As the pressure for climate action fades, what is driving investment in clean energy? | The big talking points from New York Climate Week

0:04Hello and welcome to The Energy Gang, a discussion show from Wood Mackenzie about the fast changing world of energy. I'm Ed Crooks. I'm an energy analyst and the host of the show. We're back in New York City for Climate Week, one of the biggest events in the energy year. It's bigger than ever this year. And we're talking to industry leaders, investors and policymakers about the biggest issues in clean energy. Climate Week has in recent years taken place against a backdrop of confidence in the transition to low carbon energy. When we were here last year, market sentiment around renewables was still riding high.

0:36Policy support for solar and wind and other low carbon technologies was strong. So where are we now? This is the first in a special series of episodes of The Energy Gang that we're recording at Climate Week. We're going to be talking about all the biggest issues in energy right now, including meeting electricity demand for AI, breaking down the barriers to low-carbon technologies, including nuclear power, and following the money to see where the investment is coming from and going to. Today we're opening our Climate Week coverage by talking to Helen Clarkson, CEO of The Climate Group, which organises New York Climate Week.

1:08Here's what's coming up from her later in the show. Climate Week NYC has always been really important because I think we've called it sometimes like the green room for the COP. It's a moment that sort of focuses the mind. It's always a bit back to school, you know, time to focus everyone. You've got two months or six weeks maybe to the COP. We need those NDCs. We need to kind of pin down who's going to be where, when, who's saying what. Really high electricity costs on the grid. In Pakistan, very little regulation about you can just put a solar panel on your house. people started putting up TikTok videos showing this is how you do it, very cheap solar panels from China, and now you can just see the solar boom in Pakistan on Google Earth.

1:44Even in rural villages, about 50 % have solar on the roof. That sort of shift is coming really, really quickly. And so I think as much as it's about the US sort of growing back, the rate and difference of change across the world, that's the difficulty, I think, for businesses looking out across that landscape and thinking how to respond to that. So that's coming up later. Welcome to today's show brought to you by Engie. Your business has enough challenges. Energy shouldn't be one of them. That's why Engie builds tailored energy solutions around real business needs to support growth, strengthen predictability and move businesses forward.

2:18Because real power comes from shared expertise and relationships that outlast the paperwork. Learn more at engiresources.com. But first, I wanted to get an investment analyst's view on where we are with the energy transition. Will Thompson is a director in the thematic investment research team at Barclays Investment Bank. Hi, Will. Welcome to the Energy Gang. Thank you for having me, Ed. Yeah, great of you to join us. So look, I wanted to think about Climate Week in the context of investors and the way investors are viewing the energy industry right now. It feels like probably if we'd been having this conversation two years ago, all the talk would have been about climate, would have been central focus climate and sustainability probably more broadly.

2:59Now it feels like the focus has shifted very dramatically. Climate is being de-emphasized in investors' decisions. And the thing that's absolutely dominant, certainly dominant in the conversation, is AI, electricity demand, all the changes that are flowing from that. Do you think that's right? Yeah, I mean, we at Barclays hosted our annual energy conference earlier this month. And two central themes was obviously the integration of AI, both in the upstream side, but also in terms of advancing clean tech technologies. But secondarily, was addressing this growing load growth that is associated with data centers and our AI ambitions.

3:35And how big is that opportunity, do you think? You have a lot of numbers being thrown around. People talk about very steep growth in electricity demand just out to 2030. Maybe there's another 100 gigawatts or so of demand that could be there on the system and therefore very significant need for investment in new generation capacity and in transmission and distribution infrastructure as well. Do you think that's right? No, we're seeing it. We've been actually, even despite the January events of DeepSeek, saying that this train has really left the station. There's been a paradigm shift in the AI scaling laws.

4:11It was really leading to both training compute needs, also inference compute needs, and I can get into that later. But the Department of Energy, I think, pegs data center load growth of 4.4 % of U.S. powder demand, I think that was in 2024, and potentially growing to 9 % to 12%. And I do think that's consistent with how we see the market playing out. Right. And just a footnote on that DeepSeek point. So this was DeepSeek. This was the Chinese AI model that was announced earlier this year, appeared to have radically lower power consumption to achieve the same results as some of the other models from Google and OpenAI and others.

