The battle for nuclear and data infrastructure with Mustafa Latif-Aramesh (TLT LLP)

Hello. I'm Ed Porter, VP of Insights at Moto Energy, and you're listening to Transmission, where today we'll explore the story behind the pace of nuclear deployment and whether we can really enable the wave of data centers looking at the UK. In the energy world, there's a sentence we hear all the time. We need to build more.

More grid, more clean generation, more storage. And now because of AI, we also need to build more compute, more data centers, and fast. But the uncomfortable truth is in Britain, the limiting factor isn't only engineering, it's permission and permits. Our guest today is Mustafa, a partner at TLT, where he works on infrastructure planning across everything from transmission lines and batteries to wind, solar, and nuclear.

And recently, he's been right inside government. Most of us served on the prime minister's nuclear regulatory task force set up by Keir Starmer to figure out why building nuclear in Britain takes so long and what to do about it. They came out with forty seven recommendations, but the diagnosis is pretty simple. The system is risk averse, full of duplication, and not always set up to reward delivery.

In fact, the rules can create perverse incentives like encouraging ten thousand lorry journeys just to avoid a site being labeled as landfill. Then we get to the story that may have made it into your group chat. Hingley Point C's fish protection system. Yes.

It's the fish disco. Wildly expensive mitigation that doesn't clearly match the environmental benefit. Imagine spending seven hundred million pounds with the impact measured in fractions of a salmon. None of this is a pitch for weaker environmental standards.

It's the opposite. Mustafa argues that if you care about nature, you should care about effectiveness, not box ticking. But the conversation doesn't stay inside nuclear. The same regulatory machinery is about to get stress tested by data centers.

The UK went from roughly fifty data centers in the year two thousand to around four hundred and fifty today with a pipeline of roughly five hundred more in planning or construction, each with local imperfections of their own to navigate. How will the planning inspector cope with that growth? How do we build more without losing the outcomes we actually care about? Welcome back to transmission.

Hello, mister Phil. Welcome to transmission.

Thanks, Ed. Thank you for having me on.

Our pleasure. And as ever, please could you introduce yourself and your company to our listeners?

Yes. So my name is mister Philiti Faramesh. I'm a partner at TLT.

TLT is a law firm. My team specializes in infrastructure planning, so we help get consent for every type of development you can think of from data centers to transmission lines to battery, solar, wind, nuclear.

Okay. And last but by no means least, nuclear. And, obviously, you've been working on that a lot over the last few years. So what has been your role in the nuclear space?

So the thing that's worth talking about is the most recent escapade into nuclear, and that's my role on the government's nuclear regulatory task force.

The task force was set up by the prime minister, and it's comprised of our chair, John Fingleton, who's an eminent economist, and then three others on top of myself. Three others span from a former chief nuclear inspector, two people with detailed experience in the defense realm, and a professor from the University of Manchester. And we were established by the prime minister to help give some recommendations on how the regulatory planning and environmental sphere in nuclear could be reformed to help expedite the delivery of new nuclear.

Okay. Now we have a report that's recently come out. How does that change the potential for nuclear delivery in GB?

So it's probably worth saying that the report diagnoses a general issue in the delivery of nuclear, and it's comprised of three elements. The first is risk aversion. The second is duplicative requirements and a lack of incentives. And the last is ensuring that the mindset is geared towards delivery.

And that manifests itself across forty seven recommendations, which I won't outline straight away. But effectively, what we're looking at is a system in which we're trying to ensure that the outcomes that we're trying to achieve, good environmental sustainability, good safety records, is put at the top of the priority list rather than the process that leads to them. What we've seen over recent years is that the process is often considered to be more important than the actual outcome that we're trying to achieve. And one of the issues that we identify is that this is not an issue with individuals in the system, and we don't single out any single entity, any single organizational type of organisation.

We call it a system problem. And so the recommendations are geared towards fixing those issues by removing the duplication, ensuring that everything is geared towards the outcome we're achieving, and also ensuring that where there is risk aversion in the regime, that it is proportionate to the level of risk that is actually involved.

That's that's pretty clear. I suppose if I was as a as a listener to this, I think you'd be thinking, well, this sound these all sound like really good changes. Maybe we could put a few kind of concrete examples on it. So you mentioned that sort of the process had been more important than the outcome.

Is there an example of that that you could point out in today's process?

