Good afternoon. Welcome everyone to the 2026 Legrand Capital Market Day. We met two years ago in 2024 in London, and we are back here in Singapore. Thanks to all who made the effort to come here. We see a number of faces we know well. I hope you enjoyed the booth tour this morning as we enjoyed it from the Legrand team. Today, a pretty busy agenda. We will start with a short overview from the CEO, Benoît Coquart, around the strategic roadmap of the group. Then Brian DiBella, the head of North and Central America, will go through the data center strategy for the group. We will have a short lunch break, don't worry, at 1:00 P.M., 1:10 P.M. Then in the afternoon, a pretty detailed technology keynote around our data center offering and product roadmap.
Then a focus on Malaysia, that is a strong success story for Legrand, and with very nice growth perspective. Then Franck Lemery, the CFO, will walk you through our upgraded 2030 ambitions. Then a conclusion from Benoît, and then a Q&A session where we will take in priority the questions from the room, but also questions online because the event is a webcast. I would like also to say good morning, good afternoon, and good evening to all the people who are connected and following the event. Now, let's welcome on stage Benoît Coquart, the CEO of Legrand.
Hello, everybody. Thank you very much for coming to Singapore, which is the place to be for at least two days for the data center world. A big hello also to those who are connected remotely. I will try to be fast because we have quite a packed agenda.
I would like to start with a few basic slides about Legrand that all of you know well. I will go quickly through those slides. As you know, we are in a large market. We estimate our market to be EUR 150 billion. We have a lot of competitors, some of them attending this DCW event. We believe that we have at least 6,000 competitors, and probably a lot more than that. Some people might see that as a threat. We see that as a fantastic opportunity because some of those competitors are, of course, natural targets for acquisitions. Last, EUR 150 billion of market, Legrand sales EUR 10.5, EUR 11 billion. You might think that, well, 6% or 7% market share, it's not a lot, but the reality that we do two-thirds of our sales as number one or number two on our market.
With leadership position, and this two-thirds metric is valid both for building and for data centers. We have a large portfolio of close to half a million product SKUs, which is probably the strongest and biggest asset of Legrand. We have catalogs which are big like that. You see as a breakdown of our sales. North and Central America represent now close to 50% of our sales, Europe 35%, rest of the world 18%. Sales by end market, 32% data center, 29% resi, 39% non-resi. By solutions, again 32% data center, 22% energy transition, 5% digital lifestyle, and 41% essential infrastructure. Last basic information about Legrand. We were IPO'd 20 years back, so we are celebrating the 20th birthday of our IPO.
You see that we had a pretty decent performance with our sales multiplied by three, EPS by 10, and share price by six, with a total shareholder return of 13.5%. We are not NVIDIA yet but over a period of 20 years, this is not that bad. A couple of words on how we stand compared to the ambitions we released two years back in London. Not a surprise to you, we have exceeded our ambitions. Top line, we intended to grow from 6% -1 0% per year, half organic, half inorganic, where last year we did 13.2%. If we take our latest updated guidance, we are shooting for 16% - 18%. We are doing basically twice as much as we thought we would do in terms of top-line growth, which is nice.
In terms of profitability, we set a guidance two years back, long-term ambition of 20%, and we were at 20.7% last year and from 20.5% - 20% this year. Both in terms of top line and in terms of bottom line, we have done significantly better than our 2024 CMD ambitions. How it is compared to our peers? We did not spend weeks thinking of the peer group. We just took ABB, Eaton, Schneider Electric. In terms of like-for-like growth, we are growing slightly less than those three guys, but we are bridging the gap. We used to grow a lot less organically, and you see that now over two years, we grew 17%, they grew on average 21%. We are doing a lot more M&A than they do. In total, like for like plus M&A, we are basically growing almost one-third faster than those guys.
With a total growth of 33% over two years, as opposed to 25% for our peers. When it comes to profitability and cash generation, those numbers are well-known from everybody. We have an EBITDA margin, free cash flow to sales and cash conversion, which is significantly above our peer group. Both when comparing to our 2024 ambitions and when comparing to our peer groups, I think that we have done a pretty good job for the past two years. Now let us jump into our strategic roadmap. As you know, our strategy is designed around three focus areas. Of course, data centers, and we will deep dive into data centers quite a lot today. As a reminder, 76% of our sales in data center come from North and Central America, 12% from Europe, 12% from the rest of the world.
We are very much North American-driven when it comes to data center, which is not a big surprise because that is the hottest area worldwide in terms of data centers. There is a second focus area, which we are probably not talking enough, and we intend to talk a little bit more about that in the quarters and years to come, which is energy transition. I will say a word a bit later on energy transition, which is more geared to what Europe, 46% of our sales in Europe, 27% in North and Central America, 27% the rest of the world. Then we have essential infrastructure. We actually embedded our digital lifestyle small business into that, which is more the traditional Legrand business.
