The AI data centre backlash: can the boom survive contact with local politics?

Paul Diggle

Hello and welcome to Macro Bytes the economics and politics podcast from Aberdeen. My name is Paul Diggle.

Lizzy Galbraith

And I'm Lizzy Galbraith.

Paul Diggle

We are going to be talking about AI, but from a slightly different angle to the way we normally treat it. Of course, we've spoken a lot about the transformative technological change from AI, potential risks that brings. It's one of the defining features of the economy, of financial markets, of politics, of geopolitics at this juncture. But AI has something of a reputation problem with many ordinary Americans and indeed populations globally because there is something of a backlash in parts of the US about where data centres are being built. And it's part of a broader political fight about who bears the costs and who feels the benefits of AI in which data centre construction itself is perhaps a proxy for broader public anxiety about AI's impacts. But that backlash could have genuine impacts on what we care about as economists and investors - the returns that could be earned from data centre investment, the speed at which AI technology itself develops, whether the potential positive economic impacts it could bring in terms of investment productivity are ultimately realised and it's really important to the sustainability or otherwise of stock market performance driven by AI. So it's I think a really important topic to consider when we're thinking about this thematic, this huge driver of macro markets from all its different angles. So that's what we're going to dig into today on Macro Bytes. Lizzy and I are delighted to be joined by Nancy Hardy, sustainable investment manager here at Aberdeen, and Tim O'Reilly, a partner at Tritax, which is the UK data centre developer, and it's part of the broader Aberdeen group. So welcome, Nancy. Welcome, Tim.

Nancy Hardie

Thank you for having me.

Tim O'Reilly

Hi. Thanks for having us.

Paul Diggle

So Lizzy, let's start by perhaps level setting on this idea of a sort of backlash or public opposition to AI. Tell us where public opinion on AI and data centres sits at the moment.

Lizzy Galbraith

So it's fairly negative. And what's interesting is it's becoming more negative over time. So this seems to be something where, particularly in the US, the more the public hear, the more it's reported, the more negative their responses have become. So if you look at polls last year, they weren't necessarily good, but we have seen a marked deterioration in the number of Americans that are, as an example, willing to allow a data centre to be built near their home. You also are seeing more negative reputational impacts as well. So people are more suspicious of the potential knock-on effects of having a data centre near them as well. So, the polling hasn't necessarily started in a particularly positive place, but it has been deteriorating over the past six months or so. And you now are seeing actually quite a small minority of people who consider these data centres to be a positive, kind of, on net, particularly if they're asked about a data centre potentially near them. So it's not in a great place. And there's a lot of reasons for why people have those types of opinions. The different factors people are citing are fairly broad based. And we're going to get into some of those later in the pod. But the other thing to note with this backlash is that it is interestingly quite bipartisan in a time where we know an awful lot about political polarisation, people don't agree on very much necessarily anymore. Actually, having a sort of a negative view of data centres, particularly the potential of new data centres in your local community does appear to be quite interestingly a bipartisan issue. And you're starting to see that in some of the political responses as a result.

Paul Diggle

We're going to get into, as you say, Lizzy, some of those specific issues people, populations have, what data centre construction itself can do to mitigate those. But I think a really important theme of this backlash is that arguably it's not just or really about the buildings themselves. It's perhaps tied up with broader public anxiety about AI as we were talking about - jobs impact, which we've talked about on the podcast, the way it could change society in all sorts of ways. But on the physical build out itself, then how much is actually being held up or prevented by this opposition?

Lizzy Galbraith

So this is where it gets a little bit tricky because estimates vary. There's not necessarily a single source of information that we can be confident captures absolutely everything going on because of course in the US it’s all done at a state level. But current estimates would suggest that something in the region of $64 billion worth of data centres are either delayed or cancelled as a result of local opposition. And some other alternative estimates would suggest that of data centres that are currently planned around about half are subject to delay or cancellation. So they're quite significant in terms of the potential impact here. And certainly something that does appear to be rising as the data centre build out expands across new states or concentrates in certain areas of the US. So with that kind of introduction to where public opinion is at, now we want to actually get into the reasons why the public are having this kind of response. So we want to go through some of the issues that are commonly cited by people who are opposing data centre build out, particularly kind of the local level of opposition, and go through some of the potential factors that are driving this kind of opposition. So Nancy, I just wanted to start with some of the concerns around the environmental impact. So particularly things like power use and emissions, which are coming up really commonly in these kind of anti-data centre movements. What's actually driving that concern and what's the sort of the source here of this concern around data centres and power?