So one of the case studies that we have in our report is about the decommissioning of of of nuclear at the Winfred site, which is just just on the south coast. And the issues that we identified there were, as part of their decommissioning, they would have to apply for a permission to their local planning authority, they'd have to apply for a separate permission to the environment agency, and they'd have to apply for an entirely separate permission from the Office for Nuclear Regulation. So immediately, you have three regulators involved in one activity. That activity is something worthwhile.

Decommissioning of nuclear fuels, waste, etcetera, is a very good thing to do for the environment. And what you find is there are slightly different requirements across different regulators and across different regimes. And they would be having to go through these processes and have to produce slightly different documents. But the real problem with this approach is that, say, one of the permissions that's required from the environment agency incentivizes them to rather than keeping the rubble that they produce as part of their activities on the site, they would be incentivized to send ten thousand HGVs, lorries onto the road network to avoid the site being designated as a landfill.

That doesn't seem like the right outcome. It doesn't seem like a good use of public resources to have three separate regulators looking at one specific activity. And it's certainly not good for the environment to have ten thousand trucks going onto the road when we could be doing it in a much more sustainable way. So our recommendations in that context are very much focused towards, there should really be a lead regulator.

We also have a recommendation for a single body that would be potentially responsible for the activities on the site. And we also have a desire to make sure that every one of the people involved is thinking about the outcome, which is not to put ten thousand trucks on the road, but to ensure the sustainable and quick delivery of the decommissioning on the site.

Okay. And is that one body that would oversee each individual site, or is that one body that oversees the sort of total nuclear fleets in GB?

It's the latter. So what we've recommended is the establishment of what we call the Nuclear Regulatory Commission. And effectively, that would sit as a super regulator that would have the powers of all of the existing regulators.

It could deal with applications, issues by itself. It could also be used as a way of determining any issues that needed escalation, if there needed to be appeals. It's effectively trying to put everyone under one roof so that you have a greater amount of coordination across the different regulators.

Okay. Let's then instead of talking about just just the sort of large nuclears that we've seen in in sort of recent history, what about this new wave of SMRs that are coming in? So small modular reactors. Are they bound by the same sort of double double take on the the planning process, or or do they have an ability to be delivered at a faster pace?

So as it as it currently stands, most of the rules, and I I I mean close to ninety percent of the rules, are identical for SMRs as they are for large gigawatt reactors. The slight caveat to that at ten percent is simply that if you're below a particular threshold, you go through a slightly different planning route. But so far as the regulatory context is concerned, it is the same. So the issues that we have identified, the recommendations that we have made, will be applicable to small modular reactors.

Okay. So we so we, you know, we build small modular reactors because they are modular, and we think that modular things are supposed to be delivered faster. But in reality, we might see a small modular reactor going through all of the same process as these much larger reactors, and so still could be sort of bogged down in that in in in in the sort of historically slow pace of deploying nucleosides.

Exactly right. And I think part of the issue that we've identified is that a lot of our rules are geared towards the gigawatt reactors. So one particular aspect, is just a kind of microcosm of this issue, is that the methodology, the test that everyone uses to decide where to site a nuclear reactor is actually based on a paper from the nineteen sixties, which is based on a reactor that was being built in the nineteen sixties. Safety has changed. Decay rates are different. Topography is different. All of these things mean that some of our rules which have worked so far for the gigawatt scale reactors are not suitable for the smaller the smaller reactors, but they continue to apply to them.

Okay. So given sort of free choice and the fact that we're refreshing some of these rules from the nineteen sixties, did the did the group think about where you might put a nuclear reactor today given free choice?

So I I I think it's worth saying that we don't make any specific recommendations on on exactly a specific site.

The way that the planning policy in this country worked up until probably about two years ago was that we had designated nuclear sites, which are the sites that you will have heard of at Oldbury, Wilver, Dungeness, which are existing sites for nuclear. And the government changed that policy from saying it's just just these that have the starting presumption to one in which all sites are suitable provided they meet a particular criteria. Our view is that the criteria that has been put forward as part of this refreshed approach needs to be adapted to ensure that more nuclear can come forward.

I should also say that even when we had these designated sites, which were the existing ones, there was still an ability to to build out somewhere else, but it didn't have the starting presumption that those sites were suitable. And so what we're recommending is effectively, it's right that you should be able to show that any site is suitable. The criteria approach is correct, and it should be grounded in safety, and it should be grounded in environmental considerations. But the current rules are not well adapted to making sure we get the most that we possibly can. And that example of part of the methodology being based on a paper from the nineteen sixties is exactly the kind of thing that we're trying to to push the government to change.

