Introducing Talent Talks, the podcast where we dive into all things talent in the real estate, construction, and infrastructure world. From the boardroom to the building site, we uncover the stories, insights, and people shaping our industry today. We shine a spotlight on data centres, one of real estate's fastest-growing and most in-demand sectors.

In this episode of Talent Talks, hosts Sarah Davenport and Oli Coote sit down with John Booth, Managing Director of Carbon3 IT and a veteran authority on data centre operations and sustainability. The discussion explores a sector at a critical "fork in the road," balancing an ageing workforce and a culture of aggressive talent poaching against the explosive, power-hungry demands of the AI "bubble". John details his vision for the National Data Centre Academy, which aims to replace one-dimensional recruitment with a multi-disciplined, "3D" apprenticeship framework. By addressing the staggering reality of power requirements, projected to jump from 5 kilowatts to potential megawatts per rack, John highlights the urgent need for a new generation of cross-skilled experts capable of navigating the rapid evolution of global digital infrastructure.

Watch this episode on YouTube:

Listen to this episode on:

Spotify

Apple

Amazon

Deezer

You can read the transcript of the discussion below:

Sarah Davenport

Today on Capstone Talent Talks, we are joined by John Booth, Managing Director of Carbon3 IT and a long-standing voice in the data centre industry. John has spent decades working across data centre operations, standards, and sustainability, and is also closely involved with the National Data Centre Academy, helping develop the next generation of talent entering the sector. Today we'll be talking about how the industry is evolving, what that means for skills and talent, and how future careers in data centres are being shaped. John, welcome to Talent Talks. Thank you so much for joining us.

John Booth

My pleasure.

Sarah Davenport

Let's start with you. There's lots to talk about with the work you're doing with the National Data Centre Academy, but let's begin with how you found yourself in the sector.

John Booth

I kind of fell into it, like so many people working in the data centre industry. It's a relatively new industry, only really around 24 years old as a professional discipline. A lot of data centres were stranding capacity after the dot-com boom, and then somebody had the bright idea of using them as backup sites for big clusters. Then 9/11 came along, and there was a real need for the development of highly secure facilities away from banks and financial institutions.

I started my career as a BT apprentice back in 1981. I did 3 years as an apprentice, and a lot of my approach to how the National Data Centre Academy is structured is very much based on that experience. We wanted it to be a more practical, engineering-led training programme than what is currently available elsewhere, though I don't see others as competitors, I see them as partners. We're all there to try and deliver the people we will need for our future digital infrastructure requirements.

I will say, though, that I'm not 100% convinced that the amount of development that has been announced over the last 2 years will actually be built. But there are a lot of transferable skills we can draw on. Data centre engineering is a highly specialised discipline, and we don't have enough people, which has been paramount and really clear for the last 10 years or so. Every event I've attended has raised the question: where are we going to get the staff from? We've muddled through it, but now we've got this huge fork in the road with AI. We will need more design engineers, more installation and construction people, and more operations staff as we move forward, and I think that will happen regardless of whether some of those projected projects don't get built, because we need those people.

Sarah Davenport

Yes, and we've got an ageing workforce in the sector at the moment. So from an attrition perspective, there's certainly going to be a greater need in terms of volume and also in terms of capability and diverse talent coming in.

John Booth

Absolutely. Because we're a relatively new industry, I think we've been quite poor at engineering recruitment and talent-seeking expertise. Operators don't tend to have what I'd call aligned apprenticeship programmes, as we had at BT. The approach has very much been to grab someone from another data centre operator and dangle an extra £20,000 a year in front of them.

That is prevalent throughout the industry, from C-suite all the way down to data centre technicians. I was chatting to someone at DCD London last September, and he said he had technicians literally walking 350 yards up the road in Slough to a competitor and earning £5,000 more. With the cost-of-living crisis, who wouldn't be persuaded by a bit of extra money? But that is a very short-term view. We can't continue to poach from each other because eventually that pool will be diminished. And we're competing in the engineering market with a lot of other sectors, high voltage engineers for the Great British grid upgrade, EV charging, and heat pumps. We’re going to be competing with ourselves and our clients.

I know of at least 3 occasions where people working for a data centre operator have been poached by their clients with an extra £50,000 a year on offer. Who wouldn’t want to do that?

