- 32 minutes ago
The AI race is increasingly becoming an infrastructure race. Beyond models and applications, competitiveness now depends on access to compute, energy, cloud capacity, and secure systems capable of supporting AI at scale. Energy is no longer just a constraint — it is a structuring force shaping where and how AI can be deployed, and at what scale. For Europe, this raises a deeper question: can it build the backbone required for sovereign AI? As cloud providers, energy companies, and cybersecurity leaders become part of the same strategic stack, new forms of coordination—and new dependencies—are emerging. What would a truly European AI infrastructure look like? And how do you build systems that are not only powerful, but trusted and resilient by design?
Category
🤖
TechTranscript
00:00Please be ready to welcome Oktav Klaba from OVH Cloud, Catherine McGregor from Engie,
00:09and Andrei Kudelsky from Kudelsky Group.
00:12And moderating this session is Tom McKenzie from The Opening Trade and Bloomberg Tech Europe.
00:18Please welcome them all on stage.
00:41Okay, Oktav, Catherine, Andrei, thank you very much indeed.
00:47Let's start, Catherine, I think, with you.
00:50Frame for us, from your perspective, what the energy needs are when it comes to AI right now.
00:58Where are we in this moment?
01:00Where are the gaps?
01:02Frame it for us in terms of that energy impulse.
01:05Yeah, so good morning, everybody.
01:08Very happy to be here.
01:10So, indeed, data center is, you can't separate data center from energy need,
01:18and we are obviously starting to see that in the reality.
01:23Last year, the power demand increase from data center was around 17% globally.
01:2917, 17, 17.
01:3117, 17, which is to be compared to an increase in power demand of 3%.
01:35So, you have an outgrowth of power demand from data center compared to your average power demand globally.
01:43And that is expected.
01:54And that is expected, thank you very much.
02:13So, you have an outgrowth of power demand from data center, which are around energy, which are around the speed
02:18under which the data center, CODs, or the start of the functioning of the data center.
02:26So, you have speed.
02:28So, you have, indeed, energy.
02:30You need connection points.
02:32And then you have the whole theme of acceptability.
02:35And acceptability catches a lot of things.
02:38It catches, indeed, does this data center going to have an impact on power price?
02:43Is it going to consume too much water?
02:47And how about emissions?
02:49Is climate situation going to be worse from data center?
02:52So, there is a lot of challenges that are converging on the data center development challenge or opportunities and challenge.
03:02And here, the energy partners are, indeed, a perfect combination for the data center development.
03:09So, what we're trying to do is to shorten the time to development.
03:14Data center development, today, I expected to bring their infrastructure online in two or three years.
03:20It's very, very challenging if you start with brand new greenfield energy projects.
03:24The most competitive thing you can find in terms of energy projects is probably solar and wind in certain countries.
03:32And it takes you three, four years when you are really, really good in terms of development.
03:36So, what we're trying to do is work together with the development, with the data center development, bring them together,
03:43shorten this time by either using existing infrastructure that already have, for example, connection to the network.
03:51Or develop projects that are under development, and typically, these would be renewable projects that are very fast to develop,
03:59but are already, for example, positioned in the queue because there is a queuing system to get to connections to
04:06help accelerate, of course, the data center development.
04:09And renewables is very important because it brings additional capacity to the energy system, and therefore, it alleviates this whole
04:19story about data center.
04:20It's going to make my power very, very expensive.
04:23I was visiting that booth earlier, and I stole the term unicorn-powered data center.
04:31The unicorn power is powered from, there is not one magic asset that can power the data center.
04:37You have to bring the combination of assets and solutions, and this is typically what an energy company like Engie
04:45can do because we have existing assets, assets under development, portfolio, capacity to source in the energy market, and we're
04:54working with our partners to bring this solution together so that they can have, you know, this.
04:59To be clear, because you brought this up, are data centers in Europe currently being built, are they pushing up
05:04prices for consumers and households and businesses?
05:07I don't think that generally we have seen that in Europe, but, you know, Engie is very present in the
05:13U.S., and so we obviously are confronted to making sure that, you know, power prices in certain markets in
05:19the U.S. have increased.
05:20And so we're very, very closely working with the data center development, specifically in the United States, to make sure
05:27that we can bring online new power solutions to avoid having power price increase, because that's, frankly, a big showstopper.
