How Much Investment Has Europe Announced for AI Infrastructure?

Live from Morgan Stanley’s European Tech, Media and Telecom Conference in Barcelona, our roundtable of analysts discusses tech disruptions and datacenter growth, and how Europe factors in.

Read more insights from Morgan Stanley.

----- Transcript -----

Paul Walsh: Welcome to Thoughts on the Market. I'm Paul Walsh, Morgan Stanley's European Head of Research Product. 

Today we return to my conversation with Adam Wood. Head of European Technology and Payments, Emmet Kelly, Head of European Telco and Data Centers, and Lee Simpson, Head of European Technology. 

We were live on stage at Morgan Stanley's 25th TMT Europe conference. We had so much to discuss around the themes of AI enablers, semiconductors, and telcos. So, we are back with a concluding episode on tech disruption and data center investments. 

It's Thursday the 13th of November at 8am in Barcelona. 

After speaking with the panel about the U.S. being overweight AI enablers, and the pockets of opportunity in Europe, I wanted to ask them about AI disruption, which has been a key theme here in Europe. I started by asking Adam how he was thinking about this theme. 

Adam Wood: It’s fascinating to see this year how we've gone in most of those sectors to how positive can GenAI be for these companies? How well are they going to monetize the opportunities? How much are they going to take advantage internally to take their own margins up? To flipping in the second half of the year, mainly to, how disruptive are they going to be? And how on earth are they going to fend off these challenges? 

Paul Walsh: And I think that speaks to the extent to which, as a theme, this has really, you know, built momentum. 

Adam Wood: Absolutely. And I mean, look, I think the first point, you know, that you made is absolutely correct – that it's very difficult to disprove this. It's going to take time for that to happen. It's impossible to do in the short term. I think the other issue is that what we've seen is – if we look at the revenues of some of the companies, you know,  and huge investments going in there. 

And investors can clearly see the benefit of GenAI.  And so investors are right to ask the question, well, where's the revenue for these businesses? 

You know, where are we seeing it in info services or in IT services, or in enterprise software. And the reality is today, you know, we're not seeing it. And it's hard for analysts to point to evidence that – well, no, here's the revenue base, here's the benefit that's coming through. And so, investors naturally flip to, well, if there's no benefit, then surely, we should focus on the risk. 

So, I think we totally understand, you know, why people are focused on the negative side of things today. I think there are differences between the sub-sectors. I mean, I think if we look, you know, at IT services, first of all, from an investor point of view, I think that's been pretty well placed in the losers’ buckets and people are most concerned about that sub-sector… 

Paul Walsh: Something you and the global team have written a lot about. 

Adam Wood: Yeah, we've written about, you know, the risk of disruption in that space, the need for those companies to invest, and then the challenges they face. But I mean, if we just keep it very, very simplistic. If Gen AI is a technology that, you know, displaces labor to any extent – companies that have played labor arbitrage and provide labor for the last 20 - 25 years, you know, they're going to have to make changes to their business model. 

So, I think that's understandable. And they're going to have to demonstrate how they can change and invest and produce a business model that addresses those concerns. I'd probably put info services in the middle. But the challenge in that space is you have real identifiable companies that have emerged, that have a revenue base and that are challenging a subset of the products of those businesses. So again, it's perfectly understandable that investors would worry.  In that context, it's not a potential threat on the horizon. It's a real threat that exists today against certainly their businesses. 

I think software is probably the most interesting. I'd put it in the kind of final bucket where I actually believe… Well, I think first of all, we certainly wouldn't take the view that there's  no risk of disruption and things aren't going to change. Clearly that is going to be the case. 

I think what we'd want to do though is we'd want to continue to use frameworks that we've used historically to think about how software companies differentiate themselves, what the barriers to entry are. We don't think we need to throw all of those things away just because we have GenAI, this new set of capabilities. And I think investors will come back most easily to that space.

Paul Walsh: Emett, you talked a little bit there before about the fact that you haven't seen a huge amount of progress or additional insight from the telco space around AI; how AI is diffusing across the space. Do you get any discussions around disruption as it relates to telco space? 