Nancy Hardie

So it's a very legitimate concern, but not all communities will have the same concerns. So for example, if the energy demand from a given data centre requires upgrades to the grid or new power generation, then there might be a worry that some of the costs could ultimately feed through to households and businesses. In other places, a key concern could be emissions, for example. So particularly if new electricity demand is met by using on-site turbines using natural gas rather than from low carbon generation. So which of these issues matters the most, and often there's a combination, really depends on the facility's size, its location, how it's powered and what cooling technology they're using.

Lizzy Galbraith

So Tim, your background is slightly different to everyone else on the pod in that you were a former National Grid engineer. How do some of the UK data centres and the UK connected projects differ from some of these US kind of data centres? Are there different concerns and different mitigating factors involved if you look at geography to geography?

Tim O’Reilly

Yeah, sure. So that's right. So I spent about a decade sort of planning the network for these kind of industrial uses. So, you know, what data centres, they have quite an interesting way of operation. So effectively what a data centre is doing is it's taking electricity and converting that into its data. And so it is effectively a bit of a heater. It creates heat from electricity and that cuts into several other bits of the commodity use cycle. In terms of how they're powered, the data centres operate in a mode that's sort of what we refer to as sort of base load. So where they kind of operate on a 24-7 basis, their output tends to be fairly level. And if you take them in clusters, they tend to be quite level across a cluster. So they have sort of consistent jobs going all the time. What that makes them is quite useful to the system in terms of their ability to sort of absorb renewable energy in the system, but they also act as a bit of a buffer to other generation that's happening. So where that shortfall from renewables is made-up by nuclear or gas, that gas generation as the marginal may be associated with, say, a big data centre project. However, that gas generation still needs to run to make sure that there is enough spin up in the system so that when load varies, somebody turns a light, they put their kettle on or they turn on their oven, the system can react to that and it doesn't sort of collapse because a wind turbine can't go any faster. The wind isn't on tap if you like, whereas the valve to a gas turbine is. So while we can sort of look at these kind of different use cases in isolation, the way we have to look at the system and plan the electricity system is in a sort of total aggregate sense. So all of those different bits of the system work together to create an amorphous mass. And so when we look at decarbonisation and when we plan the system to be decarbonised further, we look at what total demand is. So we have allocations for different types of technology in the generation system, but we look at total demand. So everything needs to kind of work together as a system of systems. When we look at that across different jurisdictions, the UK is relatively unusual in that we have a highly interconnected system. So all of our generators are sort of connected together with not a huge amount of constraint. A lot of the talk about, as Nancy kind of pointed to, sort of that network build is about removing those constraints so we can more efficiently move electricity around the system around the UK. So at the moment, there are some restrictions in bringing very high wind output south to where, for example, data centres are located, but actually where the population centre is in the UK in the southeast of England. So that is currently constrained. That requires more cables or wires to move that to move that energy around. When we look at the competing markets, say West Virginia or the eastern seaboard of the US, they typically have much less interconnection between those large generation sources. And quite interesting why the industry built up around West Virginia. It's because you're close to the political center of the US, but also close to the North Virginian coalfield, where there are lots of power stations and the ability to build more power stations to supply that local demand. What's not happening there is importing, you know, energy from different states. So that is a sort of catch up that the US market is playing in certain senses where they're building lots of wind in the Midwest and then trying to import that into other bits of jurisdiction, but they just don't have that kind of industrial focused network that we have in the UK and to the most part Western Europe.

Paul Diggle

Yeah, I think that difference between the UK system being much closer to being a genuine national grid and the US system being a little bit more regionalised is quite critical to why local communities in the US, some of where there are issues with electricity and power usage and perhaps local grid capacity and local cost, energy cost comes from and why that picture is a little bit different in the UK.