You see that it is more balanced in terms of geographies, close to 40% of our sales in North Central America, a little bit more than 40 in Europe, and close to 20 in the rest of the world. Three focus areas. I will say a quick word on each of them. Data center, this is a topic of the day. Pretty nice performance. We grew from less than EUR 1 billion of sales back in 2021 to more than EUR 3 billion in 2026. You have couple of nice numbers on this slide. 32 acquisitions, an organic CAGR of 21% from 2021 - 2026. Accelerating, if I may say, because last year we did close to 40% organic growth. This year, our last guidance was 25%-30%. Pretty nice growth. We have 140,000 SKUs.
You could see for those who are attending physically to this event, you could see some of them on our booth. Thousands of customers. Of course, the big guys are the big customers, whether hyperscalers or colos. We are a global player. Even though three quarters of our sales are made in the U.S., we have data center sales in 45 countries. You will not find any single Legrand country without a data center approach, a data center dedicated team chasing opportunities. I am extremely proud of this slide. You could see downstairs that we have built a comprehensive product offering. Actually, I sometimes have the feeling that the financial community is looking at Legrand with the 2019 glasses. I sometimes hearing, "Hey guys, you are a PDU maker. Hey guys, you are a busbar maker." We do a lot more than that.
I hope that you could feel that in the booth. For example, who knows that we are able today to do container with gen set, which can be gas-powered or oil-powered. One container, 2.25 MW each for backup power and main power for offsite data centers. Unfortunately, we couldn't show it on the booth, but we are doing that with Girtz Industries. Who knows that we are the world leader in load banks, and that we were the first one to introduce liquid-cooled load bank in order to test not only the electrical loads, but also the liquid installation of a data center. Who knows that we have software which is installed on millions of devices to manage the load, the compute load, to compute the energy, to bring security, to report information to the DCIM system.
Today we have a product offering, which is much larger, much deeper than what you could think of when you look at Legrand again with the 2019 glasses. This slide is important, and this will be my last slide dedicated to data center. We will capitalize on new capacity and architecture. It is worth spending a few minutes on this slide. Today, you might hear about a lot of different numbers. We believe that today we have 80 GW of IT loads. I mean, today, last year, 2025, which are installed. When you put all the announcements which were made by the hyperscalers and the big guys, you end up with a potential total load of 420 GW by 2030. Most of the industry analysts discount that because you have a gap to demand, a gap to supply.
Not all the data center going to be built as expected. On top of that, you might have some bottleneck here and there, chips or whatever. If you look at the Omdia and other market intelligence people, they discount that to 250 GW. When you talk to our industry peers, they even further discount that to 180. We take the more conservative number, 180, which means that from 2025- 2030, you're going to have 100 GW of additional capacity being built on IT loads, which is a massive opportunity for Legrand. An absolutely massive opportunity. If you look at year 2030, itself, 30 GW of additional capacity is going to be built. The question that some of you raised on the booth tour was, how does it translate into content for Legrand?
We look at our current offering and the roadmap, and you have here the numbers. We believe that the current AC architecture brings a total accessible market for Legrand. Again, if we have 100% market share, which we never have, basically. It's a total accessible market of $2.5 million per MW, and you have here approximate breakdown between critical power, compute, and so on. Hybrid architecture with a site card, it's more than EUR 3 million, and full LVDC, it's more than EUR 3 million. Let me address one question I had downstairs. I was asked, "What about SST?" SST are not in those numbers because we are working on an SST. We have yet to decide whether we launch it or not. It depends on many things, including the economic conditions of an SST.
Without SST, the LVDC TAM for Legrand would be a little bit more than EUR 3 million, 3.2 or 3.3 to be precise. With SST, we believe that it's going to be 3.8. Even without an SST, we have an accessible market which will be in excess of a total accessible market, in excess of EUR 3 million. Simple math 100 GW built over the next couple of years. EUR 3 million per MW. That's it. The opportunity is absolutely massive, and now the challenge for us will be to grab as much as possible of this opportunity. I pause here on data centers, and I will let then the rest of the Legrand team to deep dive into this topic. Pretty busy slide. The two slides dedicated to energy transition do not really give justice the potential of this business.
Energy transition, as a reminder, we are not selling heat pumps. We're not selling electric-powered injection press. We are selling a bit of EV charging station, but not a lot. Every time you have a heat pump, electric press, or EV charging station, you may have somewhere upstream, circuit breaker, switchgear, transformer, load shedding, measurement, load management, and so on and so forth. We are selling the back end, if I may say, of the energy transition and the electrification. If you're looking at the three zones, you have massive opportunities ahead of us. Take Europe. The war between Ukraine and Russia, and then the war between the U.S. and Iran, acted as a wake-up call of the fact that Europe is too dependent upon gas and oil, and that it was an absolute geopolitical potential issue. There was the plan Electrify Britain.
There's now a plan, Electrify France. Last week was announced a plan, Electrify Germany. Every single European country is moving toward more electrification. You see that the share of electricity in the Europe energy mix will move from 23% - 46%, so doubling in 15 years. We have a full action plan, which I will not comment, to make the most of that. North and Central America. We are cracking the code of energy transition. Until very recently, we considered that energy transition was already occupied by a number of big players that you know well, and we didn't really find how we could be part of this market. Still, it represents less than 10% of LNCA sales.