And maybe maybe one final one final question on nuclear that's more of a serious one, and then you might have to humor me on a on a less serious one.

But we are trying to build a lot of nuclear sites right now because we're taking a lot of sites offline and trying to sort of bring some on so that we don't see too much of a dip. But right now we do have a bit of a dip sort of for a couple of years.

Are the recommendations, are they sort of are they potentially fast acting enough to bring some of those new projects slightly closer to delivery? Or are those project are those sort of are those projects a bit out of scope in terms of things that will be hit by these recommendations?

So the the the desire and the kind of ambition is that it would be useful for projects coming forward in the immediate term. The report itself against each recommendation has a target date for when the recommendation should be complete. And it's worth saying that the prime minister, the publication of the report, has accepted the recommendations. The government is working on an implementation plan, which will specify the details of how they're going to go about actually implementing things. But the timeline that they're working to includes very specific aspects, and the anticipation is that all of the recommendations will be implemented by December twenty twenty seven. So it should have an immediate effect. Many of the recommendations could be completed, you know, halfway through next year, and that would have a real tangible effect on the delivery of new nuclear.

Okay.

Quick break. If you listen to this show, then you probably work in energy.

And at Modo Energy, we're not just talking about the energy transition. We help our users to actually make it happen. All energy storage, solar, and wind assets on the global balance sheet need to be valued and benchmarked, and that's where Modo Energy comes in. Our benchmarks and forecasts are transparent, bankable, and trusted by the world's leading banks, asset managers, utilities, and developers. So if you want to learn more, go to modo energy dot com. And if you want more content like this, sign up to our weekly dispatch newsletter. Enjoy the conversation.

And and I asked you to humor me with a less serious question.

Did the group talk about fish discos at all? Yes.

We spoke about the fish protection system.

Okay. We did we don't refer to it as the fish disco. You can, of course. But the fish protection system was put in place at Hinkley Point c, which is one of our new nuclear power stations that's being built in Somerset. And as part of their planning approvals, they had to include some mitigation for the potential impact on a number of fish.

Our report has a case study about the fish protection system, which is not quite as catchy as fish tiske. And in our report, it is an example of where the environmental process has led to a position where we have a solution, a mitigation measure that is not optimal, and it's not based on the outcome of protecting the fish, as the name would suggest. Our report found that seven hundred million pounds was spent on a fish protection system, which is comprised of many different elements. Not all of them relate specifically to the acoustic fish deterrent, which is what is being called the fish disco, and not all of it is specifically for mitigation. But the use of seven hundred million pounds, you'd expect to have a substantial environmental gain.

And we don't think so the amount of money that was spent or is being spent on it justifies the impact that's being achieved. So EDF's own assessments say that it protects approximately zero point eight three salmon, which is not a high number.

Now, it's worth saying that the figures for the number of fish which might be affected is subject to some kind of, you know, disagreement, and there are differences in the figures. Even at the highest levels, seven hundred million pounds is not, in our view, proportionate. And we highlight in our report very good examples of environmental groups using substantially less money to achieve very good environmental outcomes. And so one of the recommendations is specifically, don't focus on the process, don't focus on the specific thing that is being affected, but instead focus on what is the overall environmental goal that we have, and instead fund that, make sure that we're focused on that environmental goal rather than getting into a position, which is not the fault of any of the regulators or the developer in that case where you're spending and, you know, a very high value on something that Yeah. Is not the most effective thing to do.

You've been very kind to me. It was a it was a very silly question. You've given me a very sort of detailed and diligent answer. But you sorry. You said naught point eight three salmon?

Did did did is that It might be zero point zero eight three.

Zero point zero eight three. And is that a percentage or is that like a number of salmon saved per minute?

Like, what does because it's something you want, like I I believe that's the figure saved per per year.

Okay. So it's effectively saying that you might need to wait just over a decade to to save one.

It's just a totally from the outside looking in, right, I I obviously don't know how we got to this point. But from the outside looking in, it just looks totally crazy. Right? People eat salmon very regularly at the cost of, I don't know, ten pounds.