Tim O’Reilly

Absolutely. So I think if you stand in a data centre campus in the US, you can probably see the cooling tower of the power station that's powering it. And I think, and that is a much more kind of local direct impact that you can kind of see, which doesn't really carry over in the UK context as much.

Paul Diggle

Nancy, let's talk about water as well, which obviously is another area where there are a lot of concerns about data centres and it's a particularly emotive topic as well. What is the nature of the concern with water and data centres? Is it a justified concern?

Nancy Hardie

So the concern really varies depending on if the region is water stressed or not and also what technologies the data centres using. The amount of water a data centre uses can be really significant. And just to give a sense of the upper end, the Environmental and Energy Study Institute, they have estimated that a medium-sized data centre can use up to 110 million gallons of water per year for cooling. And the largest hyperscale data centres, they can consume up to 5 million gallons per day, which is the equivalent of a town with around 10,000 to 50,000 people, and that's per day. So you can see that water consumption really varies by the size of the facility. On the technology side, one of the main concerns is that many data centres rely on evaporative cooling, which uses a lot of water. And this is where water evaporation helps remove heat. And one way it does this is through cooling the air, which then cools the servers and other equipment. Some newer facilities, they use closed loop systems and air chillers and that can reduce on-site water consumption. So closed loop systems, that's where they circulate the same water around the cooling system so that very little water is lost. And air chillers are basically like air conditioning and so they generally avoid on-site evaporative water consumption. There can be a trade-off between water and electricity use. So evaporative cooling can reduce cooling related energy demand, but they consume more water. But some systems that use less water, they can use more electricity. I think it's also important to remember that a data centre's water footprint also extends beyond the facility itself. So for example, lots of water is used in electricity generation and semiconductor manufacturing, meaning that actually the total water footprint can be a lot larger than the water used directly for cooling. And in terms of comparing water use from data centres to other sectors, such as globally, data centres actually use far less water than in agriculture or heavy industries, for example. But seeing things from this perspective in this global context can actually miss that important local context.

Paul Diggle

Tim, Nancy spoke about the difference between evaporative cooling and closed loop cooling and the impact that has on local water usage, at least in a data centre. Tell us more about that choice from a sort of engineering design perspective.

Tim O’Reilly

So, I think Nancy really set out, it is a choice. So you can build a big, cooling tower if you've seen a sort of 1960s power station, that's what you're sort of thinking about, big concrete tower and steam coming out of the top. That's taking that energy away from, you know, the chips via a water system within the facility typically, whether that be air-to- air or air-to-water or water to the chip in more modern data centres. That's a very efficient method of cooling things. It's compact, it's quiet, and kind of just does its thing. It's pretty failsafe. And it's typically what you'd see in a US-style facility. So that's much more typical in the US. What you would more likely see in a European and typically a UK-based data centre, you'd see that closed loop cooling system being put in. So there are lots of environmental regulations around water usage. You know, we in the UK, you apply a hierarchy of solutions to your environmental impacts. And therefore, if you can use a closed loop cooling system, you should use a closed loop cooling system. There are sort of hybrids where you can have these closed loop cooling systems with big radiators on the roof of the data centre. So from a volumetric perspective, it limits the shape of the data centre that you can build and potentially the number of stories that you can go up and fit lots of high-density computing. So it kind of defines your land take area somewhat. Those closely radiators can operate in two modes, so you can kind of just have big fans that are pushing lots of air through them and the higher the load on those radiators on a hot summer day, the faster the fans run and push more air through. It's kind of quite straightforward physics. The alternative option is when they get very hot, you can spray water. So you can have this kind of hybrid thing where effectively you're using a closed loop system, but you're limiting the running of the fans by spraying a bit of extra water on to then evaporate off in a hybrid fashion. We don't design that at Tritax. That's not something we've gone down the road of. But you can build large facilities with closed loop cooling systems. It does have other impacts on communities. So obviously fans running, there's a lot of mechanical load. So you've got big pumps to run that. You've got big fans that are high up, typically high up on the building, which then radiate out a sound pattern that needs to be managed and it needs to be well considered within the design. But those facilities are very water efficient and they're not, and they're not using a huge amount of possible water in those exact jurisdictions. But as Nancy points out, obviously there are huge amounts of water being used in energy generation and for the equipment that's being manufactured.