05:36Because we're seeing a pushback in the U.S. right now, I think about $150 billion worth of data centers
05:40planned build in the U.S. being stalled because of local opposition.
05:44So it's becoming a real point of contention, and part of the framing in the U.S. is that this
05:49is pushing up our local energy prices, and it's leading to opposition at a local level.
05:52This is why additionality, decarbonation are very important themes that need to be integrated in the data center, early involvement,
06:03working with communities.
06:04It sounds familiar to us, because that's what we are confronted when we do renewable project development.
06:10It's all about early engagement, community, making sure that we decide with them where we are going to put this
06:19infrastructure.
06:19These are all the things that we do days in, days out when we develop new renewable projects.
06:24Otherwise, they are stopped.
06:25And I believe data centers are confronted to a very similar project, which is why, you know, we have a
06:30team dedicated now to work very closely with them to make sure that some of these practices also, you know,
06:36are translated into the data center, where you cannot just arrive in a town and say, okay, we're going to
06:41put our data center here.
06:42It doesn't work that way.
06:46Octave, you manage and run data centers.
06:50Do you align with what Catherine is saying?
06:53Are you seeing that?
06:53Is it a constraint for you right now?
06:54How much of a constraint is energy?
06:56You do not have to say yes.
06:58Actually, yes.
07:01Very diplomatic.
07:02So, you know, we've been building the data center for 20 years more.
07:07So today we have 250 megawatts of the energy in our data center in Europe, in US, Canada, and also
07:15in Asia.
07:16And we started doing that in this moment that's 10 years ago, 15 years ago, where it was easier to
07:23find the places.
07:23It was easier to accept them, and it was cheaper, much more cheaper than it is today.
07:29But what we see today, because we're still building and we're extending our capacities, that in Europe, you have the
07:37delay.
07:38You've been talking about the timeline.
07:40So in France, it's okay, because there was really push on building the data center in other capacities.
07:49But still, even if you talk about the very big projects, you're talking about two or three years, just on
07:55the energy to bring the right lines and to build all these.
07:59Two or three years.
08:00Yeah, yeah, two or three years, because, you know, you can bring the 20 megawatts, you can bring the 40
08:06megawatts of the energy very quickly.
08:08But when you talk about 100, 200, 300, you have to change really power, the lines, the kind of the
08:14lines, the transformators, that it takes six months, nine months, 11, even two years to build and to deploy.
08:21And so this kind of things in France, still, it works not bad in Germany, it's really, really hard, okay,
08:29it's really, really complex.
08:30We have data center, but to extend the power, the only answer is, we don't know, we don't, we cannot
08:38answer you when and how many in the, so.
08:42And of course, Germany does not have nuclear power anymore.
08:45So those different countries in Europe that they are struggling already with the with the capacities.
08:51We are only also in Canada, in Canada, it's really a specific country that they manage the power in a
08:58specific way with the dams.
08:59So they have the lot of dams, and they produce the very green energy.
09:04So we are very close to a dam, very big dam.
09:08And we have a very big place where we have 100 megawatts.
09:13And it's, it's, it's not complex to upgrade, but they are struggling in the next two, three years,
09:19because there was so many demands that the, what it's not used, it's just decreased progressively over the month.
09:28And then if you don't take that now, if you don't commit, if you don't pay, you will not get
09:33that.
09:34The U.S. is more complex.
09:36It's really depends on the different regions.
09:38So, yes, we are struggling with the energy and the different countries.
09:43This is why France is trying to take this issue as the opportunity to say, hey, you want Europe?
09:52Go to France, because we have nuclear energy, because it's easier to deploy, because it's cheaper.
10:01We have the very stable prices.
10:02We want to be in this nuclear energy.
10:06So let's do that in Europe.
10:08But on another side, you probably will ask about the sovereignty.
10:14In Europe, we have 27 countries, and the sovereignty means that you have to deploy the data center in every
10:22country, three of them,
10:23because it's a regional free data center.
10:25So, in fact, for the cloud provider like OVH, it means that in a few years, we'll have more than
10:32100 data centers just in Europe.
10:35It will not be the very big data center, but it will be a smaller one, because every country wants
10:40to have the sovereign cloud.
10:41Or in totally full control.
10:44So that means that you have to go to every country and to try to figure out how you will
10:49build 5 megawatts, 20 megawatts, three times data center,
10:54because you will not go on the 100 data centers, every one 100 megawatts.
10:59Yeah.