It is it is worth saying as well that this is this is not just an issue that exists in the nuclear context or indeed the energy context. The example that that caught the public's attention several months ago was the HS two bat tunnel, which is this idea that HS two would build this one hundred slightly north of hundred million pound tunnel to ensure that bats had somewhere to go so they weren't affected by the new railway. What we have found in the fish protection system is something that is the equivalent of seven times the HS2 bat tunnel. And this is a systemic issue, which is driven by particular application of the rules, the risk aversion, the duplication, the attempt to put the process over the actual outcome that we're trying to achieve.

And I think maybe moving on from nuclear just to just to more much much more general. Do you think this is something that's kind of everywhere throughout the planning system and that we are very much focused on the process and that means we're getting fish protection systems as well as sort of bat tunnels and that is slowing us down as a country. If it it feels very much like those two have been in the in the public domain a lot recently. But it feels like if if you were to just kind of scrape back a few layers on a few projects, we might see a lot of similar stuff to this in other places.

So we do we do say in the report that there are many lessons which could apply to different contacts. I do think that there is a more general lesson to learn, but it does vary depending on the sector and depending on the specific process that each particular project is following. So we're very clear that our recommendations because That's what our remit was for about nuclear projects. But there are probably some lessons to learn that reflect a more general concern with how infrastructure is delivered in this country.

Yeah. And and for for balance, like, we obviously are keen to make sure that that salmon, bats, whatever it might be, are also supported, but it's done in an effective way. Right? So we're not spending lots of money to save fractions of a salmon. Let's do let's do something that actually helps salmon rather than fish protection systems.

Exactly. This isn't an attempt to say that the environment doesn't matter. This is quite the opposite. We want to do better for the environment, and the current system is not delivering.

Okay. Let's move on from nuclear to data centers because your your job expands a range of these topics. And so let's talk about the just the sheer scale of data centers looking to build out in GB. And maybe say looking to build out in GB because we have some today, but not as many as we think we might have in the near future. So so what does that build out look like?

So it's worth saying that historically, the UK had fifty, around fifty data centers around the year two thousand. You fast forward to today, and it's closer to four hundred and fifty. And it's worth saying that forty percent of those are around London. So we've gone from we've we've had quite a substantial expansion in the number of data centers.

If you compare that to our European brothers and sisters, what you find is that the number of data centers in London is almost twice the amount that you would find in Amsterdam, Paris, Frankfurt. And so the UK is in a very unique position where we have already taken a bit of a head start on the number of data centers. And really, when you're looking for people who or countries who are doing better than us, you're looking to the US and China. So we're in a very good position, and the demand has already risen.

If you look at the pipeline of data centers, there are probably another five hundred or so which are either in the process of construction or in the process of planning. So there is already quite rapid expansion.

And those sort of five hundred coming, are they more like the hyperscalers, like the massive data centers in one place that could maybe be a gigawatt or a gigawatt and a half of of power demand? Or are we more talking about sort of edge scale, smaller data centers very close to their sort of end user?

So there's a lot of regional variation. So to answer your question, it really depends where you're talking about. So historically, one of the issues that data center providers have had to deal with is latency. So this is the idea that the distance between a user asking a machine to do something and then getting the result back.

What how long is that delay? How long does it take? Technology has moved on, but again, if you if you go back even just a decade ago, what they were seeking to do was to reduce literally the physical distance between where a data center would be or where any compute functionality would be and where it was needed. So you did have this kind of collation around the city of London primarily for financial investment purposes where you need very, very quick systems in place.

What we find moving closer to now is that, again, there is regional variation. If you've got the kind of hyper scale infrastructure, one of the things that they look for is, generally speaking, high-tech connectivity. So it's not just that you need power, it's not just that you need to make sure that you can be in a location which will attract the talent that is already there, but you also need to be close to a fiber optic system that will allow you to operate in a way that is most advantageous.

So what we find is that the hyperscale infrastructure to date has been based in very particular areas. And then the more kind of cloud storage, the slightly more run of the mill infrastructure can be placed in other locations where they prioritize low power prices. So short answer, loads of regional variation, but it's driven by exactly what the demand is, what the latency is, and what the power supply needs to be.

Okay. And to maybe put some sort of not not specific names from projects that you've worked on, but maybe just to help people at home to understand kind of who might use each one of those. So would you have a few examples of like a a company that might build a hyperscaler or would you have a a few examples of a company that might look at more sort of AI training, slightly higher latency use cases?