Now, we've made a number of acquisitions dedicated to data centers that have brought us also the products, the engineering capabilities, the customer connections to sell those products into other verticals. Off-grid power generation, solar, industry, and so on and so forth. We have here, again, a lot of potential because, as most of you know, the grid in the U.S. is of very low quality, and massive investments will come to renew and upgrade the grid. The rest of the world, which is quite an obvious one, you have growth in population, growth in industrialization that should also pull the demand for energy transition businesses. We really see energy transition as a second pillar, which should bring us a lot of additional growth opportunities. This slide is interesting. It's an example of what I've just said.
We made a number of acquisitions in data center, Davenum, Avtron, Kratos, DTS, SRS, and so on. They brought us sales in data centers of about EUR 700 million, but they also brought us sales in other energy transition verticals. Again, industry, infra, PV, and so on and so forth. Avtron, that you saw downstairs, half of the sales in data centers, half of the sales in other verticals. Now we have a market position. We have the teams to start growing significantly our business in energy transition in the U.S. and in Asia, which we didn't have before. Well, I see the clock running, so super fast. We are not giving up on essentials. We still believe there's a lot of growth and profitability potential behind our traditional business. In Europe, markets are supposed to grow again.
We have put the Euroconstruct numbers that show that progressively the market is improving. It will be slow, it will be progressive, but it will come, and we have a number of initiatives to make the most of it. North and Central America, the numbers are more muted. Both in resi and in office, the numbers are not getting any better. The fact is that we have significantly reduced our exposure to resi and office building. We have put the numbers here. Back in 2023, office and residential represented 45% of our sales. This year, it should be something like 25%. Of course, a lot of that is coming from the fact that we have grown a lot our data center business, but we have also been able to diversify into other verticals, health, education, and a few others.
Rest of the world, that is where the opportunity is, of course, absolutely massive. If you look at the GDP growth and population growth in Africa, Southeast Asia, India, Middle East, we have a lot of opportunities. Again, we have put a number of examples. If I take one, which is, for example, India. India, it is Legrand's fourth largest market. So U.S., France, Italy, India. It is already 6% of our sales. It is growing double digit. Of course, we have the data center opportunity. Of course, you have an energy transition opportunity. But, for example, you have 10 million houses a year missing in India. So there is a huge wave of construction coming, and we will be able to sell a number of product, wiring devices, and so on and so forth. Our growth engines are working well.
Innovation, we have increased our spending in R&D, and we intend to keep running the group with a level of R&D to sales of about 5%. Customer experience. Every year, we survey about half a million customers. We get 20,000, 30,000, 40,000 responses. We have 80% satisfaction rate. Here again, we are shooting for ambitious targets. We want, by 2030, to maintain this customer satisfaction rate at 80%, to have NPS of at least 50%. Pricing. Since we started to record these metrics, we have only experienced year-on-year average increase in prices, and we started back in the 1990s. So last year, 2.2%. Going forward, we still intend to have a positive pricing every year, which magnitude will, of course, depend on many factors, raw mats, and so on and so forth. Another growth engine, M&A, which is really a great area of expertise for Legrand.
We are a good M&A machine. So we have put a couple of numbers to show you how selective was the process last year, which is a typical year. So we screened 1,000 targets. We engaged with 120 companies, and we closed eight deals. So we see how selective we are. We are closing less than 1% of the potential opportunities we have. It is a pretty industrialized approach, big pipeline, clear strategy and financial criteria, including something that the market seems to have forgotten, which is that sometimes it is better to have an EVA accretive deal. So to be higher than you WACC within a reasonable timeframe, 8%, 9% within a reasonable timeframe. Well-managed docking process. We have a track record which is pretty in line with the strategy. If we look at the past two years, we have invested close to EUR 1 billion. No.
We have acquired, sorry, close to EUR 1 billion of sales. We have paid on average 12 x EBIT, and it is even better for 2026. It is closer to 11. Again, it is EBIT of the year of the acquisition, pre-synergy. It is not a forward-looking, a 2030 normalized EBIT including synergy, but it is really the synergies, the multiples at the time of the acquisitions. We have a team which is engaged, motivated, young. So this is the executive committee of Legrand. On average, 52 years old. So those are the people that will have the responsibility to handle the plan. A good mix of Legrand experienced veteran as well as newcomers. Diverse in terms of profile and background. So you already know, of course, Franck, Brian, and Blandine will speak.
Some of you have met Juan, especially those of the Group One with our EVP Strategy, Brands, and Digital, who worked a lot on putting together this event. If you happen to go on the booth this afternoon or tomorrow, you will probably meet also the two other zone leaders, Frédéric Xerri for Europe and Jean-Luc Cartet for the rest of the world, who are also attending this event. Well, last slide on my side. I have one minute left to tell you about our 2030 ambitions. Do not worry, Franck will give you a lot more granularity in those targets. We are upgrading our organic growth ambitions. We are shooting for 3% -5 % per year. We are now shooting for 6% - 8% per year. We are upgrading our M&A targets. We are shooting for 3% - 5%.