4:47And as you say, there was a sort of, it was like a DeepSeek moment when there was a big correction fallout back in share prices for some of the power generation companies. Everyone had this sort of moment of worry about would that increased power demand really be there. That didn't seem to last very long. It seemed like pretty quickly after that, people got back to reassuring themselves that, yes, that demand growth was really going to happen. Yeah, I think there was a market shock and the understanding of this Chinese startup come out with this pretty efficient model, both in terms of model-level architecture design, but also on the training costs.

5:22What it really showcased is a paradigm shift in the AI scale and loss. So historically, the belief was that we needed more compute, more costs, more data centers to advance AI intelligence. What DeepSeq really showed that is, yes, we continue to use more data, more compute on the training side, but the models are getting incrementally more intelligent on the inference side. So on the back end of once you train these models, we're teaching these models to reason and think like humans through complex problems. well that requires time test scaling in terms of actually spending more time in compute letting the models think and that's actually leading to a large surge in inference demand associated with AI.

6:01So what are the implications of that then for electricity demand? I mean it feels like when people talk about this issue there's sometimes the implication that effectively demand is infinite that as many new data centers as you can build will be needed and as much new electricity infrastructure as you can put in place to support those data centers is going to be needed. And so that's really the thing you have to do is just deploy as much capacity as quickly as you possibly can. There is also obviously then underlying that the sort of the worry that maybe it's a bubble. Perhaps at some point this is all going to become much more efficient and either the computing is going to be done more efficiently or the data centers are going to become more efficient and use their electricity in smarter ways that mean they don't need as much.

6:50And you're going to end up with stranded assets, and there will be data centers and power plants and transmission lines and everything else that aren't actually being used. How big is that risk, do you think? Yeah, I think there's a lot of investor handling still on the return on AI, so the ROI of AI, for obviously concerns that we don't, you know, the hyperscalers don't know what the return is. I think they're optimistic. They are seeing tangible, measurable returns in terms of their own operations. But in terms of gen AI, we don't know what the sort of perfect model is for that and how to monetize that.

7:24From their body language, my sense is they actually consider AI an exceptional risk to their business models, right? Meta's talked about how AIs can impact everything they do. And so I think there's an arms race, clearly, not only within the tech community, but also autocratic versus democratic AI, because obviously the threat of China. And so I think they're throwing everything they have behind it. These are well-capitalized companies. And based on their body language, they are very power hungry. And I think their actually biggest concern is not the return of these investments, but actually access to power to actually meet their AI ambitions.

7:55And we're of the view that maybe the risk to hyperscale CapEx is not the return profile of AI, because we're still early on the return curve, but actually accessing power and we're maybe actually hitting a power wall. Right. So when you talk about the power wall, this is effectively what, as I was saying, if you think about the demand for AI and AI-related services as being basically unlimited, or at least we don't know what the limit is, you're saying there's going to be a constraint on the supply side with people being able to get the power they need. Yeah, I think we are hitting the power wall.

8:29There's indications that data center capacity under construction has plateaued, if not started to decline, US and globally. And that suggests to me that there are power constraints really because the pipeline of data center projects continue to grow. These are supported by extremely high pre-leasing rates. But actually moving those projects from sort of feed to construction has been a challenge. And we're increasingly seeing a lot of even the hyperscaler projects move to distributed power. So what we like to call bring your own power, BYOP. And we're seeing it with XAI in terms of their Colossal 1 and Colossal 2 data centers in Memphis, Tennessee.

9:05Meta is building a one gigawatt data center in Ohio that will be largely powered by distributed power solutions. So I think this move to sort of grid independent power solutions for data centers suggests that they are realizing that they can't get access to the power they need. I want to come on to that question of distributed power in a moment. Before I do, just to think about the constraints then that are forcing the data center operators and customers in that direction. What are the issues then? What is it that's stopping people getting the power they need? A lot of it is we have concentrated data center build out in these primary markets, and they've absorbed any excess power that was already available.