So an example of a a hyperscaler would be think big. So think Google, Microsoft, where what they are building is a system which can effectively process very, very complicated computer processes. So one of the things that many people in the industry talk about are advent of large language models, but also this idea of frontier model. And these are almost at the cutting edge of of of what can be produced in the AI sphere. So when we're talking about hyperscalers, we're talking about fairly sophisticated systems. Then you have something that's slightly different in that data centers can comprise cloud storage.

That's not to do down cloud storage. Cloud storage is incredibly important, and it effectively fuels and helps most parts of our economy function. There is less of a need to deal with the connectivity latency issues in something like that compared to a hyperscaler. You also mentioned AI training. Now, that is a a form of the hyperscale infrastructure, so it's much, much closer to the first kind that I've mentioned. But they're slightly different in that they are being used to teach the systems that are making you know, carrying out the requests that we make of it to be better, to do better, and that requires a huge amount of of compute functionality.

But doesn't necessarily have to happen exactly immediately and so could be based in another country or or slightly further away from I mean, we're recording this in London. So, yeah, in this case, could be further away from London.

No. And and and I think that's the other thing about the the storage element to this is that what you're doing is effectively asking for something to be stored somewhere so that you can access it, but you don't necessarily need it to run as quickly as a buy or sell on a financial market.

Okay. Now let's merge your skill and expertise in planning with the growth of these data centers. So if I have an idea to bring forward a data center when it's just that original idea, what what do the steps look like before that data center is actually constructed and operating?

So there's probably three key elements to this. The first is just the financial aspects. Obviously, committing to develop a data center is a very high financial cost and you need to make sure that you've actually got a financial model that supports a return on the investment. The return on the investment is driven by the factors that we've already discussed.

So that's things like latency, connectivity, location. Location also matters not just because of the land value that's involved in the site that you buy, but also the geographic variations that we might have in electricity prices. So your financial model needs to stack up. Once you've done that, then you pick a site.

Now, when you have a site, you need to go through the steps of acquiring the site, and then you need to start doing your preparation for the planning application. Now I'm sure we'll come on to talk about some of the details of a planning application, but in many cases for a planning application, also need to produce an environmental assessment. The environmental assessment has to be based on surveys of the site. You need to go out maybe once or twice a year to see if there are any birds on that particular site.

That then is fed through into your environmental assessment, which is included in your planning application. You submit your planning application to the decision maker, which could be a local planning authority, or if some of the changes come through, it could be central government itself. And they will make a decision.

Once that decision is made, you then implement your project.

One of the important things to mention in this context is that the implementation of a planning permission often has conditions attached to it. Now one of the things that we've seen particularly around West London and Slough is that when data centers get planning permissions, they will have conditions relating to water. Water is very important for most or many types of data centers because they are needed for cooling the systems themselves.

That means you need a good water supply.

And what these planning conditions say is that you need to make sure that your local water company has effectively signed up to giving you that water supply.

There are water constraints in the southeast, and so that is often quite a hurdle that needs to be get gotten over.

Some data centers don't need water cooling. They use air air cooled systems, which avoids that issue, but then you still need to go through the steps of implementing and effectively making sure you've complied with all the other planning conditions.

It's interesting because it's the same problem, right, for the nucleosides. They need water in order to be able to cool the steam and that's why they're all on the coast. But obviously with data centers because you need the latency, you can't do that. So you you kind of need to kind of balance the the availability of whatever you're using to cool the system with whatever the thing is actually going to be doing, which is very interesting. You now you mentioned you sort of dropped a a sort of a leading point into there around how those planning permissions could be granted. And so data centers have just been added to the NSIP regime. What's an NSIP, and why does that matter for getting projects approved?

So an NSIP is a nationally significant infrastructure project. And it currently comprises the the really big types of projects. So think of major airport expansions, major new roads, railways, gigawatt scale nuclear reactors. It's a system that came into place in two thousand and eight, which is designed to ensure that you can have all of the things you need in one permission. And the end result of going through this process is that you have something called a development consent order.

You don't have a planning permission, but you have something that is the equivalent. It gives you the planning consent that you need, and it's subject to a slightly different process. So the conventional approach is that you submit your planning application to your local council, and they spend some time looking at it and then make a decision. Nationally significant infrastructure projects are different.