We are now shooting for 5%, which is going to be financed, number one, out of our existing cash flow. Number two, we have identified a couple of assets which we believe are very interesting assets but less core to Legrand, and that we intend to sell for an average representing sales of between EUR 0.5 billion and EUR 1 billion on a yearly basis. That will help us to accelerate the move toward the energy and digital transition. A word on those divestments, because this is, I believe, the first time we announce a divestment. So margin level consistent with group average, so not dilutive to the group. Again, good businesses, but part of the essentials product family, not as core as other businesses. So instead of growing 6% - 10%, we intend to grow 11% - 13%, excluding FX and divestment.
To give you order of magnitude, I will let you do your math, but it will probably lead us to something like between EUR 16 billion and EUR 18 billion of sales by 2030, depending on the assumption you take for FX and for that divestment. In terms of adjusted operating margin, we were shooting for 20%, and we are now shooting for 21% - 22%. Of course, same definition, so including dilution coming from acquisitions and including any restructuring expenses. No change in definition. That is what I wanted to tell you. On that, I am now turning the mic to Brian DiBella, President and CEO of Legrand North and Central America, that will give you more granularity about our data center strategy. Thank you.
Okay, well, good afternoon to everybody that is here in the room, and good afternoon, good morning, good evening to our folks that are remote. Really glad to be here to talk data center. As if we have not had enough this morning. Let us get into it. Okay. Well, I did recognize some names. By show of hands, who was in London two years ago?
Yeah. A lot of folks, and I know we have some folks remote. Well, we have been awful busy, and I was reflecting on a lot of the questions and conversations that we had in London. At that point, it was 2024, we were just at this beginning stages of this AI era that we are now very much in, and there were a lot of questions about where Legrand was going to play, how much growth potential, and well, what we have learned is a lot.
The first thing, the numbers, Benoît shared the overall growth data center business since the close of 2023, last full fiscal year, 2.5% growth. Given this is a financial audience, that's probably the thing you're most interested in.
250%.
What's that?
Not 2.5%. 250.
Yeah, 2.5 x 250%. Thank you. Good one. But it's the other numbers that to me are more meaningful here, because this really defines, one, how we got there, but more importantly, as we face this AI era, we understand it much better. We know what we need to be doing and where we need to play. We have positioned ourselves. We've transformed ourself in terms of our offering. We've completed 15 data center acquisitions since 2024. So in two years, 15 acquisitions. What has that done? It's increased our addressable market by 3x. By 3x. And we haven't forgotten how to innovate. In addition to the acquisitions, 20 new platforms launched. And we use that term platform, so think about a next generation in technology. It is a DC busbar that you saw downstairs. It's a next generation of our Xerus Technology Platform.
These are not just individual SKUs, but entire platforms that are built to do what? Well, to address the radical changes in infrastructure that AI is requiring. This has been about solving that equation of what is it going to take to be successful in the next several years in this AI era. Because of this activity, we are now positioned in what we call the critical systems layer, and I will explain a little bit more what that means. Ultimately, this is about being relevant to customers, solving their biggest problems, and helping them ultimately achieve their objectives. So we saw this data before, the sales by region, 76% in the U.S. That is reflective of two things. Number one, it is the overall strength and size of the market. Number two, it is a demonstration of our effectiveness of establishing key positions with the market leaders.
We are focusing on where the market activity is, the hyperscale companies, the co-locations that come out of U.S. We are establishing a strong foundation with those key customers, and we have done that. As we look at where we are today, 76/12 to 12, as those hyperscale companies, those co-location companies, look to expand, which they are doing right now, and I will share some stories with you in a few slides, we are going to see those numbers increase.
We are going to build off of the reputation, the proven solutions that we have. Then we see this sales by solution. This is, again, a reflection of that adaptation of our product offering. We are not just in PDUs, if you will. We have got a third of our business in critical power. Physical compute, 25%. Monitoring, management, and control, that is that Xerus and other technologies, 25%. Advanced cooling, so that is liquid cooling.
Then testing and lifecycle services, 10%. So a much different. It is not a 2019 lens on Legrand. We are a very diverse player with a very meaningful participation in this market. So this is the slide that Benoît shared, and I will go into a little bit more detail where this fits into our strategy, but I want to sort of share the philosophy of how we got here. We are selective at where we are choosing to play. Our history, our foundation was built sort of in what we would call the white space area. I think what that allowed us to do is understand the relationship immediately between compute and infrastructure. When you are directly connected, you are in the cabinet, small changes in compute have direct impact to us, where companies that were way upstream in the gray space maybe did not have to deal with.
There was no change to them. As this AI era has evolved, we have identified those products and systems across this infrastructure where differentiation on the product. So being able to handle generational changes like we heard in the D.C. area today, or the interconnection between gear customization add real value. They solve real problems. We have worked our way back all the way across the powertrain to that source of energy from generators or utility power coming into the building. We have taken positions in what we call the most critical parts of that infrastructure. That does include commissioning. You can have all the products in there, but if they do not work together, if you cannot test and validate, you cannot turn the data center on. So our footprint.