9:40We essentially haven't grown power generation in the United States for several decades, right? The demand growth has been flat for two decades because of the efficiencies we've been able to get in terms of LED lights, high-efficient appliances, industrial efficiency efforts, even though we've had GDP growth. I think my sense is that low-hanging fruit has been picked, and now we're talking about gigascale data centers just showing up and asking to interconnect to the grid. And we've seen the growth in the interconnection queue for a lot of these projects. And we've grossly, from my perspective, underinvested in grid infrastructure to move the electrons from generation.

10:10So we can focus as much as new power generation. But you can't move solar or wind generation to where the data center is. Then you have a grid interconnection issue. And so if you need new generation and grid infrastructure, some of the issues obviously that come up, So permitting and siting being problems, that still seems to be a perennial problem in this country. There's forever talk about doing something in Congress to streamline permitting. That talk's been revived again recently. There's talk about maybe bipartisan legislation passing. I believe that when we see it, I guess, have to wait for that to make a bit more progress and to become more real.

10:48So that's one issue. Another issue clearly is the question of how that infrastructure gets paid for and what the impact is on customer bills. And certainly you see it is generally the case, I think, that a lot of people in the United States right now feel that their personal finances are strained and rapid rises in electricity bills. I think electricity bills rising on average by 7 % or so in the US this year. That makes a lot of people unhappy. People notice that. You get political pushback, certainly in quite a few areas where people are looking to build data centers. You're finding political resistance.

11:24I was very struck on the social platform X the other day. Ron DeSantis, governor of Florida, Republican, you would think of as generally a kind of business-friendly type of person. He was saying, you know, we're worried about these data centers. They suck power off the grid and they don't contribute anything to our economy. So I'd say lots of different signs of resistance coming. Do you think that's something that's really going to obstruct data center build-out? Yeah, there's already been NIMBY resistance, particularly somewhere like Northern Virginia, where, again, they've impacted reliability.

11:58They're leading to higher energy prices, particularly in the PJM market. We saw the impact of the PJM auction and the potential pressure that we'll have on utility bills for consumers. And I think I do worry that this will get highly politicized. And so this could be, as we try to figure out, I mean, there's clearly bipartisan support to lead on AI against China. But I think where we lack consensus is how do we do that? And then the other thing that we follow quite closely at Wood Mackenzie is equipment and equipment supply chains and whether companies building infrastructure are able to get the components and equipment they need.

12:36It seems like that's becoming a big problem. If you look at wind and solar, actually availability there is not so bad. But we've had the administration putting various regulatory barriers in the path of wind development in particular. also its extent for solar. We've had, of course, tax credits for wind and solar on their way out as a result of the legislation, often known as the one big, beautiful bill that was passed in July. So the administration very much trying to encourage investment in new natural gas generation, because they say you need reliable baseload power that's available 24-7, that's dispatchable and can support the grid, therefore, in that way.

13:19And that supply chain is pretty tight. And I think our numbers are showing that, on average, if you order a new gas turbine, you're going to wait four years for it today and can be significantly longer than that. And so that, again, is another reason for significant delays in getting new capacity built. How do you think that gets resolved? You were talking about distributed generation as being part of the answer. So that maybe gets around some of these issues with NIMBYism to an extent, and certainly could be a way to get around concerns about new demand on the grid, creating pressure to drive prices higher.

14:02But still, you need the equipment, right? You have to buy it from somewhere. Yeah, I think we obviously are seeing the demand signal today that we need this new generation and the transmission to move the electrons. I think there's obviously the continued overhang of the uncertainty around the sustainability of hyperscale CapEx. We have seen capacity additions added from the major OEMs who are producing these turbines. But it is across the entire supply chain because we've been talking about large power transformers that are critical, both on sort of the generation and the receiving end of these transmission wires.

14:34And we hear three to four year lead times. We're highly dependent on European imports for a lot of this equipment. and it still requires skilled labor to produce. And a lot of it requires specific steels that we don't actually have a lot of suppliers. We have one supplier of what's called greener electric steel in the United States. So I think there's a whole lot of headwinds to standing up the supply chain to grow our generation, which we haven't done again in several decades. So I think there's a broad recognition that Solon Wind can get us so far, but as a cheer point, we need baseload power to support these data centers, which are considered essentially a block load.