Lizzy Galbraith

Tim, you mentioned noise impact and some of the trade-offs in different types of technology that can mean that data centres maybe create more or less noise depending on the technology that they're using. Noise is something that does come up in some of these local oppositions to data centres. Can you talk about what actually creates the noise that data centres can emit and how different design and operating practices can create less noise or more noise depending on the trade-offs?

Tim O’Reilly

Yeah, absolutely. So I think the thing to point out is that, compared to say an industrial or manufacturing facility, data centres in the kind of the peak noise generated aren't hugely noisy. They're not that noisy. What you have to consider when we look at noise is, you know, the radiative pattern of that noise. So where does that noise go and how does it impact people? What times of day is it noisy and what is the kind of nature of the noise? So there's quite a big difference between, say, blasting in a quarry compared to a kind of low, gentle hum from a motorway, for example. They affect people in very different ways. When you look at a lot of data centres in the UK, so particularly in Slough, they're in the Heathrow flight path, so noise is already quite a consideration for those communities. Data centre noise tends to be 24-7 constant. Heathrow doesn't fly in the night, for example, to try and mitigate that noise risk. That kind of doesn't work for data centres. They need to operate 24-7, albeit load typically is a little bit lower overnight because there are fewer users, you know, lots of people are asleep. But that tends to be when a lot of backend sort of computer systems run and parse data and all that kind of stuff. So I think what you're kind of looking at in that sense is if you've got a sort of residential area, you're trying to probably direct that sound away from local communities, away from residential so that they're not having that sort of impact overnight, for example, and you're moving it more towards a commercial area. So moving that noise towards a commercial area, a low hum from a data centre during the day isn't as necessarily as kind of intrusive as you might find elsewhere. So there are lots of considerations around that in terms of the design of those facilities, but it comes off to other impacts. So, effectively you have to sort of orient these systems in a three box system. So you've got an office to, you know, house everybody. You've then got the data halls which contain all the compute and then you've got what we call the mechanical and electrical gantry which houses a lot of the generation equipment, you know, the backup generators, the switch gear and stuff and the coolers are then on the roof. Often visual amenity is a really big part of planning in the UK. So we try and orient buildings so that they kind of fit in with the existing road layout or the visual sight lines that people have. So there's this trade off between how does a building look compared to how does it actually function and where does that sound go towards. So you can do quite a lot in terms of noise mitigations, makes buildings larger, makes them less efficient. And I think that a lot of this stuff, all of these kind of issues can be dealt with reasonably. But what it does drive in is cost and time. So cost, when we're sort of competing in the UK and lots of other jurisdictions are all sort of competing to have these large-scale facilities which pay lots of rates and they provide quite a lot of jobs. You know, it's all about efficiency. It's all about the cost, ultimately the cost of those services that we all use to the end user.

Lizzy Galbraith

So, Nancy, how do you go about weighing up those types of trade-offs? So how should we consider the kind of the noise impact, the implications of some of these environmental factors on communities versus the national or regional economic benefits of some of these infrastructure projects?

Nancy Hardie

Economic benefits, they matter, but that doesn't mean that local impact should be discounted. As Tim's explained, developers should prevent or minimise harms that materially affect well-being and local livelihoods at the outset of any projects. But where some impact is unavoidable, there should be tangible benefits to communities in return, whether that's through things like infrastructure investment or local employment. And projects that do this well are more likely to secure local support.

Lizzy Galbraith

So just on that benefit point, it's something that comes up an awful lot when you hear about sort of opposition to data centres is that there's often a lack of understanding from local communities about what exactly they receive in return for hosting this data centre and for having this sort of new, often very large sort of industrial building hosted near them. So what exactly do local communities get in exchange for hosting a data centre in terms of employment, financial incentives? What options are there that you've seen offered in exchange for some of these buildings?