Oli Coote

Do you think the apprenticeship structure you had at BT, compared to what's available now, is part of the difference? I appreciate that some companies are doing good things on the apprenticeship front, but BT was a very different kind of organisation compared to the many smaller, more entrepreneurial businesses in the market today that are still finding their feet.

John Booth

Yes, BT obviously had the power of scale. At the time, I think they recruited something like 1,100 apprentices in the London area alone and had a workforce probably in the millions. But there were apprentices we met in Scotland, Sheffield, Bristol, and other locations too, so there was definitely a broad apprenticeship pool.

You're right, Oli. What we're trying to do at the Academy, in partnership with some of the operators, is create a proper data centre engineering apprenticeship framework. Some operators do recruit apprentices, but they tend to take on maybe 2 a year and link them with their local technical college, and they’re doing standard electrical, mechanical, or ICT apprenticeships. That's a start, but what we want to do is create a data centre engineering apprenticeship where you get the training needed to be a data centre technical expert, cross-skilled across mechanical, electrical, and IT.

One of the EU Code of Conduct best practices, group involvement, talks about having everyone in a data centre, the stakeholders, being able to discuss any aspect of data centre engineering, whether it's a retrofit, an upgrade, or a new client coming in, and understanding the full consequences of each decision. If you've got someone in the room who understands all 3 disciplines, that's only a good thing. Too many times I've been on customer calls at weekends and asked, "Have you got the networks up?" and you can see the penny drop, they'd forgotten the network. Or asking whether they'd done a power assessment to understand the impact on the facility. Having someone multi-disciplined in the room to ask those 3 questions makes an enormous difference.

We've fleshed out what the apprenticeship looks like; it's a 3-year programme with additional modules that can be applied as people move into more senior roles. We've even included a module on sales, because I'm constantly surprised by how technical some operator websites are. They talk about PUE, dual-resilient power supplies, and tuning templates, but the person visiting just wants to buy some space. Sometimes the conversation just needs to be: how much space do you need, what sort of power do you want, and when do you want it?

Oli Coote

Keep it simple.

John Booth

Keep it simple, stupid, which is a principle I've held on to throughout my career.

Sarah Davenport

So, the difference between what you're building and what some operators are doing from an apprenticeship perspective is that you're trying to build a 3D approach, rather than 1-dimensional routes.

Talk us through how the apprentices find out about you, how they are, because it's an increasingly popular route, which I'm very excited to see, because it's moving away from the traditional degree route, which is where everyone was kind of shoved in, whether you wanted to or not, me included.

Whereas this is obviously suitable for a much broader section of society where you can earn and learn, and that's really appealing for many.

So how does that work in practice?

John Booth

Nothing has been published yet because we're still liaising with government, operators, and other partners about the structure, and we are in the early stages. The government knows that the current apprenticeship scheme is flawed and has been trying to address it; they've done some policy updates and stripped out some existing courses. We're trying to slot something new in at the same time as this rationalisation exercise is happening.

We've also been talking to the British Computer Society, who are doing some work with computing in schools, and we've suggested that a digital infrastructure module needs to be part of it. We held a big data centre skills summit in March last year and gathered a lot of information. We know what the framework looks like, but pulling all of the strings together from a national perspective is complex.

You can't restrict it to 1 college; it has to be available nationwide, which creates challenges when you look at sectoral and regional factors. We know the AI growth zones are currently planned for 5 locations: 1 in the North East covering Newcastle, 1 in Scotland, 1 in North Wales, 1 in South Wales, and 1 at Culham. The local colleges supporting engineering in those regions need to be able to deliver these courses. But right now, nobody has actually committed to building anything in most of those locations. It would be premature to ask a local technical college to invest in the programme when there aren't jobs available for students to go into afterwards. It's going to be quite premature to get a local Technical College to look at the program and to publish it when there are no jobs for them to go to afterwards.

The only current initiative is the Universal Technical College in Slough, partly funded by some local operators. It covers the Thames Valley Corridor up to Oxford, and it has a 14-to-18 track, which includes data centre-related content. But it's not an apprenticeship, it's more of an enhanced school track, similar to a sports or engineering specialism. They've actually come to us and said there's nothing for their students to move into after that. They won't get directly employed by an operator because they don't yet have the knowledge. That's the missing piece, a proper data centre engineering apprenticeship, or potentially a degree.

