on how much safer new nuclear is than old. I say this to say it would not surprise me if we don't see further publicity just out in the general public, even out of the energy sector, just trying to make the case that Tony just made and also make the case that modern nuclear is different than that of the '70s. Anyway, it seems like there's a story being told here, and it would not surprise me at all if we continue to hear that even louder. But anyway, I know we're up on time, Berko, so I will be quiet now.

Thank you, Dan, and thank you, everyone. I know we covered a lot today. So for the audience, if you have any follow-up questions, please use that Q&A widget, and we will get back to you shortly. If you're watching the replay, please reach out to us via the Request Demo link under the Related Content widget. And if you want to revisit any of the material we discussed today, we've recorded this session. Tomorrow, you will receive an email with a link to access the replay at your convenience. Before we go, I'd like to remind you that we recently acquired Visible Alpha, which provides sell-side models and granular consensus estimates. Click the link in our Related Content widget to learn more on that. And while we are here to close out the webinar, you will be routed to a survey. We'd love to hear your feedback, so please take a few moments to complete it. So that's all. And the only thing left to say is thank you, Dan, Tony and Katherine for presenting, and thank you all for taking the time to attend the session today. We look forward to you joining us again soon. Thanks, everybody.

Thank you.