15:09but they run 24-7, 365. And so, yes, we've seen strong demand for aerodinative gas turbines, industrial-scale gas turbines, even fuel cells and gas engines. So there's a whole supply chain that we're seeing a surge in demand growth for to support the move to distributed power. So how do you think this plays out then both for the industry and in policy terms? For the industry, as we've been saying, these tech companies see themselves as being in a race to develop the most effective AI capabilities as quickly as they possibly can. In policy terms as well, there is a sense for the administration and on a bipartisan basis in Congress that there is an AI race between the US and China now, just as there was a space race with the Soviet Union in the 60s or a race to develop the atomic bomb against Nazi Germany in the 1940s.

16:03These are kind of very high stakes, apparently, that we're playing with here. So something has to be done, it would seem, to get through that power wall that you're talking about to make it possible to accelerate development of AI and the power infrastructure needed to support it. What do you think can happen? I think whoever wins the energy race can win the IRIS. And that's obviously a global phenomenon that we're facing. The move to distributed power, I think, has been just a focus on speed to power. And so the hyperscalers are less sensitive around the cost of electricity and more focused on when can I get it.

16:38And so I think there's a worry. There's obviously a competitive nature in this industry to get B first to have the best frontier model. Meta even alluded to the fact that, you know, we have AI making AI better. And so being on the front edge of the AI race will give you a huge competitive advantage. And so getting that competitive advantage will require getting access to power. Yeah, very interesting. You use that phrase speed to power. I was very struck by a statement that came out from the Energy Department very recently talking about a new speed to power initiative. They have clearly identified this as a big priority for them.

17:11Do you think we're going to see material consequences from that in terms of measures that will actually accelerate the deployment of energy infrastructure? We're seeing, I mean, you mentioned earlier about the challenges on infrastructure permit reform, right? We have state, federal, local permitting, environmental reviews, then you have cost allocation and you have to rate-based this stuff. Like, it's very complicated. We have 3 ,000 utilities in the United States. There's no master plan. Obviously, the best option for speed to power right now is solar plus batteries, but we need to stand up a lot of battery capacity to do that.

17:42So I think we're seeing policy efforts in terms of Texas State Senate Bill 6, which focuses on having backup power on site. We're seeing an effort in West Virginia for similarly to have backup power on site. So I think the recognition that we will add power capacity in state, but when the grid gets constrained, you need to have your own power available so that we don't constrain the grid at peak times, particularly somewhere in August where we have everyone running their air conditioners. So I think that's the back end way to get around this. The fact that we can't add utility scale generation and transmission fast enough is that we're essentially having the data centers become utilities with their own grid independent solutions.

18:22So a final thought then, taking it back to Climate Week again, what are the implications of that for emissions going to be? Given that, as you say, it seems like there's increasingly a focus on distributed power. Some of the time that distributed power is literally just diesel generators. Some of the time it might be gas turbines, gas engines of various kinds. It feels like on the whole, doesn't it, that this push for speed to power and significant backup power for data centers to make it possible to get them online, that's just inevitably going to drive up emissions, isn't it? Yeah, I mean, the default form of cheap baseload power is natural gas.

19:00And so most of these projects are being supported by natural gas. So clearly there's an emission implication there. we still see the hyperscalers essentially meet all their annual energy demand needs in terms of virtual power purchase agreements for renewable and solar projects but again there's there's a mismatch in terms of the timing so i think the stick which the industry should be measured by is 24 7 supply demand balancing on the same grid network we've gotten some disclosure from the hyperscalers on that and we know that they're far behind the 100 target by 2030 and so i think Again, their emphasis is speed to power today.

19:34We're going to use natural gas. We'll try to figure out how to decarbonize it down the road. In the meantime, we'll still sign a lot of virtual power purchase agreements to essentially say we're using 100 % renewables in terms of annual matching. But then the long-term goal is to use AI to sort of advance the clean tech and decarbonization efforts. Right. And so in the long term, that means things like advanced nuclear. It might mean long-duration storage. It might mean geothermal. might even potentially mean hydrogen, I guess. But in the short term, it's, as you say, solar, batteries, and gas, right?