Emmet Kelly: Very, very little. I think the biggest threat that telcos do see is – it is from the hyperscalers. So, if I look at and separate the B2C market out from the B2B, the telcos are still extremely dominant in the B2C space, clearly. But on the B2B space, the hyperscalers have come in on the cloud side, and if you look at their market share, they're very, very dominant in cloud – certainly from a wholesale perspective. 

So, if you look at the cloud market shares of the big three hyperscalers in Europe, this number is courtesy of my colleague George Webb. He said it's roughly 85 percent; that's how much they have of the cloud space today. The telcos, what they're doing is they're actually reselling the hyperscale service under the telco brand name. 

But we don't see much really in terms of the pure kind of AI disruption, but there are concerns definitely within the telco space that the hyperscalers might try and move from the B2B space into the B2C space at some stage. And whether it's through virtual networks, cloudified networks, to try and get into the B2C space that way. 

Paul Walsh: Understood. And Lee maybe less about disruption, but certainly adoption, some insights from your side around adoption across the tech hardware space? 

Lee Simpson: Sure. I think, you know, it's always seen that are enabling the AI move, but, but there is adoption inside semis companies as well, and I think I'd point to design flow. So, if you look at the design guys,  they're embracing the agentic system thing really quickly and they're putting forward this capability of an agent engineer, so like a digital engineer. And it – I guess we've got to get this right. It is going to enable a faster time to market for the design flow on a chip. 

So, if you have that design flow time, that time to market. So, you're creating double the value there for the client. Do you share that 50-50 with them? So, the challenge is going to be exactly as Adam was saying, how do you monetize this stuff? So, this is kind of the struggle that we're seeing in adoption. 

Paul Walsh:   And Emmett, let's move to you on data centers. I mean, there are just some incredible numbers that we've seen emerging, as it relates to the hyperscaler investment that we're seeing in building out the infrastructure. I know data centers is something that you have focused tremendously on in your research, bringing our global perspectives together. Obviously, Europe sits within that. And there is a market here in Europe that might be more challenged. But I'm interested to understand how you're thinking about framing the whole data center story? Implications for Europe. Do European companies feed off some of that U.S. hyperscaler CapEx? How should we be thinking about that through the European lens? 

Emmet Kelly: Yeah, absolutely. So, big question, Paul. What… 

Paul Walsh: We've got a few minutes! 

Emmet Kelly: We've got a few minutes. What I would say is there was a great paper that came out from Harvard just two weeks ago, and they were looking at the scale of data center investments in the United States. And clearly the U.S. economy is ticking along very, very nicely at the moment. But this Harvard paper concluded that if you take out data center investments, U.S. economic growth today is actually zero. 

Paul Walsh: Wow. 

Emmet Kelly: That is how big the data center investments are.  And what we've said in our research very clearly is if you want to build a megawatt of data center capacity that's going to cost you roughly $35 million today. 

Let's put that number out there. 35 million. Roughly, I'd say 25… Well, 20 to 25 million of that goes into the  chips. But what's really interesting is the other remaining $10 million per megawatt, and I like to call that the picks and shovels of data centers; and I'm very convinced there is no bubble in that area whatsoever.

So, what's in that area? Firstly, the first building block of a data center is finding a powered land bank. And this is a big thing that private equity is doing at the moment. So, find some real estate that's close to a mass population that's got a good fiber connection. Probably needs a little bit of water, but most importantly needs some power. 

And the demand for that is still infinite at the moment. Then beyond that, you've got the construction angle and there's a very big shortage of labor today to build the shells of these data centers. Then the third layer is the likes of capital goods,  and there are serious supply bottlenecks there as well.

And I could go on and on, but roughly that first $10 million, there's no bubble there. I'm very, very sure of that. 

Paul Walsh: And we conducted some extensive survey work recently as part of your analysis into the global data center market. You've sort of touched on a few of the gating factors that the industry has to contend with. That survey work was done on the operators and the supply chain, as it relates to data center build out. 

What were the key conclusions from that? 

Emmet Kelly: Well, the key conclusion was there is a shortage of power for these data centers, and… 

Paul Walsh: Which I think… Which is a sort of known-known, to some extent. 

Emmet Kelly: it is a known-known, but it's not just about the availability of power, it's the availability of green power. And it's also the price of power is a very big factor as well because energy is roughly 40 to 45 percent of the operating cost of running a data center. So, it's very, very important. And of course, that's another area where Europe doesn't screen very well.