And it's hard going sometimes. You speak to 1 operator, and they say they like it, but they need changes. You speak to another, and they say the same thing. But guys, it's a data centre. All of you use the same types of technologies in broadly the same way. We're trying to help by creating a pool of people; don't put unnecessary restrictions on it. Some operators in the US have made the same point: look at the job descriptions on data centre operator websites and ask yourself how many people actually have all those skills. Very few. And usually they're really just looking for someone working at the data centre next door whom they can poach with a £20,000-to-£30,000 carrot.

If you continue down that route, you'll stay in exactly the same situation you're in now, struggling for staff. So, I'd advise any data centre operator to re-engage with us, look at what we've developed, and see whether it fits their needs. We're quite happy to license the course for other colleges to use once it's developed, but many colleges want to see a high student count before committing, and that's not guaranteed. It really depends on the location. In Newcastle, for example, with the Blyth project under construction, by the time apprentices have finished their programme, there could be around 600 jobs available. In Scotland, it might be more like 7 or 8 years away because they haven't even broken ground yet.

The other thing we need to be is nimble, because the technology being deployed is changing rapidly. Gone are the days of 5 kilowatts per rack. At a recent tech event, a colleague from NVIDIA mentioned they're looking at 1.5 to 2 megawatts per rack. That's a huge, fundamental change in design physics. But the course has to point in that direction, give apprentices a good grounding in the basics of air cooling, liquid cooling, and immersed compute technologies, so that in 5 years' time, if we've all moved to immersed compute, they'll know exactly what health and safety protocols are needed and how to deploy at scale.

Oli Coote

Presumably from a skills shortage perspective, it's not one-size-fits-all either. You could have someone far more junior taking the full apprenticeship route, going through that data centre engineering programme, but then you could also have people coming from a related discipline who just need upskilling to make an impact in data centres.

John Booth

You have an interesting point. The framework is designed to accommodate that. We're looking at taking one year's worth of the commercial element of the electrical apprenticeship, for example, because if you do a standard electrical course, around 90% of it is domestic, which is of no use to a data centre engineer whatsoever. What he really needs to be focused on is high voltage awareness, low voltage awareness, and then qualifications like the AP (Approved Person) or Approved Engineer designation, especially when doing mains failure tests or major facility upgrades.

Our vision is that the apprenticeship is a series of modules, and when you complete a certain number, you receive the apprenticeship certificate, but nothing stops you from taking more modules to become a kind of super-qualified engineer across multiple disciplines. And if we work with a university, there's potential to create a pathway to an actual degree, a postgraduate certificate in one discipline that's a recognised qualification, with 3 or 4 of those adding up to a full degree. Do more, and you might reach master's level; do more still, and you reach PhD. So you can have people who are embedded, heart and soul, in data centre engineering from its origins right through to where it might be in 10 to 20 years' time. That's the vision; whether we can get there is a lot of work, but we've been building towards it for some time and have tacit agreement from most of the people involved. I think they're waiting to see what the output actually looks like.

Sarah Davenport

Do you think part of the challenge is that there's so much noise in this space? I love the data centre industry, let's make that clear, but I really struggle when it's all chat and no action. There's so much chat around talent, new sites, new joint ventures, and so on, that ultimately many operators, developers, and others are probably thinking: I don't know how to sift through the noise to figure out what the real capacity need will be.

Because ultimately, if I'm only going to need 5 ongoing engineers, then that's going to impact whatever it might be, depending on the site or whether it's retrofit or whatever.

But actually, it’s then filtering out the noise and going, what’s actually going to land here? Because what's talked about and what actually gets built are two very different things.

John Booth

Yes, it's a tricky one. I'm not going to be controversial; I'll simply say that's what happens in a bubble environment. And I do think the data centre industry has been in a bit of a bubble for about 3 years now. A lot of talk, not a lot of action, and this has largely been given by AI companies. Pre-AI, most projected announcements were actually customer-driven, customers saying they wanted extra cloud infrastructure, that type of data centre to be located in certain areas to give them the availability zones they required.

And that was also driven by those companies wanting to expand their actual footprint into places where they didn't have a footprint before. And that footprint could have been, you know, largely it could be Africa, it could be Latin America, where traditionally there's not been that much data centre activity.

The data centre industry is literally driven by the states. Europe's kind of saturated part of APAC is saturated. But there are still certain kinds of developing countries that don't have the necessary infrastructure. So you can absolutely understand why they want to develop data centres.