They are submitted to the government who appoints an independent panel that will hold hearings and they will ask questions about your project. And at the end of that, they will then make a recommendation to the government. And the government, once they've got that recommendation, looks over it and makes a final decision. It's intended to be a kind of one stop shop with a number of fixed time frames within it compared to the more conventional approach, which is, you know, to your local authority doesn't have as much certainty around the time scales and also means that you can get powers for things like compulsory purchase and ensures that you have a specific process that you follow for getting to the point of your permission.

And so there is this process for bringing forward these data centers. At the same time, we're trying to do sort of Clean Power twenty thirty as well as sort of adding much more wind solar batteries to to the to the energy system as well as building out sort of transmission lines.

All of this feels like it's kind of being squeezed through a planning committee or a planning body. Does it does it feel like the planning inspectorate, which I think is probably the right term, are then running out of bandwidth to actually review all of this stuff? Because it just feels like we're trying to build so much all at once. Do they have the ability to actually get across all of these projects?

So I I I think on the planning inspector who you're right. They are the body that comprises that independent panel I mentioned for the nationally significant infrastructure projects.

They do have additional resourcing that's been made available to them. And I think the wider context for this is that the government has said before the next election that they want to make a decision on a hundred and fifty nationally significant infrastructure projects. In the first year, they did, I think, somewhere between twenty and twenty five. And so the government has acknowledged that in order to, you know, meet their ambitious target, that they will need to give additional resourcing. And I think, to their credit, the planning inspectorate, as part of this process, has been delivering some improvements to the time scales that are associated with this. So a combination of central government leadership, additional resourcing, they should be capable of of of meeting the target that's been set by the government.

Okay. Really interesting. And I suppose we also see some data centers coming forward and saying, well, we can't. Let's say we can get the land and our our environmental assessment is okay.

And we've got water, so this project is looking good. But we can't get the grid connection and we can't get the energy we want. So the solution is, well, we're gonna do something else on-site. So maybe that's gonna be, I would hope, solar and batteries perhaps.

Maybe it is reciprocating gas engines, which we see quite often. But also and to combine two topics from this conversation, SMRs and data centers. That feels like something which is incredibly difficult already in terms of bringing nuclear forward as well as a data center, which is, again, another challenge. Do you think there's a a likelihood that we'll see SMRs paired with data centers over the next, say, ten years in in GB?

So I do think that we will see colocation of power generation in data centers. Hopefully, we won't have to wait a decade, but in in the next kind of, you know, five five, seven years.

One of the announcements that the government made when president Trump visited was was actually a SMR data center development at the Cottam power station site. So clearly, people are thinking about it.

I know many people in the industry are talking about colocated power generation in data centers. I think if we talk about the constraints, we can talk about some of the solutions. One of the constraints is that data centers have this rule called the five nines.

And the point of five nines is that your power supply to a data center needs to be on ninety nine point nine nine nine percent of the time. I think that was the right number of nines. That means that you really need a stable power supply, and you need a backup for your primary power supply.

One of the issues that many industries have to grapple with is the intermittency of their power generation. I think what that means in practice is that whilst we will have colocated power generation and data centers, we're also likely to have a continued need for direct grid connections even if it's just for a backup. Because in order to meet that five nines test, you can't take the risk of just having one colocated site power generation. As things currently stand, as I mean, you've you've just highlighted, even the existing data centers which have a direct power supply will have some form of battery or some other form of generation on their site so that if anything goes wrong with their grid connection, there's something that can step in.

The colocated power generation element, no, I think, is increasingly becoming an option because one of the things that will help in getting your grid connection is having viable power projects being delivered. So it's it's slightly chicken and egg, but the grid reforms that are going through are clearly targeted towards first come, first deserved. And if you can make a case for why your particular project, which in this context is a colocated development, then you have a potentially a stronger case for then getting the backup that you need from the grid connection.

Yeah. It's it's it's super interesting that sort of that five nines. We talked a little bit about data centers on this on this podcast before, and I think the there's the five nines of availability. But sometimes, the data centers, they actually their their consumption, their utilization of the generation of the grid connection doesn't necessarily match that.

Some of them run at say sixty or seventy percent of the of as as as a load factor. So even though they demand the sort of availability, they don't necessarily always use the energy, which I think is a a sort of a fascinating part of kind of how we are using data centers. And one thing for me that I'm really hopeful for is that we can be flexible with our data centers in terms of how they use energy. We see this a lot in the in the US actually where the flexibility of data center consumption is one of the ways that we can essentially run our grid slightly more flexibly.