I was looking at statistics just last week on the countries that have got the highest power prices in the world. And unsurprisingly, it came out as UK, Ireland, Germany, and that's three of our big five data center markets. But when I looked at our data center stats at the beginning of the year, to put a bit of context into where we are…

Paul Walsh: In Europe… 

Emmet Kelly: In Europe versus the rest. So, at the end of [20]24, the U.S. data center market had 35 gigawatts of data center capacity. But that grew last year at a clip of 30 percent. China had a data center bank of roughly 22 gigawatts, but that had grown at a rate of just 10 percent. And that was because of the chip issue. And then Europe has capacity, or had capacity at the end of last year, roughly 7 to 8 gigawatts, and that had grown at a rate of 10 percent. 

Now, the reason for that is because the three big data center markets in Europe are called FLAP-D. So, it's Frankfurt, London, Amsterdam, Paris, and Dublin. We had to put an acronym on it. So, Flap-D. Good news. I'm sitting with the tech guys. They've got even more acronyms than I do, in their sector, so well done them. 

Lee Simpson: Nothing beats FLAP-D. 

Paul Walsh: Yes. 

Emmet Kelly: It’s quite an achievement. But what is interesting is three of the big five markets in Europe are constrained. So, Frankfurt, post the Ukraine conflict. Ireland, because in Ireland, an incredible statistic is data centers are using 25 percent of the Irish power grid. Compared to a global average of 3 percent.

Now I'm from Dublin, and data centers are running into conflict with industry, with housing estates. Data centers are using 45 percent of the Dublin grid, 45. So, there's a moratorium in building data centers there. And then Amsterdam has the classic semi moratorium space because it's a small country with a very high population. 

So, three of our five markets are constrained in Europe. What is interesting is it started with the former Prime Minister Rishi Sunak. The UK has made great strides at attracting data center money and AI capital into the UK and the current Prime Minister continues to do that. So, the UK has definitely gone; moved from the middle lane into the fast lane. And then Macron in France. He hosted an AI summit back in February and he attracted over a 100 billion euros of AI and data center commitments. 

Paul Walsh: And I think if we added up, as per the research that we published a few months ago, Europe's announced over 350 billion euros, in proposed investments around AI. 

Emmet Kelly: Yeah, absolutely. It's a good stat. Now where people can get a little bit cynical is they can say a couple of things. Firstly, it's now over a year since the Mario Draghi report came out. And what's changed since? Absolutely nothing, unfortunately. And secondly, when I look at powering AI, I like to compare Europe to what's happening in the United States. I mean, the U.S. is giving access to nuclear power to AI. It started with the three Mile Island… 

Paul Walsh: Yeah. The nuclear renaissance is… 

Emmet Kelly: Nuclear Renaissance is absolutely huge. Now, what's underappreciated is actually Europe has got a massive nuclear power bank. It's right up there. But unfortunately, we're decommissioning some of our nuclear power around Europe, so we're going the wrong way from that perspective. Whereas President Trump is opening up the nuclear power to AI tech companies and data centers. 

Then over in the States we also have gas and turbines. That's a very, very big growth area and we're not quite on top of that here in Europe. So, looking at this year, I have a feeling that the Americans will probably increase their data center capacity somewhere between – it's incredible – somewhere between 35 and 50 percent. And I think in Europe we're probably looking at something like 10 percent again. 

Paul Walsh: Okay. Understood. 

Emmet Kelly: So, we're growing in Europe, but we're way, way behind as a starting point. And it feels like the others are pulling away. The other big change I'd highlight is the Chinese are really going to accelerate their data center growth this year as well. They've got their act together and you'll see them heading probably towards 30 gigs of capacity by the end of next year. 

Paul Walsh: Alright, we're out of time. The TMT Edge is alive and kicking in Europe. I want to thank Emmett, Lee and Adam for their time and I just want to wish everybody a great day today. Thank you.

(Applause) 

That was my conversation with Adam, Emmett and Lee. Many thanks again to them. Many thanks again to them for telling us about the latest in their areas of research and to the live audience for hearing us out. And a thanks to you as well for listening. 

Let us know what you think about this and other episodes by living us a review wherever you get your podcasts. And if you enjoy listening to Thoughts on the Market, please tell a friend or colleague about the podcast today.