10:59That doesn't sound very efficient.
11:01No, but it's Europe.
11:03You know, this is when we have our culture.
11:08We have our history.
11:10And we tried to build this European community, but we're still in this kind of fight.
11:20We are not killing each other anymore, but we are still trying to build up and trying to compete together.
11:28You don't have this issue in the U.S.
11:29You put very big data centers on the East Coast, West Coast, and you match everybody in the U.S.
11:37André, you feel strongly about the need to act quickly, with speed, with efficiency.
11:43Weigh in on this conversation, because I know that's top of mind for you, is the gap between Europe and
11:50the U.S.
11:50when it comes to actually acting and acting quickly and efficiently.
11:54First element, it's always important to compare to the best and not just to the neighbor.
12:01And one of the big issues in many people think with the old borders, to think, oh, I have to
12:09do better than my neighbor, rather than to think that we are in an interconnected world and we should compare
12:16to the best.
12:17Now, for AI and data center, one of the key elements is speed on one side and scalability.
12:25And fundamentally, each element where you try to fragment the way you bring a response will make you much less
12:36competitive.
12:37And fundamentally, or you believe that Europe is something that is real, and then you need to think at the
12:45European level.
12:46Or you still think just country by country, but then you are not really competitive.
12:52And fundamentally, the big challenge that we have is that first, AI is moving at light speed.
13:00So it's not something where we can wait and think that we'll do that right, but it will take time.
13:07We need to have a response much faster and to be able to be much more agile.
13:13And here, one of the big issues that we have in Europe is that we try always to do the
13:19things, if not perfect, at least really right.
13:23U.S., first try, if it's bad, try something else.
13:28If it's good, we keep it or we perfect it.
13:31And there is a cultural gap between the two.
13:34But with AI, if we are not able to be faster, we just will have no sovereignty at all.
13:42Because fundamentally, the question is that if you don't have the best technology, you will be left behind.
13:49So, or you adopt the technology that is the best that is existing, or you are able to be, in
13:56a way, in a driver's seat in order to have something that is competitive.
14:00Now, we cannot also consider that we can be champions everywhere.
14:04So, fundamentally, to choose our battle to see what are the elements where we can make a difference
14:11and where we should be able to align and use existing technology that can come from other places.
14:19Okay.
14:19So, that speaks to a structural challenge for Europe then, 27 different regulatory bodies, ultimately, 27 different sets of regulation,
14:27but also national interests versus the European interests.
14:30Catherine, you talked about renewables.
14:33When we look at the U.S., they seem to reach for energy, any energy source to power their data
14:38centers.
14:40Can we balance sovereign AI with climate goals?
14:52One of the very important things is that the commitment that most of the companies have made in terms of
15:00CO2 reductions remains true,
15:02despite the little noise and sometimes the perception that there is a backlash on climate.
15:09So, a lot of the corporations that are using today AI, they will have to face the challenge.
15:15They are facing today the challenge of, okay, how much of this AI consumption is contributing to my CO2 emission?
15:22And there is no choice, but we're going to have to be very demanding in terms of the trustability, the
15:28visibility,
15:29not just on cost of token, but also CO2 emissions of token.
15:35Today, nobody talks too much about that topic, but I can tell you that at Engie, we have an objective
15:41to decarbonize ourselves,
15:42and we are going to put pressure on our suppliers to make sure that they give us not just the
15:48fin ops,
15:49but the CO2 ops of the AI models and AI services that they give us.
15:54But if the U.S. doesn't care about this, then that holds us back, doesn't it?
15:58Well, so two things.
15:59First of all, you know, a lot of the U.S. hyperscalers do have operations globally.
16:04So, I think, you know, markets and customers are going to be very important.
16:08And what we find is that a lot of our U.S. providers do want to listen to their European
16:12customers.
16:13They also often have made their own commitments.
16:17Companies have made their commitments.
16:19So, it's also, you know, very important.
16:21And then the third thing is that economy prevails in this situation where some people think climate is not so
16:29important.
16:29So, okay, let's look at the economy.
16:32And today, solutions based on renewables are beating both cost and time to market,
16:39which is very important for what we're trying to do here, right?
16:43So, you know, I was in Texas, in Abilene.
16:45I was visiting one of our projects, 600 megawatt, beautiful solar projects.
16:49It was a couple of weeks ago in May.
16:53The project started in January construction.