Nancy Hardie

Yeah, in terms of temporary jobs, depending on the size of the data centre, construction can create hundreds or even thousands of jobs, but once they're operational, they tend to employ far fewer people and that's because a lot of the facility is automated. Permanent operational roles, they're mainly focused on keeping the facility running, so things like maintaining power and cooling systems, also supporting network and IT matters. Of course, there's variation depending on the size of the facility. So small and medium sized facilities, they might employ only a dozen or a few dozen people, while the largest hyperscale campuses they can support hundreds of permanent roles. But in terms of other benefits to communities, we're increasingly seeing proposals include commitments to local hiring, education partnerships, infrastructure investment, for example, which will of course, very project to project, but unsurprisingly, the projects that are more likely to secure local support are those that develop benefits by really understanding that community context from the outset rather than it being an afterthought.

Paul Diggle

Tim, can you give us a sense of community involvement in the planning process in the UK where your experiences and perhaps how that differs from from the US. And one of the thorniest issues within this area is sort of cash incentives to local communities and the way that sometimes that can backfire because it's seen as sort of buying consent. So some thoughts on that as well.

Tim O’Reilly

Yeah, I mean, you know, kind of the UK planning system is quite, you know, quite specific in the way in the way that it operates. I think as Nancy says, from the outset we have to engage with local communities and we do do that. I think one of the first things that we do when we enter on the planning journey is go and engage with the local community, and that's about really inviting the community to come and review the plans and feed in, with the architects, with the planners really at the outset of the project. And listening to those local issues, I think it is really important. Projects are incredibly localised. I think the, and I can give you some examples in a minute about, sort of what the community might ask for versus what we might think we would kind of put as a headline policy. When we look at that kind of backdrop in the US, the US is a very different system around allocation. So the UK, obviously you're taking a piece of land, the automatic answer is no you can't develop it until you have met a burden of proof to prove that your harm is appropriate and your economic impact is appropriate in that trade-off. In the US and in lots of other jurisdictions, these commercial areas tend to be zoned. And once that area is zoned, there's a much lighter touch challenge and review of some of those projects going forward. So, some developers, and you're in the hands of the developer at that point, for want of a better phrase. Some developers will be really good at engaging local communities and bringing forward really positive developments, whereas obviously, you know, in a light touch sense, you can equally bring forward a really poor development. We're not immune from that in the UK. You know, local authorities are stretched to assess these large developments. We're under a lot of economic pressure to bring these forward. They are nationally significant pieces of infrastructure and that has a sort of different planning trade-off which, potentially takes some of the focus away from a local scheme and moves it into more of a national context. We've seen that in quite a few data centres with the revised national planning policy framework and the policies of this most recent government. So, you know, what we have to do is be sensible developers to bring forward appropriate development. The burden is on the developer to bring forward those developments in a sensible way in the first instance, but obviously we need checks and balances to maintain that. In terms of buying support of local communities, we take the view in Tritax that that isn't necessarily the best thing to do. We want to make positive, enduring benefits to the community with our schemes. And that will be different in different locations and across different scheme types that we look at. We brought forward a data centre recently, a proposal that's gained consent in West London. One of the big things from the community when we engaged was that they would like from a nearby residential area, they'd like better access to a nature reserve sort of on the other side of quite a large parcel of land. So what we've done is then incorporated that into the master plan for the site, provided a cycle path and footpath all the way down to the nature reserve. And that's been really well received by the community. It makes a really positive benefit to the sort of green space and the accessibility of some of those really important nature reserves and gives access down quite a picturesque river in an otherwise very dense urban environment. So, the context of that was really well received and also well reviewed by the planning inspector who really found that a positive attribute to the design of the scheme.

Paul Diggle

Lizzy, let's pull the lens back a bit then from specific scheme designs to thinking about the politics and the geopolitics. One of the things you said early on was that the opposition today tends to be something of a bipartisan issue in the US. Why is that? And does the sort of critique, as it were, differ between, you know, the Republicans, the Democrats, the left and the right?