Those dots, those are approximately 60 manufacturing locations that are primarily or exclusively tied to data center. Legrand has more than that. This is really a reflection of what I will call the heritage of Legrand. It is a local specialist. We have grown through acquisitions, but also organically. When we go into a market, we developed very close relationships with key contractors, with the people that are influencing the market. We understand the codes, the standards, the practices. That means we have that local knowledge, but we are not a holding company, we are an operating company. There is a network that sits on top of this that allows us to coordinate globally, to take a global view on solutions with customers, but to have the agility, speed, and decision-making that happens in a local market.
I will go into a little bit more detail about what this means from a strategic standpoint, but this is where we are today. Now I want to talk about a few of the key trends that really influence and impact our view of the future and our strategy. The first one, AI. Top of the slide, workloads are driving increases in rack densities and power demand. You can go back 20 years in the data center market, and there was always a story about rack densities were increasing and increasing power, but it is different this time. You can see the shape of the curve is steepening. There is a completely different level of power requirement. That means we have got to use new technologies, and that is primarily DC power and liquid cooling. It is also a good story because there is a lot more power needed.
Benoît shared the figures in terms of the value per gigawatt. What have we got here? Here is a look, and this is our view of what we think the mix will be in 2030. New architectures, direct current, and for those in the room that had a chance to go through the booth, really that understanding of the different generations of power that we can anticipate. This is our view of 2030, and what you can see is they are all still here. What makes us unique is we are not trying to advocate for 800-volt LVDC or a sidecar. We can handle all of it. We have the ability to be in the present, in the future, and in the distant future based on what our customers ultimately need, based on the deployments of individual projects or the availability of technology.
The same goes true for cooling here. Big shift, and this is 2025. If we went back a few years before that, it would all be air. We know liquid is coming as these power levels go up. We have got a lot more power to cool. I think one of the important parts of this that I think for us has been a, not a surprise, but a good evolution. Once you invest in a liquid cool infrastructure for that direct to chip cooling, there is an obvious solution for everything else that is in the cabinet, and that is a rear door heat exchanger. We can see where maybe a little bit in the past, a little bit more of a niche solution.
We think regardless of single phase or two phase in the future, that role of rear door heat exchanger as providing the best cooling solution from an efficiency and from a cooling capacity is here to stay, and we have a very good position there. As if all that technology and infrastructure change wasn't enough, we've also got to go faster. There's some practical parts of this. Obviously, we know the big hyperscale companies, the AI companies, are racing to compete with one another for model training. We now have this concept of time to token, so there's an economic return for it. I think beyond that, there's a practical limitation, which is labor and resource availability. So things like modular construction, prefabrication, they're new to the data center. They're innovations in data center construction.
I would tell you from a Legrand perspective, who's been in the construction game for a century plus, they're familiar to us, and they make good sense. They make the products efficient to manufacture. They again allow for rapid deployment, consistent quality. Again, it's state of the market right now. To wrap up the market view, the trend view, this really on one slide, and we'll frame up this AI era, not surprising, with two major driving forces. Number one, the one that's up front is power. We need more power. That means, number one, different technology. We need DC power. The physics require it. The efficiency is there. So we're going to see that, but it's not going to be an overnight switch. I'll talk a little bit about our perspective on that in the coming slides.
Obviously, when you have more power, you need a different cooling technology. That brings us to liquid cooling. Then power generation. Benoît shared some data about energy transition, the U.S. market in particular. We just don't have enough. So what we're seeing is more and more bring your own power being required for large data centers. Speed to build. So number one, what do customers want? How do they want to deploy? Again, prefabrication. We saw the structural containment. These are all parts of what I'll call productivity on the job site. They create speed. Modularity, again, a key part of that. Then safety. Two dimensions here. The power levels are going up, so they're inherently more dangerous. When we're trying to go fast, we want to make sure, both in the construction process and the operation, that we're keeping people safe. Okay.
So this is our strategy on a single page. The way that this section is going to work is I'll go through each of these five pillars or vectors of the strategy and then share some customer examples because this isn't just words. This is really happening today, and I think those customer examples will give you some insights into what it really looks like and how the Legrand differentiation is coming to life. Now, fundamentally, what does this all mean? Very practical focus. It's helping our customers solve this technology transition with infrastructure that can be designed, built and manufactured, deployed, and commissioned. We know there's going to be technology changes at the chip. We know that's going to lead to a lot of disruption and change in the infrastructure. We're here to solve it for them.
We want to help them achieve their objectives, whether that is token factories or cloud services or communications. Pragmatic DC transition with our customers. Now, again, I think we had a good sense of that. The DC transition is here, but it is not coming cleanly. It is not coming all at once. I will explain a little bit more about how we are going to do that.
You got a sense for the folks that are here in the booth thinking about the different generations of technology. Owning the critical infrastructure layer. Again, we want to play in areas where differentiation on service, on design makes a difference. It is important to the success of the project. Again, we have been in construction for over 100 years. We have proven we know how to manage projects and deliver on time. That is an absolute reputation that we have in the market.