16:56It will be put online in June next year.
17:00And the team was telling me maybe it will be earlier than that.
17:02So, in less than 18 months, you get 600 megawatt of top quality solar.
17:09Not even talking about storage because, obviously, storage will be very important for solar.
17:13But that's the type of time to market and cost that you can provide with renewables.
17:18Sorry, how long does that take in Europe to do exactly the same project in Europe?
17:24So, you see, the issue of Europe is that to do a 600 megawatt project on solar is very tough.
17:31We have one underdevelopment in France, just to give you an idea of scale.
17:35It's like the biggest ever project.
17:37It's 200 megawatts.
17:39And it's the one-off and it's not even started.
17:42So, typically, the average size in Europe of solar projects is much smaller.
17:46Just look at the continent.
17:47So, it's very difficult to compare.
17:50But Europe has other benefits.
17:51You know, Europe has hydro.
17:53It does have solar in Spain, of course.
17:57We have very nice quality wind.
17:59We have offshore wind.
18:00We have nuclear, of course, in France, for example.
18:03So, you know, you also have to think about portfolio.
18:06Portfolio.
18:06And this is why, you know, the whole integration of European market is, indeed, very important.
18:12Europe is the right scale for energy as well, not just for AI.
18:16Totally agree with you.
18:18And so, that's, you know, so, basically, I would not oppose U.S. Europe.
18:22Very different regions.
18:23But what's really interesting is that low-carbon energy actually wins even if you don't care about climate.
18:30And the data center guys know that.
18:32You know, if I look at what we're doing with data center providers and hyperscaler, it's a big part of
18:39our supply.
18:40You know, we are number one.
18:41Last year, we were number one in PPA.
18:43So, long-term contract at ENGIE.
18:46A big majority of these contracts were signed with hyperscalers.
18:50So, they do care.
18:51And they are right to do so.
18:52Well, that's good news.
18:53Andre.
18:54Now, just to say it's not because each country is not wrong.
18:58It's that overall in Europe, we are right.
19:01So, fundamentally, to think at the scale, at the European scale, is very important because we have really two major
19:08issues.
19:09One is to have something that is cost-effective for energy and that is absolutely key for AI.
19:15And the second one is to have a carbon footprint that is as low as possible.
19:20But for that, we should look at the scale of Europe.
19:25Because fundamentally, to do each type of renewal in each country is, in terms of carbon footprint, maybe not the
19:34best solution.
19:36Fundamentally, by choosing the site, by choosing the place and the type of energy, we can, at the European level,
19:42be much more efficient than each country trying to do its best.
19:47Octave, in terms of the question as to whether or not we can be sustainable and competitive at the same
19:53time, Catherine's making a pretty clear case that we can be both.
19:57Is that what you're seeing?
20:00Can we be sustainable on data center build and compete at the top level?
20:06There's two answers for that.
20:07The first is, I believe, yes.
20:10This is what we've done for 20 years more.
20:13One of the examples is that we don't have air cooling in our data centers for 20 years more.
20:20Why?
20:21Because we applied, we deployed water cooling.
20:25And the water cooling, it can extract heat with very low energy.
20:31So we don't consume so much energy in our data center except for the workload.
20:35So what does it mean in a company like OVH?
20:37OVH is generating 1.2 billion revenue per year.
20:40We spend today 70 million per year for the energy.
20:45If we don't have water cooling, we have air cooling, that would be 140.
20:52So it's double.
20:53We reduce by 50% consumption of the energy because we just apply something that makes sense for us more
21:02than 20 years ago.
21:03So now it's something that everybody is saying, wow, it's a great idea.
21:09Yes, but for 20 years, our customer was saying that we are crazy guys deploying water cooling in the data
21:16centers.
21:17It's something that doesn't work together.
21:19Okay?
21:20So this is one part of the – the second part is that, you know, when you think about the
21:26CO2, you have scope 1, scope 2, scope 3.
21:28Okay?
21:30Okay?
21:30So there was a lot of numbers of the website of cloud cobriders that they provide the CO2 per hour,
21:36whatever.
21:37But usually they apply just scope 2.
21:40Some cases, scope 2 and scope 1.
21:43So this – what kind of energy you consume and what is your internal usage.
21:48But nobody except OVH had in these numbers scope 3.
21:54That is, what is this CO2 that you bought from your suppliers that they made to provide you the chips?