What we've got at the moment is that we're looking at AI as a thing, and we're being told by the AI companies that if we don't deploy AI, they will be left behind. And I’m thinking, left behind from what, exactly? If you look outside the data centre press and the event rhetoric, something like The Economist, and look at the underlying fundamentals. They were saying that for every dollar deployed in AI data centres to date, the revenues have been around 3 cents on the dollar. That doesn't seem like good business sense. I also read a report saying that for AI to recoup what has already been spent, it would need to generate cloud-level revenues twice over. That is a huge ask.

And then you look at these big hulking data centres, 100 MW-plus, with the latest talk of an 11 GW campus, and you just think: let me know when you get the power. The second question is whether anyone has fully considered the impact of all those projects on the global supply chain in terms of capital plant items, generators, transformers, chillers, CRAC units, leak detection systems, DCIM, and think about the impact on the employers. How many people do you need to A: design, B: build, and C: operate those facilities if it comes off right?

Back in 2023, the Uptime Institute published a statement estimating a need for around 2.5 million people globally working in the data centre sphere of influence. That figure didn't include the 50% projected to retire by 2030. So, there's a potential shortfall of around 1.5 million people, and that was before AI drove a quadrupling of projected global capacity. The industry may have to do one of two things: automate far beyond its current imagination, proper logistics-style automation like you see in distribution warehouses, or recruit a lot more people.

I suspect it's going to be the latter. But I also suspect that right now, much of the industry thinks it's somebody else's problem. Building big data centres and think it’s alright, I’ll just sub-contract facilities managers, and I’ll try to keep my team tight. All of the senior people within an organisation may be trying to stick to say 40/50 people that are directly employed by that operator and then sub-contract everything else.

That, on the face of it, looks like great business sense, and I’m sure for all of the facilities management providers, all it means is that our impact, our really critical staffing problem, falls down a level.

Sarah Davenport

That's exactly what I was going to ask about, where is the demand actually sitting? Is it within in-house teams, or further down the supply chain, and how is that balance changing project scale and project risk?

John Booth

Most data centre operators will have their own small design team, but those teams are largely working from cookie-cutter designs, something that's been deployed elsewhere and worked, which you can simply pick up and drop into a new location. The design work is minimal because you're essentially cutting and pasting.

What's changed is that the latest AI chipsets are demanding a lot more power, which means big redesign of the power side of things. Physics will tell you that you can only deploy so much voltage to a location at a point, and we know what the theoretical limit is. Most of the current electrical supply chain offering runs to around 120 kilowatts per rack. And even that creates problems: taller ceiling slab-to-slab because you've got more buzz bars in place, which affects the height of the fan walls, and what you’re deploying down there and everything else. Like, how do you reconcile fire suppression systems? Do you need enhanced sensors? You have a lot more power in a smaller space, how do you do that?

Sarah Davenport

And heavier equipment.

John Booth

And heavier equipment, which can be a good or a bad thing. The next generation of that particular brand is going to be 4 times as powerful. I had a supplier at the Academy for an event last year and asked him the question: who was going to deliver the 630 kilowatts per rack that you’ve just mentioned? He said, "Oh, the market will do that." I said to him, “What happens if it can't?” and he just said “Moving swiftly on.”

And then I saw one of his colleagues at a TechX event, saying that we’re going to do 1.5 to 2 megawatts per rack. I said, “No, we're absolutely not going to do that. Not in the time scale you want, and not with considering all of the health and safety aspects of delivery and that much power in that rack.” They were talking about moving to DC, direct current, and there's a very good reason we don't use DC in data centres: because it kills people. At that level, you bring in a whole new tier of health and safety scrutiny.

I've been calling for a radical redesign of data centre engineering for some time. The data centres we're building today have their roots in facilities from 30, 40, or even 50 years ago. I recently visited one built in 1977, and I was quite surprised, it used free cooling, it had water towers, which was interesting as water towers are a relatively new thing. But then, a couple of weeks later, I visited a brand new data centre and walked down the corridor thinking: nothing has changed. We're still delivering the same types of infrastructure we were 50 years ago. We haven't innovated.

In some of the presentations I’ve done for other webinars, I talk about the 1977 Ford Fiesta, obviously built in the same year as this data centre. Compare it to a 1923 Fiesta, and it's a completely different car, more streamlined, with a more fuel-efficient engine, far better safety features inside the cabin, and better in-car entertainment. Pound-for-pound its probably a little more expensive, but we haven't seen that level of product innovation in the data centre industry ever.