Lizzy Galbraith

So I think it's partly because all of the issues that we've discussed do on some level sit at the intersection of several broader anxieties that we know are very common among Americans at the moment. So particularly things like the cost of utility bills, they've been rising for a number of years in the US. We know that cost of living is a very, very significant concern to a lot of Americans and it affects a lot of wider decision making on politics and other issues at the moment. Another factor is also potentially distrust of large corporations and a feeling that large companies can sort of come into often quite rural communities, very disconnected from those types of businesses and start making decisions. As Tim said, because an area of land is potentially zoned, there isn't necessarily going to be that much engagement. And we've seen that kind of corporate backlash be a factor in some people's decision making. It also cuts across left and right in a lot of cases. Local democracy concerns, the potential for communities to not really have any engagement again depending on that zoning issue and not really have any of their concerns directly addressed by the development itself prior to permission being granted and potentially then it going to review only as a result of organised kind of backlash following that initial approval. Resource use also does potentially become a bipartisan issue. Sometimes the language can differ. So if you're of a left-leaning persuasion, then potentially you're kind of voicing those concerns more around the environmental impact, sort of protection of local nature, concern about greenhouse gas emissions, for example. If you're on the other side of the political spectrum, then it could potentially be around preserving agricultural farmlands or protecting land use rights. So a lot of these issues are potentially described in slightly different ways by people with slightly different politics. But at the core of the issue is actually something that generally speaking is actually remarkably similar. And I think there's also underneath all of that, particularly for a lot of these often quite rural communities, a concern about AI itself. You're not necessarily seeing people who have experienced any benefit from AI so far if you're building a data centre in a sort of a very rural part of a predominantly rural state. So you could still have a lot of concerns about the way that AI is going to affect the overall economy and job prospects in the future without necessarily having any immediate experience of that yourself. So I think there's an awful lot of different potential sources of this, and particularly if you have a system that doesn't automatically imply a level of community engagement to try and address those concerns, sometimes it can snowball. And I think that's partly what we've been seeing over the past few months.

Paul Diggle

And there's a midterms angle to this, right? Obviously, we're going to talk about midterms increasing on the podcast in the run up to November. But tell us why, you know, gubernatorial races matter for data centres, how AI policy is sort of being the different candidates have been scrutinised, how AI donations is a bit of a focus now in the US too.

Lizzy Galbraith

So as we said a fair few times, this is often a very, very local issue. You're talking about local planning, local politicians and local democratic accountability. So a lot of the time you're not necessarily thinking about what's going on in Washington, although it should be said that this debate is playing out there too. And there's often quite active debate about AI in data centres there as well. But a lot of this is actually decided at a very local level. And the governor of any particular state therefore has the ability to make blanket planning decisions. And we have seen instances where previously relatively supportive or at least relatively hands-off governors facing re-election this year have introduced temporary moratoriums or new restrictions on data centre applications. So Kathy Hochul in New York, Greg Abbott in Texas, previously very, very supportive of data centres, and more recently, Josh Shapiro in Pennsylvania, all up for re-election in November, all of whom have announced some level of new restriction or pause on data centres. And this also matters for people kind of lower down the ballot as well. So there are currently active data centre applications in 200 competitive house districts in the US and a lot of the time you're starting to see people be asked about their opinion on whether that data centre should be approved or not. It is starting to become a more active political choice for politicians rather than just something that they can say, that's for, this county's approval boards to deal with. It is becoming a broader issue. People need to have a view or are increasingly expected to have a view. And that comes into question, particularly I think on the Democratic side where there is definitely greater scrutiny at the moment based on polling of who you're taking campaign donations from. This is always a question that we see in American elections, you know, what can we see about a candidate and the political choices that they want to make based on which corporate entities or kind of high-earning individuals they accept campaign donations from. And that does start to become a bit of a lightning rod depending on which races we're talking about and how closely these candidates are scrutinised. So it is definitely something that I think is going to be a factor in November. And I think even more than that, it's actually likely to be something that we see crop up in presidential primaries of both parties ahead of 2028 as well. Nancy, one thing that is important to kind of get our heads around with this is what actually happens when a project meets local resistance. So we've talked about some of the reasons why that resistance could come about, what it's doing to national level politics, particularly in the US. But what actually happens to data centre projects when they run into local resistance? Do we see developers make different choices and change their plans? Do we see them just move to different areas? How have we seen the response to this backlash play out from a developer perspective?