We have got to continue to be excellent at that. Win liquid cooling in racks and commissioning. The liquid cooling market is still evolving. There is a lot of questions. We have made some choices here about where we want to be the best. Glocal, and I think this is the most unique part about this strategy. It is a really big differentiator as far as Legrand. I shared the map that showed all those different local locations. We have teams, and again, I will bring another map up to show a different view of that. But that ability to have deep expertise, deep relationships in-market, but a global coordination, global scale, and capability. We can be local and fast, but large and efficient. Okay. This is a repeat of Benoît's slide, and I think the important part here is that thinking across the generations is an essential part of this.
We put this one up front because we know it has been a question on your minds. I am going to move quickly through this one because, one, the folks that were in the booth that are here, I think saw this in action. For the folks that are remote, we are going to have our technology keynote after the break, and Blandine and Rebecca will talk a little bit more about the specifics. So what does that look like? What does pragmatic DC evolution look like? This is a real example for those that were in the booth this morning. We actually talked about this. This is a hyperscale company that we have been working with, and you heard the term rack and stack cabinet. This was a very specialized cabinet that we had developed for this hyperscaler over several years.
This is their preferred way to deploy 5,000 lbs, fully populated, so that is as much as a very heavy car or truck, fully populated. A lot of other unique features that we had designed before DC power into this cabinet. They said, "Legrand, we need to go OCP." This was a Blackwell Ultra, so that generation of NVIDIA chip, and they wanted to use OCP standards and 48-volt DC in the rack. They said, "Can you do it? But do not change anything about this cabinet that was never designed to do that. Just make it the same, but make it DC-capable." We were able to do that. We worked directly with them from design through prototype. Again, they came to our factory to validate and test the cabinet. So we now have met all of those load and transportation tests.
This particular one is both UL and CE certified because this is a global platform. This is what pragmatic looks like. It is take what I know, don't make me change too much, but power up the stuff that I need inside my cabinet, and that is what this has been all about. Here is that view of critical infrastructure, and I think, again, that history of starting in the white space and understanding the relationship between the IT equipment that this is all about and where, as you go upstream, any sort of break or interconnection or schedule issue can cause a critical path failure on the project. Again, as we understand the evolution of this DC architecture, where do we want to play? Where are the products and systems that really support that critical path to bringing these data centers online? That is what is reflected here.
Again, global opportunities for us from IT compute through load banks and commissioning. We will talk a little bit more about that. All the way back to generators and cable bus, which this pathway system that, again, essential to powering generators, and I will go into our first example of that one. We had a safety moment down in the booth. For folks that don't know, as a manufacturer, we take safety really seriously. That absolutely translates to job site and operations at data center. This one, a little less technical, but really important. This is a cable bus. This was a large co-location facility. There are about 30 generators that feed the facility, and they are set about 20 meters back from the building. Think about that. These rows of generators and then these cable bus systems, about a meter wide, running 20 meters to the building.
Now they need to be inspected regularly. There is maintenance that has to happen. They are running low on the ground. You can imagine it would take a while to walk up and down the rows. The practical reality is the guys that are doing that, they are going to hop over, and one, it is a trip and fall hazard. Two, they could damage the cable bus. We designed for them basically a ladder system, a step system that allowed them to go over and shortcut across to be able to inspect very efficiently to have a better vantage point. Now, is this highly technical? No, but it makes a difference. It is the reality of what happens on the job site about the application. The other piece with this product, not unique to this project, these are separate structures.
Those gen sets are out on pads 20 meters away, then I have got a huge building that it is connected to. You can imagine, when there is a weather event or a seismic event, that is when you need your generator most. Those buildings and structures could move separately. Again, one of the things that we design in that is unique is 6+ in of flexibility on three different dimensions. Not just for the outer structure, but the conductors inside. Again, that insight, that expertise. There were some questions we had this morning about, "So what? Is it really different?" It is, because we have learned about what really matters over time. Now this more traditional example, and I would say if the other one was a job site installation-related insight, in this case, we are effectively an extension of this hyperscaler's internal engineering team.
That powertrain, there is a lot of different places that you can put technology in there, from monitoring to switching. This particular hyperscaler uses our low-voltage switch gear and power panels. It is the heart of their powertrain. It is where all of the most important parts, again, not just the power distribution, but monitoring, protection, switching, is all designed into this. Like everybody else, they are coming out with their new generation to begin their AI deployment, and they said, "We need to completely redesign the product. It needs to be capable of handling higher loads, and we need you to increase your capacity by at least 50%, and we need it as soon as possible." That is what this really came down to, is working with them.
In one year's time, we went from a conversation to a fully vetted, listed product, UL, seismic rating, very flexible circuit protection, modular design built into it, and we increased our capacity by 60% out of our core factories in the U.K. and Ireland. We are now adding two new production sites, one in the U.S. and one in India for this key player. We are, again, their number one most important supplier. Design for their data centers start with us. So our track record of service. Again, 100 years in the construction business. We know that we have got that global picture, the ready-to-serve factories that I shared. We understand project management. That does not mean just doing things in spreadsheets. It means being on site. That is what construction project management is. It is in our DNA.