22:04And when you think that GPUs, CPU, RAM, there's a lot of investment, a lot of CO2 that has to
22:11be consumed, it has to be extracted.
22:14But in fact, there was a big part of all these numbers – this is bullshit.
22:19Okay?
22:19The numbers, they are bullshit.
22:20You have to apply some ISO just to explain how you – what is your impact on the climate.
22:29So there was still big room to go and to be improved.
22:34And here in Europe, we are ahead of this transparency that is just applied for the European companies, not for
22:43the American companies.
22:45Yeah.
22:47It's interesting that you were early on the water cooling systems that are now prevalent.
22:52What is your biggest constraint right now?
22:54Is it chips or is it energy?
22:57You know, today we have this 250 megawatts to go to – if you have all these data centers, you
23:10have to deploy 4 billion CAPEX to generate 3 billion revenue per year.
23:16So I'm not constrained today about the data center.
23:21I have the capacities.
23:22Of course, you could say, okay, you were lucky, deployed that 10 years ago for nothing.
23:27But actually, I spent really low money to do that.
23:31Today, we would probably spend 10 times more.
23:34But yes, today we don't have these constraints.
23:37Maybe we'll have that in five years.
23:39But for the next three, four, five years, I don't have this kind of constraint because I was a step
23:45ahead of all the revolutions, creating all these capacities when nobody did that.
23:54And everybody was just shocked or surprised that I was doing that.
23:59Yeah.
23:59Is it still – so you have excess capacity when it comes to data centers?
24:03Because what I hear all the time is it's a supply issue, not a demand issue.
24:07Oh, you can buy the GPUs.
24:11You can – today, the biggest issue is RAM, hardware.
24:15Okay, because we are in this shortage of supply and the prices, they are just crazy.
24:22It went five times more expensive in just six months.
24:26Yeah.
24:26Six months, five times more expensive.
24:29The hardware is double, 2.5 times more expensive.
24:34Wow.
24:34That will have a big impact on the cloud, on the GPUs, on the AI because it will impact global
24:41economy because everything is now in the data centers.
24:44Everything now is in the cloud.
24:46So at some point, when all this cloud will increase the prices, all the economy will have the inflation.
24:54Andre, as we think about sovereignty around energy, I can't help but think about sovereignty when it comes to the
25:00frontier models and what that means for cybersecurity.
25:02So Mythos, for example, that the U.S. now has access to, certainly the U.S. government and the glass
25:08wing, which means that a number of Anthropics partners do.
25:11But we do not.
25:11At least Europe does not.
25:13How vulnerable does that make Europe when it comes to cybersecurity?
25:16So I would first like to express you have two dimensions in cybersecurity.
25:23One is to protect your data against some form of spying.
25:29And another one is the vulnerability for disruption.
25:33And there are two different types of cyber attacks.
25:36And it's true that we see that with the example of Mythos is a very good example where you have
25:44a time advantage to the ones that get access to this model.
25:49And fundamentally, what we understand is some of the elements that were under the radar screen where people were using
25:56elements.
25:57Now it's a little bit more public, but fundamentally, in terms of cybersecurity on the European continent, we need to
26:05get real progress in a way that we are better protected.
26:10Because fundamentally, with these two dimensions, take the example of the academic research.
26:15But Europe, continent, has some very good research.
26:20And if we are not careful, we can get all the information stolen and then lose some of the competitive
26:28advantage.
26:29And here, it's true that we are behind.
26:32But it's not by reading books when you are at war that you start to solve the problem.
26:38So we need now to take the different steps in order to be, a few years from now, really good
26:45enough.
26:46But now, if we think about sovereignty, it's true for energy.
26:50It's true for cybersecurity.
26:52It's true for fundamental AI.
26:55You really have to first look at what is the cost of not being sovereign or to be sovereign.
27:01But you need to balance the element in a way to not say, through sovereignty, you are left behind in
27:11terms of competitiveness.
27:12And in reality, you need to be able to take the best model and the best elements that you find
27:18outside of the sovereignty zone,
27:20but at the same time, prepare the element to get your sovereignty or by being able to supervise to see
27:29if the element coming from other parts of the world are safe enough.
27:33And on the other side, to avoid that you have some, how to say, distortion.
27:38As an example, in our company, we have tried a model with dog robots coming from China and how to
27:47make it secure from cybersecurity point of view in a way that you are not vulnerable to cyber attack.