Nancy Hardie

So developers can move, but moving somewhere else isn't always the first response. They might first try to address concerns through project redesigns, community engagement, as you said, they might challenge planning decisions or delay the project. Where projects do move though, things like power, connectivity, land and a viable permitting environment, these are key deciding factors. So there are limits to how mobile they can really be. At the same time, some jurisdictions are actively trying to attract data centres. So the EU, for example, the EU increasingly treats data centre capacity as strategically important to competitiveness and digital sovereignty. But many of the same concerns around power and water and other local impacts still arise there. Projects in the EU often have to navigate a more extensive set of environmental planning requirements, so many of these issues are debated earlier in the development process than they might be somewhere else. That can make projects more complex and slow the permitting process, but it can also provide more opportunities to identify and address concerns before construction goes ahead.

Paul Diggle

So Lizzy, let's talk about the geopolitics as well, because we've spoken on the pod, and written about the way in which AI might be a race between the US and China. It's certainly part of the strategic competition between those two countries. We've also spoken on the pod before about the idea of choke points that the global economies become increasingly defined by these sort of bottlenecks in the economy. Is compute, compute capacity, what data centres provide, the next strategic choke point, analogous to critical minerals or control of semiconductor production or control of the Strait of Hormuz and oil flows?

Lizzy Galbraith

I think it's definitely got that potential. I think particularly if we do start seeing the US build out be meaningfully constrained by the political dynamics that we've been discussing. And the question then becomes, where do you go that would allow US companies to continue building on the trajectory that they had planned? Then the question then becomes where? And we have seen instances of some countries say, well, we're happy for you to build data centres here, but we'd like access to your most advanced chips. And that suddenly becomes the trade off. So I think that's definitely got potential again, if we do see the US build out start to slow meaningfully in order to keep the momentum going, that does become a more frequent conversation. So Tim, do you think that the UK could be one of those places that US developers would potentially look to if US build out becomes displaced? Are there factors that would sort of help or hinder that in the next few years?

Tim O’Reilly

Yeah, so I think looking at the UK from a kind of raw commodity perspective, if you like, from a data centre market, the UK is uniquely well placed, sat between the US, where the majority of data centre infrastructure in a state is, and Europe, where there are obviously there's a huge user base of sort of these forefront technologies. Adoption rates are high in Europe compared to internationally and the UK sits on those critical bits of telecoms infrastructure - so big undersea fibre cables that link all of this infrastructure together. The UK is a really attractive place typically for the US to do business. Increasingly we're seeing lots of additional regulation from the European. union in terms of data sovereignty and the kind of how that has to operate. The UK seems to take a much more pragmatic approach to that. We share a language with big American tech firms, which again makes it easier for them to come and develop here. We share that kind of sort of cultural overlay as well where we can do business together in a quite straightforward fashion. It's a well-trodden path for lots of industries. What's constraining the UK from sort of really grabbing hold of that? I think we've talked about planning a lot. From a US tech firm perspective, they tend to want to deliver stuff in the way that they deliver it in the States and they tend to be quite inflexible. Leads to quite a bit of consternation as to how we sort of operate in the UK, even more so in Europe. And then if you look at the kind of constraining factors more broadly in the UK, typically it's access to capital, access to capital can be quite tricky in the UK, particularly for early stage build out of infrastructure, which is sort of coming first in that AI race. You know, a lot of the AI companies are still, you know, in their sort of startup or build out phase, and they're not kind of big household names that make it really easy for institutional capital to get behind. And when we look at the kind of the key commodity that they need, electricity, increasingly we're seeing delays from the National Grid in NISO in terms of building out that infrastructure, so that physical build to move that electricity around and build more electricity generation to supply these large facilities - that network build out is really that constraining factor. And the more quickly the UK can get on top of that and get it built, then it frees up locations in the UK that can support data centres. And it doesn't then just need to cluster around existing locations, which potentially eases that planning burden. And then also, you know, this kind of free and easy enjoyment of to consume that, you know, decarbonised electricity that we spent the last sort of 10 or 15 years of effort bringing forward, which is really unique and really useful.

Lizzy Galbraith

So Paul, we've discussed some of the issues at play, we've discussed some of the international variation, and we've discussed some of the potential ways that US companies would be able to mitigate the US data centre backlash if it persists. But how significant risk do you think this backlash actually is to the AI theme as it currently stands, as the market currently understands it? Is this something that we need to be including in our kind of calculus going forward?