Sarah Davenport

Interestingly, they've now discontinued the Fiesta entirely and replaced it with completely different models. Which shows that if something's working, you keep evolving it, and when it no longer works, you pause it.

John Booth

I think we are there now, actually. If you look at where we started in the early 2000s, 0.5 kilowatts per rack, and most of the customers I speak to now are looking at 15 to 20, with some up to 50. And yet the latest GB300 chip has a 150-kilowatt cap. When I ask clients when they're planning for 150 kilowatts per rack, they often say  “we’re not going to future-proof”, and they'll wait for a customer inquiry, and then we’ll retrofit and build them out to 150. I understand the reasons for doing it, but in 2 years, it'll be 630 kilowatts per rack, and that has a massive impact on your main substation and how much power you procure from your local distribution network operator, which is typically capped at around 20 megawatts. If you want to deploy serious AI factories, you'll need substantially more, and that can only come from massive increases in grid capacity.

The National Grid in the UK has a significant high-voltage engineer shortfall, and the Great British Grid Upgrade project, 17 separate projects, is going to absorb all available high-voltage engineering personnel for the next 20 years, essentially. And at the same time, we've got this projected quadrupling of data centre capacity, much of which will also need high-voltage engineers. Where are they coming from?

There's another factor that could be somewhat tempered by current Middle Eastern events, but Saudi Arabia, the UAE, and other countries in the region have been talking about building AI facilities powered by their own fossil fuel resources. That's great. Now we’ve got a major conflict that looks like it’s in development. But what happens if they can’t?

If they can't find enough high-voltage, construction, or operations personnel in their countries, they'll come to the UK and Europe and offer packages at double the salary with all the associated benefits. If you're a young person without heavy commitments, going on a five-year project like that could genuinely set you up for life. We could lose a meaningful chunk of our existing workforce to overseas opportunities, and it's already happening in America and APAC, where there is genuine demand for European-trained data centre professionals. Retaining that talent just adds to an ever-rising cost spiral for the industry.

Sarah Davenport

We certainly see that. We work across Europe and the US, and relocating talent across geographies is a significant part of what we do, particularly across the construction cycle to the US from the UK. And, as you say, if you're young and haven't got loads of commitments holding you back, it's a great opportunity.

John Booth

If I were 25 again, with the knowledge I've accumulated today, I'd be on the first plane.

Sarah Davenport

Same, absolutely. John, we could talk for another 3 hours, but I'm conscious of your time. Thank you so much for sharing all of your insight and your vision, which is really clear, and you are spread across a huge intersection of the DC ecosystem, which can only add value to everybody, even if it perhaps isn't great for your stress levels.

John Booth

I'm completely stress-free, to be honest, because it's not my problem!

Sarah Davenport

Well, there you go! Are you going to be at DCW?

John Booth

Yes, I'm presenting. I'm doing an event looking at whether data centres are still worth the investment and under what conditions, which is going to be interesting. Then on Thursday, I'll be moderating a panel on power and taking part in the DCA's net-zero debate with my colleague, Nick Morris. It's going to be run like a proper debate, with Vanessa Moffett from the Data Centre Alliance as chair, and the opposing team being Professor John Summers from RISE and Professor Ian Bitterley. It's going to be a very lively conversation.

Sarah Davenport

I look forward to that, and to meeting you in person, John. Thank you again so much for your time.

John Booth

Fantastic. Thank you.

That was an insightful conversation with John Booth, Managing Director of Carbon3 IT and a leading authority on data centre sustainability and operations. John provided a candid look at the widening skills gap and the "bubble environment" currently being driven by the AI gold rush. John challenged the industry to move past the short-term habit of poaching talent from competitors, highlighting how this "carrot-dangling" approach creates a diminished pool for everyone. From the complexities of launching the National Data Centre Academy to the staggering reality of power demands, where racks are jumping from 5 kilowatts to potential megawatts, his insights were both a reality check and a roadmap for the future. Most importantly, he reminded us of the urgent need for a multi-disciplined, "3D" approach to training the next generation, ensuring our digital infrastructure is supported by experts who understand the full consequences of every engineering decision. We are extremely grateful to John Booth for joining Sarah Davenport and Oli Coote on Capstone’s Talent Talks.