27:55That's an example of how you make something that is not sovereign safe to sovereign safe.
28:01Okay, interesting.
28:02Catherine, how top of mind is cybersecurity for you when you're thinking about grid resilience and building out particularly grid
28:10infrastructure to cope with these additional energy demands?
28:13Is cybersecurity something that you have to think about every day as well?
28:16Yeah, definitely cybersecurity.
28:18And in general, the concept of sovereignty when you are a global company that is present in 30 countries, it's
28:25a little bit of an odd word.
28:28So we think about this really in terms of risk management.
28:34And so risk management very quickly translates into let's make sure we don't have one over dependency on one supplier,
28:42whether it's when we buy energy or when we obviously buy AI models.
28:46So we have this really very important view that, for example, on AI model, we need to make sure that
28:52we can have access to several model suppliers.
28:56Obviously, Mistral is one of them.
28:59We have opened at Engie an AI studio whereby all of our employees can have access to different models so
29:07that they can try things.
29:08And that also, obviously, the objective is to make sure that we use AI and that we become more and
29:14more.
29:14To be clear, do you trust Mistral more than you trust Anthropic and OpenAI?
29:18It's not a matter of – I wouldn't say it that way, frankly.
29:23I would say that it's very important to have choices.
29:26We also have a little bit of a nationalistic or European, let's say, wanting to make sure that we give
29:36a good chance to our local players.
29:38So we'll make sure that we work with Mistral, but it's not a matter of we trust them more than
29:42others.
29:43It's really a matter of making sure that we have choices.
29:47We are not over dependent on one supplier.
29:50So that's very important.
29:52And to quote Satya and Adela, we don't need a frontier model for a non-frontier problem.
29:59And, you know, while at Engie we have a lot of really complicated challenges,
30:02we also have some challenges that don't necessarily need the latest and greatest.
30:06And so we make sure that, you know, we have this fit-for-purpose model.
30:11It's also very important in terms of token, in terms of cost, in terms of CO2.
30:15So that's a little bit of the philosophy that is really, really important.
30:19And then, of course, the closer you are from operating critical infrastructure,
30:25of course, the more careful you're going to be with the solution, even with cloud.
30:31You know, we operate nuclear plants in Belgium.
30:33We're all going to have on-premise.
30:35When we look at the critical gas infrastructure that we operate in in France,
30:39we are looking at sovereign plant solutions.
30:41So, you know, it's really, we're trying to obviously adjust and be rational about the challenge
30:47and the risk at hand and have the right solution.
30:50And so we're looking at this topic, I think, in a very strategic manner.
30:54But really, the key is going to be diversification.
30:57Okay.
30:57André, very briefly, because we're running out of time.
30:59Just briefly to say, we have learned from COVID that if you have static supply chain, you are at risk.
31:05And what is key for AI is to have dynamic supply chain in a way that you can switch very
31:14quickly from one model
31:15and one supplier to another one in a way that you get the best result,
31:19especially if suddenly you have some sovereign questions that can arise.
31:25Okay.
31:26As we're running down the clock, I want to do a quick fire round very, very quickly.
31:30One or two word answers from you.
31:32If you had the levers of power in Brussels, what would you change now to get to the place where
31:37we need to be on energy,
31:38data centers, and AI, and Europe's role?
31:42André, I'll start with you.
31:43Fifty percent of the procurement going to the European companies.
31:47Fifty percent of procurement going to European companies.
31:50One five.
31:51Fifteen.
31:52Fifteen.
31:52One five.
31:53One five.
31:53Not quite as ambitious as I thought.
31:55Okay.
31:55One five.
31:55Fifty.
31:56One five.
31:56Okay.
31:56That'd be too much state intervention.
31:58One five.
31:58Fifteen percent.
32:00Going from 27, on AI specifically, going from 27 regulators to one regulator, very important.
32:08And more Europe in general on energy and, of course, on AI.
32:12Okay.
32:13André.
32:13So, first, to be competitive and sovereign, but also to rethink the way Europe regulates because with the current system,
32:24it's too late.
32:25And last point that was not addressed is to look how, in terms of AI, we can come with solutions
32:33that are much more power efficient
32:35because not only that can reduce the carbon footprint, but also it can unleash new capabilities.
32:41I think that's a great place to end.
32:42Thank you all very much indeed.
32:44Of course, André, Catherine, Octave, thank you all.
Comments