Paul Diggle

So there's clearly an excess of demand over supply of compute at the moment. because compute demand itself is just rising very rapidly with the advancements in the models, their usefulness, but also their token burn. And as we've heard, physically building that compute and all the real world complexity that comes with it is a genuine constraint. I think it's more constraint than sort of thesis breaking or like existential to the AI thematic or driver as it were. There are some bigger risks which we talked about on the pod. It could just be a bubble. There are valuation challenges. There are questions about future earnings and monetisation. A lot of the stock market relies on very large future earnings being realised. There's competition from China in terms of leading-edge models themselves. Again, another thing we've spoken about on the podcast is regulatory headwinds potentially. And then there are these real-world physical data centre construction energy constraints, infrastructure constraints as well. I think an important distinction to make is risks that would stem from AI not actually being all that useful in the end versus risks from AI being very useful and economically transformative, but the physical layer being difficult to build and to keep pace with that because a lot of the risks we usually talk about are that maybe actually it's all a bit of a flop in the end. But the more you believe in the eventual productivity economic benefits, the more the infrastructure will be required and the more a constraint the planning process, the community backlash that we've spoken about becomes a constraint. But look, I think it's more likely to increase costs, lead times, cause project delays than it is to halt the whole thematic itself, which of course has been an enormous part of performance of equity markets and a big part of what's underpinning US and the global economic growth as well.

Lizzy Galbraith

So let's close by bringing this back to data centre construction itself. So presumably the ideal outcome is something between unfettered building and total bans on data centre construction. So I just want to close by asking both Tim and Nancy, what does a good data centre construction and data centre policy actually look like? Tim, would you like to go first?

Tim O’Reilly

Yeah, sure. I think, you know, from the policy angle, we need to be clear and unambiguous about what the constraints and the solutions we want developers to bring forward are. So if that's no water consumption, then let's go with that. You know, and what does really great design look like? It looks like something that we can bring together that works for communities, works for those local residents, but also works for those technology operators in the long term. Really, really tricky at the coalface to sort of bring those two things together. But I think, these are within the realms of possibility of things that we can bring forward.

Lizzy Galbraith

And Nancy, what does good data centre planning and building look like from a sustainability perspective?

Nancy Hardie

I think good policy should recognise both the benefits and the costs. Data centres are increasingly important for AI development, digital infrastructure and economic competitiveness. They support technologies that people use every day, like in healthcare, for example. And many governments are increasingly seeing them as strategically important for national security. But good policy should also create clear rules around the issues we've talked about, so developers and local communities know what to expect. And it should also help ensure that data centres are developed in places where the energy and water resources needed to support them are available over the long term. Actually, many of these challenges, they're not unique to data centres. There's always been a balancing act between nationally important infrastructure and local impacts. Think of, you know, wind farms or power lines. But if communities are being asked to host resource intensive developments, the benefits need to be visible and locally relevant. It's not just about doing the right thing, but projects that fail to earn community support are more likely to face delays and other challenges like we've talked about, regardless of how strategically important they might be.

Lizzy Galbraith

So it's not often that we happen upon a podcast topic that can take us all the way from local planning policy all the way up to the future of geopolitics and supply chain choke points. But I think that does speak to the nature of just the sheer scale of the AI build out and its implications. And it does have, in a way, the challenge of having to deal with local politics and the very significant variations therein. As we've discussed, there really is no two circumstances that are exactly alike. And that is a challenge that a lot of these companies are only just really starting to grapple with. But the implications are truly international and have the potential to really shape geopolitical alliances going forward, depending on the way that this plays out in the next few years. So with all of that, then the only thing left for me to do today is to thank Tim and Nancy for coming on the pod and discussing this with us, and to encourage you all to like and subscribe to the podcast if you haven't done so already. And we will hopefully see you on the next episode.

This podcast is provided for general information only and assumes a certain level of knowledge of financial markets. It is provided for information purposes only and should not be considered as an offer, investment, recommendation or solicitation to deal in any of the investments or products mentioned herein and does not constitute investment research. The views in this podcast are those of the contributors at the time of publication, and do not necessarily reflect those of Aberdeen. The value of investments and the income from them can go down as well as up, and investors get back less than the amount invested. Past performance is not a guide to future returns, return projections or estimates and provide no guarantee of future results.