Technical engagement, so from design through commissioning, and then of course, post-commissioning startup and support. Again, it is how we do business as a company that has been in construction for a long time. The numbers here are actually pretty important as well. The first one, 4.5 times. So Benoît Coquart shared the numbers about our profit improvement, our great free cash flow. While we are generating that free cash flow, it also gave us the resources to invest 4.5 times our historical level in new capacity. Allowed us to be ready today and for the future to meet the demands of all the growth in this data center market. The other number, which again, I will dig deeper on in a few slides, 400 in-market support professionals.
Not salespeople, not in-office engineers, but field support people that are out there to be on the ground to make sure these projects go well. So what does that look like? Well, this one, this is a co-location company in Latin America, specifically in Chile. In addition to doing compute co-location, this was for a fiber network. So, supporting hyperscale and other enterprise businesses in the Latin American market. They were building this whole fiber access. They had a vision or a plan to build basically fiber consolidation data centers. They came to us for certain parts of the infrastructure, and we are looking at it, and the risk of trying to do this traditionally, to go out into what are some fairly remote and rugged locations and build a data center, we just said, "You know what?
Would you be willing to entertain doing this in a modular way? We can bring everything together for you in a container. We can handle sourcing all of the key systems, things like fire suppression and cooling that are not part of our expertise. We will do that directly for you. We will factory test and validate every part of the infrastructure to make sure it works. We will do that up to 10 high-level communication standards. Again, these are communication data centers, and we will arrange the deployment on-site. We have been successful. The first four have gone perfectly well. You can see what that looks like. This is the container. There is the infrastructure in there, fully built, designed, and delivered, and installed by Legrand, again, from the design phase all the way through site-level support when those things are getting dropped in.
Of course, when you do it right, you get more business. The first four will lead to six more. In this case, go back a couple years, 2024, 2025, the AI race was really ramping up. I am sure you all had a general awareness of it. This particular really big hyperscaler was in a frenzy competitively to get out in front of the market. They were securing capacity from a real estate standpoint, from a power standpoint, and from a material and component standpoint as fast as they could. They also had a bunch of co-location partners doing the same thing. They really, as a company moving fast, got a little bit out of sync. They did not realize quite how much they had committed, dumped an awful lot of unforecasted demand on us, but they had made customer commitments as well based on it.
They sat here sort of with more work, more projects, and more demand that they could handle, and quite frankly, than we were really sure what to do with. We had several phone calls with them. They actually came out to our site. We worked together to prioritize how we want to manage this massive backlog of products. We had them work with us to validate some additional suppliers to strengthen our supply chain for this higher level of demand. We also brought on some contract manufacturing to help hit the peaks, again, all with them working directly with us. From this, I will say slightly out-of-control surge of demand, five months time, we were able to triple our output. Now that, to them, means no risk to their projects because they are going to have all the components that they need.
They are going to have the systems. In this case, it was our Starline system, which is vital to the commissioning because you are providing power directly into the rack. We brought our lead times overall down 24 weeks from 40 - 16. That increase unlocked EUR 170 million of additional demand. That is sales for us that we fulfilled in the year. Winning in rack level cooling and commissioning. First one, doubling down on rear door heat exchanger. As I mentioned previously, this was back in the day, we are talking go back 2019, pre-COVID, a niche, very clever solution, very high efficiency, but you would not necessarily choose to bring in liquid cooling infrastructure for it, but it had a role in the market.
Well, what we've seen, obviously, as direct to chip becomes more prevalent and becomes necessary, quite honestly, once I've invested in that liquid cooling infrastructure, I've got a whole lot of other things that need to be cooled, right? I have power shelves. I have networking and communications equipment. There's a lot of other equipment stuff that's in the rack or cabinet that generates heat. Got to do something with it. It's absolutely the best way to handle that. Again, once I've gone liquid cooling, the price of entry comes way down in terms of complexity and cost to add that. Direct to chip, it's a reality. We're going to play there.
I think there's still a lot of questions about how that technology is going to evolve, especially when we start talking about two-phase, which again, the physics start to pull you there, at some point in time. We will continue to play. Again, I'd say there's a lot of R&D and evaluation. Then we've got load banks. This is a hidden gem in terms of participation in the liquid cooling market. There's a story I used when we were evaluating this company that I'll share with you is the easiest way to think about it. Anybody heard of the gold rush of the 1800s in the U.S., right? All the prospectors were running out. Everybody was hoping to make it rich. Well, that's kind of what's going on in the liquid cooling market. There's a lot of different competing technologies.
But in that gold rush, there were maybe a few people that actually did strike gold, but the people that got rich were the ones that were selling shovels and picks and Levi's jeans. That's what liquid-cooled load banks are. Regardless of the technology that's out there, you need to commission, you need to validate that your liquid cooling system is sufficient and secure. No matter what, we don't care what you choose for liquid cooling technology because you're going to be buying a load bank from Legrand. This is an AI specialty company, very innovative in terms of their infrastructure design, a very aggressive guy at the helm in terms of schedule. Can be very demanding, but it's a customer we really like having. It keeps us out, what I'll say, on the leading edge, in terms of technology. This was NVIDIA MGX 1.1.
