AI Sovereignty & Data Centers: Concrete Use Case

Pauline Thomson (00:05)

We are at the Ardian digital and AI conference on AI sovereignty. And so I wanted to start this conversation by asking you what you think about AI sovereignty, and how do you define it?

Gonzague Boutry (00:22)

First, let me congratulate you on this second event, AI x AI. I think it has been a tremendous success. Last year, you had the chance to put on stage a Nobel Prize winner in economics, Philippe Aghion. And clearly, today I get the feeling that we are going even deeper into the discussion. And I think also it shows that Europe is very mature. And getting to that maturity on this topic, the talent builds its own knowledge. I know we will discuss that, but I think you should also be positive and optimist. I'm Gonzague, leading digital infrastructure for the team in Europe. I’ve been in the team for 15 years and we joined almost at the same time. That's the way, we see a relationship at Ardian. Relationship with our client, relationship with the team, relationship with different stakeholders, even in an environment that is moving very fast. So it’s a fascinating world. 

And, well, obviously the topic of digital sovereignty was the key topic today at the debate. We believe that, with AI, the influence this infrastructure has on the superstructure will significantly increase. And that's, in a way, the sovereignty of AI. And that's why we need to better control the critical infrastructure we will build over the next five to ten years. Just one example, the nuclear power plant in France. The airports are usually controlled by a European entity or by European partners as a whole. And let’s say what we need to build for AI infrastructure tomorrow falls within the same category, as critical infrastructures. Here is the second element that I will mention about sovereignty, and to demonstrate that those are critical infrastructures. 

As you know, because we are working together on the development of Verne in France, and we'll come back to that, when you are developing a data center in France, and not only in France, but in Europe, you have, obviously, some fire and safety issues. So you need to get the firemen on-site. But the second, key step that you need to obtain is from the Army. Why is it coming from the Army? Because if there is a war in France, in Europe tomorrow, which hopefully will not get there, the second side, that will be attacked after the nuclear plant, will be the data centers. And that's the guidance by the French Army. So we need to protect the nuclear plants as a priority. But the second one to protect is the data center, because it’s basically the backbone of our digital and broader economy. 

So sovereignty is vital. Now, we need to act, and we need to act now because the race of AI is not one we can afford to lose. We can say that, on the cloud front, we clearly lost the cloud race. 80% of cloud infrastructure is owned by US operators in Europe, and a similar figure applies to the cloud system used by corporations in Europe: 80% are US cloud systems. So, in a way, we lost the battle on the cloud. Now, we are entering a new era, AI, and as you know perfectly well, it's a chance for us to compete in it. But, the wake-up call was six to eight months ago. Now we need to be very focused on the execution. Why am I saying the wake-up call was six to eight months ago? Because we had the Draghi report, we had the AI summit. So, I think European citizens now understand that there is a roadmap. Some may agree with it more or less, but at least there is one. And it's about executing that roadmap day to day. 

And that's also our duty as infrastructure investors to deliver that roadmap. But maybe before that, and let's keep our focus on AI before going to the infrastructure stack, you have been Head of data science for a few years at Ardian. How do you see the evolution of this role? You have built a fascinating partnership with Mistral. Again, huge congrats on that. Can you tell us how data science has evolved within Ardian? Also, can we do more with AI for Ardian as a firm and for portfolio companies?

Pauline Thomson (06:45)

Tough question. The way we have thought about data science at Ardian, as you mentioned, the first step was to apply it to portfolio companies. We were convinced that there was a lot of data for infrastructure assets, as you know, that were not really used to enhance operations. So, we thought that, as an investment fund, we had the benefit of scale automatically and could build tools that would be useful. For example, if you build a tool for one airport operation, there's a good chance that it will also be useful for another airport. So, we could scale the tools and collect more data because we invest in three verticals: transport, energy, and digital infrastructure. 

So that's the way we started, and because it worked well, we applied it to ourselves and I think we share the conviction that, for sure, AI will have a tremendous impact on the private equity industry because our job is mainly to treat information as efficiently as possible, the most information, the most efficient possible. We analyze it and, based on that, make investment decisions. So, when you have these large language models that can help you treat this information ten times faster and expand dramatically the scope of data that you can collect, of course, it's bound to have a huge impact. 

But, as you mentioned, sovereignty is an issue at the level of states and governments. But it's also an issue, or something you need to be aware of at the micro level of each company. And so, of course, we use US cloud services. But you need to know where your dependency lies and where you still need to be in full control over your own data. And you can also discuss with U.S. hyperscalers to get assurances about where your data sits and so on. When it comes to LLMs, we didn’t want to rely on a single model.

So we built an internal application called GAIA that leverages several large language models. And the users can choose the large language models that they prefer. And the first one that we put on the platform was Mistral. Why Mistral? First, because it's one of the best performers, and even more so when we launched a platform a year ago, the performance for a lot of the regular questions that we asked was really on par with US LLMs, which is quite amazing for a French startup. And second, because there is the sovereignty topic and we wanted to have, also, the choice between the US and Europe, you know, in our large language mobile providers. So, we built this partnership with Mistral, and I believe, as we discussed today, that, as European actors, we need to push the research in Europe. We have a research partnership with Mistral to apply LLMs to financial services. But it's sort of creating an ecosystem around European companies and creating more partnerships across markets and across countries. 