20:07Yeah, and fusion, I think, is the, hopefully, maybe down the line, obviously, but maybe a binary outcome. But I do think it is, there's this message from, I think, some tech executives, do not constrain AI power demand today because of the benefits, both in terms of not only decarbonization, but healthcare and advancing society in the long term. So I think there is a balance between, do we constrain an industry that's obviously leading to higher emissions than they had signaled to us? Or do we take the bet that AI is going to essentially make the world a better place? And so I think we have to sort of navigate through that.

20:40And right now, the focus really is on how do we get speed to power? Indeed, let's hope we bet correctly on that one. Certainly, as you say, some pretty important decisions to be made on that score. Every business has priorities to protect, goals to reach, and decisions that need to hold up. Energy should support all of that, not become one more thing to manage. That's why Engie takes the time to understand your business before building energy solutions around it. Your operations, your goals, your pressures, your plans for what's next. Because while Engie knows energy, no one knows your business like you.

21:15And when that expertise comes together, energy becomes more than something that powers your business. It becomes part of what helps move it forward. Engie helps turn energy plans into outcomes with solutions built around real business needs. Learn more at ngresources.com. That's E-N-G-I-E, ngresources.com. Will Thompson, thanks for having me today. Great talking to you. Ed, I appreciate it. Thank you. Thanks for joining us. So that's the backdrop for Climate Week. It's a challenging time for clean energy in some respects, but there is still plenty of optimism in the industry. Helen Clarkson is CEO of Climate Group, and she joins me now to talk about the key messages and debates that we can expect this week.

21:52Hello, Helen. Hi, good to see you. Welcome to the Energy Gang. Thank you. Thanks very much for joining us. Let's start off maybe just talking a little bit about the event. What is going on at Climate Week? And what do you think is particularly important about this event? Well, Climate Week NYC happens every year alongside the UN General Assembly here in New York. And actually this year, the fact that it's happening, I think, is really important. I'm not really one for kind of giving people marks for showing up to things. But, you know, across this year, we've been looking at what would the event look like?

22:21So much uncertainty out there. You could see the way the kind of year was unfolding. And so actually getting out there and saying to people, look, we still need to have these conversations. Climate needs to be top of the agenda. And New York is the right place to have those conversations because we've got people in town for hunger and so on. And kind of convincing people that it's going to happen. I think it's been really important to bring everyone here together. So now they're here and they are having those important conversations and moving the agenda forward. And I think I heard you say yesterday at the opening ceremony, this will be the biggest climate week in NYC ever in terms of interest, activity, attendees.

22:54on what metric is that? Yeah we have a few metrics so one thing we do as climate week as climate group is we run the week which means having our own events but also hosting the platform where people can register what we call affiliate events and those are events that happen around the city and we always joke they sort of range in size from big sort of global meetings to read poetry readings about climate in in basements but potentially comedy events which I always every year so I'm going to get to never do this year we've registered a thousand events on our website which is more than last year when you know the world looked very different this time last year also the number of government leaders that were speaking at the climate group events over these couple of days so showing that real appetite to come here and have these conversations and kind of encourage people to do more and then in terms of corporate sponsors interest from companies so a range of metrics that we look at for like gauging the interest right got it because anecdotally certainly when I've been talking to people in the energy industry over the past few weeks, quite a lot of people have said they're coming to Climate Week.

23:55And quite a lot of people have said one of the reasons they're coming to Climate Week is because they're not going to COP30, the climate talks in Boleum in Brazil in November. And there's a range of reasons for that. Logistically, it's tough. Boleum's not a very big city. It's clearly going to be strained in terms of its hotel capacity and that kind of thing to support the tens of thousands of attendees that you typically get at a COP. So I think that's one reason why a lot of people are taking the decision not to go. Another reason, though, you could say is that because optimism and confidence in what a COP might achieve has been waning.