So we don't have to build that last gas station to make sure that we can meet the very very peak in the middle of winter. Wouldn't it be good if we could just turn down something slightly right at the peak so we don't have to build that last power station?

Well, there well, there's there's two points about that. The first is you're you're right that even though there's this five nines rule, data centers won't always operate at maximum capacity. In fact, they vary a lot over time. And that actually has a knock on implication in terms of the power supply because your power generation needs to be able not just to potentially power up the maximum scenario, but it also needs to vary in line with the needs of a data center. And some technologies have a a more difficult time of varying the power generation compared to others. Actually, of the issues specifically in the nuclear context is how how quickly can the power generation be adapted to fit fit the specific needs of a particular data center.

The second thing, though, I think is more commercial, and it's this point that even though there is variability and you can take a punt on saying, well, I think I can be I can make sure this power supply operates seventy percent of the time. The actual contracts for these data centers between a power generator and the developer of a, say, a hyperscaler specifically says you should be able to operate mostly according to the five nines rule. And if you work out the five nines rule, I think it over the course of a year, it's something like your system can stop generating for about five minutes. Now, that's a huge risk to take for developers. And so you want to make sure that you're able to meet the requirements of your contract so you don't have any financial penalties. Because in the end, these developers, whether they're the power generators or the data center developers, have expectations. And if you don't meet those expectations, there's potentially a high cost.

So I'm gonna move us on to a final question before our two final questions. So in reality, three last questions. We have a global race in AI and data centers.

How do you think GB is doing?

I'm going to be slightly diplomatic and not answer your question, but I'm gonna try and answer it in a slightly different way, which probably reveals more about how I think about the world.

It's not how we're doing relative to other people, and it's more, are we doing the best that we can be doing? And I think we could be doing a lot better. And it's quite, you know, hopeful that this government has recognized the importance of AI.

They've gone about publishing various AI strategies. They've got these ideas around AI growth zones, which will hopefully help speed up the delivery of these colocated developments with data centers. I think we also need to be realistic about what is the outcome that we're trying to achieve. Are we going to be the leading nation for data centers and hyperscale infrastructure that deals with frontier models?

I'm not sure. But we could be an AI global player if we took the steps to make sure that we're dealing with the grid constraints, dealing with the wider constraints that we've and restrictions that we've spoken about in the planning and environmental context. So I don't know if that answers your question, but I think we're we're doing alright. But the comparison is where could we be, and I think we could be a lot better.

I think it's a really nice reframing and will help people think about data centers in GB in the right way. So to wrap up, is there anything you'd like to plug?

So there's two things to plug. One is the nuclear regulatory task force final report. It's worth saying that the forty seven recommendations, as I mentioned, have very specific time scales. And an organization called the Centre for British Progress has created a tracker so you can track exactly how the government is implementing the recommendations. And then the second thing is my firm has infrastructure planning blog. So if you're interested in reading about development consent orders, infrastructure planning, environmental rules, then you should definitely sign up to it.

Oh, I I definitely gave that a read in advance of this. So, yeah, can definitely vouch for that.

And last question, is there a contrarian view that you hold?

I think this is slightly contrarian, but the idea that the wider benefits of low carbon technology of new tech new kind of hyperscale infrastructure matter a lot more than the individual impacts that they cause. So what I mean by that is, yes, we could spend time looking at whether a particular data center or a particular nuclear reactor could be made perfect through a series of processes.

But really, a world in which that nuclear reactor did not exist is a much worse world. A world in which we don't have the AI infrastructure that we need is is a much worse world. So we should try to develop a system, a way of thinking about things which, again, to use the phrase, puts the outcome over the process. But also doesn't try to make every specific decision that we make a reflection of every single possible desire that we have for the wider world.

It's such it's it's really contrarian because when the the the headline hits, right, of of something going wrong at a data center or something going wrong at a at a nuclear site, the the biggest story of what that nuclear site has done over a long period of time or what the data center is doing for productivity is totally lost. And you don't see it at all. Sort of mind goes back to some of the decisions around nuclear here in in Germany over the last few years and how that's kind of impacted on on coal operation. But we we probably don't have time to cover all of that. So I wanna say a huge thank you for coming on and bringing your expertise in terms of nuclear data centers, but also planning. I think that will give our listeners a huge insight into what's working well and what we need to fix.

Thank you for having me.