And I think that's the way we approach Data Science at Ardian, we try to build partnerships with universities, with startups, with the cloud hyperscalers because they're in Europe, they need to be our partners. We shouldn't forget what's important to us, but they're here. There are many ways of creating sovereignty for us, which doesn't mean isolation.

And I think that's a good segue into Verne since we did this investment together. And I think we are very proud to have invested in a data center platform that has been focused on AI since 2012 and remains so today. In Europe, I think it's quite an amazing investment that we made. Maybe you want to say a word about it. 

Gonzague Boutry (11:28)

Yes. And before that, I think what you built with this tool, GAIA, is fascinating. And it’s true for GAIA, but it's also true for most AI products and the adoption rate. And seeing our team, our analysts, and our investment managers using the platform, I think we can be proud of how adoption can deep dive into a corporation. 

And we had the chance to have Vincent Luciani on stage, who made that reference. We are not late on the usage of AI within corporations in Europe versus the US. Where we are late is the infrastructure. It's about investment. And it's about innovation. So, we need to accelerate that: innovation, investment, and education. In a way, those are the three pillars where we need to accelerate. And again, that's our duty as investors, to support that. 

And that’s the way we do it with Verne because, like in any of our investments in infrastructure, the way we create value, the way we deliver good performance and best of class performance for our investors is to deliver massive CapEx. And we have the Chance to link the two. That’s obviously why we see billions of announcements every day about data centers and AI data centers. So there is a massive growth plan to be delivered. And we want to be part of it; we are part of it. And I think we can be proud of that. 

Just a snapshot on Verne, Verne is a data center platform that used to be in a very niche market, which is high-performance computing. And they had a vision: those data centers will consume more energy than cloud data centers. That's a fact. Because with higher density, you have more energy consumption. So it will consume more energy. Then where do I need to put, install, and build those data centers? And where do you have a lot of energy as well. A place with a lot of decarbonized energy, and a lot of efficient and cost-effective energy and initiatives. Through that, the business was founded in Iceland because it’s 100% geothermal, and then it expanded into the Nordics with Finland and Norway, where we have also secured a site. 

Now, you, me and the board, supported by the management team, and the CEO, Dominic Ward, have decided to open new markets. And why do we want to open a new market? We want to keep the DNA of the company, that vision that I just mentioned. But we want to develop, build, and close the gap versus the US in building those new strategic infrastructures. So, we are opening in France, and I'm very proud to say that France is a key market for Verne going forward. We have a massive ambition in France. We are assessing one site from the famous list of President Macron which was announced as part of the AI summit, featuring 35 sites. So we are looking at one of the sites today. I cannot disclose that site, but we are making good progress. This is a big site. Close to Paris, in the Île-de-France region. We are also looking at assessing Denmark as a new country. 

Similarly, a lot of green energy will come online over the next few years, thanks to the Danish government’s massive ambition to build offshore wind. So those are the market where Verne is operating. And clearly as a shareholders, we want to provide the tool and make the commitment to the management, local stakeholders, national stakeholders, and European stakeholders that we will control this infrastructure and maintain a certain level of control over this strategic infrastructure going forward. 

I'll stop there, and I'll ask you a question. We have been on the board of Verne for 18 months, we have changed, and we have supported the management in changing a lot of things. How do you perceive that a shareholder, like Ardian, can create value and support for the management team on a daily basis?

Pauline Thomson (17:27)

Well, I think we have both witnessed how capital-intensive this market is. So I think, for us as long-term investors, it's great because we want to invest in CapEx. And so, in a way, what we are witnessing is that we struggled to build fast enough, in fact. So, what we can bring is also the construction expertise we've built in other sectors, such as energy. We have Graham, an operating partner, who’s been helping us with CCGTs in the past, specifically on district heating networks and other very capital-intensive assets. And now, clearly, data centers are maybe the most capital-intensive that we see. 

So, that's our role, of course, as shareholders to support the management team’s ambition to reach capacity by investing alongside them or behind them. Second, I think it also helps to have a European shareholder who knows about European specificities and local specificities. We have offices in Finland. We have offices in London. We don't have an office in Iceland, but we spend a lot of time there. And so we bring this local specificity, but we also have, by our presence globally, relationships with big US and Asian players, so that we can bring all these relationships for the benefit of companies. 

I think in this data center market, you need shareholders with industry expertise, deep pockets, international reach and honestly, humility, because I think we both learn every day in this market. It's very technical. It seems very simple. People always think or seem to say that it’s warehouses, data warehouses, basically, but it's actually very complex; complex market dynamics, complex technological dynamics. But it's also fascinating to be in this market for that reason. So yeah, that's what I would say.

Gonzague Boutry (19:58)

And to wrap up on my side, I agree that it's very complex for us, what we are building. Usually, I make the comparison with, building Ferraris in Formula 1. We are building Formula 1. And we also need a track, the track being connectivity. And that's what we are building, not only for us, but for shareholders and for our community going forward. 

To wrap up, I think it's a message of optimism. I think we are entering a new paradigm. We’re entering a new paradigm with a lot of strengths in Europe, strengths to increase capacity, to invest, to deliver, best-in-class infrastructure, also to innovate, and also to form new talent. So I think it's a message of optimism. It's a message about the control of the infrastructure, where Ardian has a key role to play, and I think this key role goes hand in hand with it. 

Also, duties towards national, European, and local stakeholders. But we share some convictions, and we share the same conviction about optimism and a great future, assuming we take that very seriously and we execute now the plan and the vision that have been framed by our leaders.