Again, this is a Blackwell AI chipset. This was as of not that long ago, the latest and greatest. They needed us to, again, make a DC-compatible rear door heat exchanger. Bring DC power into it, a product that was never designed for it, but this is what they needed, and we need it right now. Again, from the request to working through prototypes and actually first article, we're able to do that in eight weeks. What does that mean? Well, they get what they need in terms of their cabinet. They get a great cooling solution for, again, everything else that's in the cabinet. Also, by going directly to DC, we save about EUR 7,000 per cabinet for a rectifier because this is now a native DC rear door heat exchanger. This one, I love this one. It's a huge co-location company. Huge co-location company.
It's their first liquid cool data center. They called us because we're the leader in load banks and said, "We got to get this right. We're not really sure about this. We want to make sure the system's going to be reliable." So before we get to product, we spent several weeks in meeting upon meeting, working out the testing scenarios with the owner, with the contractors, with the engineers. What do we need to do? How do we make sure that this whole system is going to be safe? What does this box need to have in terms of monitoring and tracking? One, strength plays to strength. We gained insights from that scenario planning that actually influenced the design of the product and certain features that are built in.
Number two, obviously, we became the standard for their fleet when it did come time to commission that data center. What we learned, the number one issue, you don't want to leak, because that'll shut down a million dollar rack in terms of equipment, could damage it. The impact in terms of loss revenue can be very significant. The other thing, yes, the primary purpose is to make sure it's reliable, but by calibrating correctly, so you can adjust your inlet water temperature, you can adjust your flow rate, you can make this cooling system even more efficient. Those pumps and those chillers, they do take a lot of energy, and we're talking about 20%, 30%, even 40% more efficient when you calibrate them correctly. It does two things.
One, make sure you are using the least energy possible, but also make sure that those GPUs aren't going to throttle down. You don't want them to overheat and throttle when you're doing training. So for us, again, great advantage. We were the first to have the product, and the product was designed with specific features that were learned through this exercise. So Glocal. Again, I would say this is a unique part. For folks that know Legrand, we've grown over the years through acquisition. Our history was in the electrical business, and I'll say local codes and standards, local practices, is really at the heart of what we've been for a long time. Those things are still extremely relevant because data centers are buildings that get built somewhere. So having relationships, understanding who is making the market, who the influencers are, is really powerful.
It's also really hard to replicate, to take a big global operation and try to drive it into a local market. I can tell you, because we've learned, it's a lot easier to build an overlay network to control and manage and coordinate across all of these local resources and capabilities. So what do we look like? I shared the number of 400 plus in-market people. You can see here by market, the U.S., 200 sales and 160 field service and support. South America, 70 and 55, et cetera. Asia, 350. A lot of salespeople in Asia, 150 support and services. Then we have a global key account team that sits on top of all of that for the largest global accounts, where they want global product roadmaps. They want assurances on product quality, et cetera. So let's talk about what that looks like.
This is actually a local one. This is a 100 MW data center. It is a co-location site for a hyperscaler. It is colo, but it is a hyperscale occupant. We got wind of this from one of our global accounts and a specific product that was specced in. The global team said, "All right, we got to do this. It is going to be a project for this particular product, Busway product. We want to make sure the local team is ready to support it." The local team is saying, "What do we know about the project?" They figured out who the contractor was. It turned out it is a contractor that they know really well. Conversations with the contractor, they found out who the engineer was. Engineer they know really well. Those conversations led to this, "Hey, this engineer is really pushing for a faster schedule.
Great that you can provide the one product, but can you help us reduce the timeline? We need to do this faster. This engineer is obsessing about this." We didn't know what it was exactly, but this really aggressive timeline, which by the way, turns out it was from a bet with his boss that it could be done. That was the whole reason that they were looking. What that led to for us is a much deeper conversation about how to take one product and engineer an entire solution. You saw the folks that are in the room, the structural rack where we had all those systems built in. Imagine getting that all done from one supplier. Everything is going in schedule.
All of those infrastructure from power delivery to cable management, PDUs, all coming from a single supplier in a single schedule, just ready to drop in and connect. That is what this project was all about. Again, this was the first one. It is a 100-MW facility. There will be more. We are now the default. You are going to hear a bit more about this after the break, from our local team members. This is the other side of this. This is a true global situation. Another hyperscale customer here, standardized on our PDUs, next generation PDU for their edge AI sites. These aren't the big training mega sites. This is when you are taking AI and you are rolling it out to your phone.
If you have that in a phone or you have an AI app in your pocket, how many folks have Grok or whatever, you are not going back to the center data center for that. That is being replicated somewhere on an edge site. Again, like everybody, we need more, we need them now, and we need global quality. We want to make sure that these are going to work. These are edge sites that they tend to be more remote management. Quality was absolutely critical. What we did, number one, we had to work with them, increase global capacity by 50%. Number two, not only trust but verify. You have heard of that statement. This was very much because these are going in remote locations, because the schedule was so critical, couldn't have any quality issues. These are intelligent, so they use our Xerus Technology Platform.
So the sensing and all those other features are really critical in addition to that power delivery. We actually, again, worked with them to come up with an approach. We built an AI-enabled, using cameras, quality control system and test system for every one of those PDUs that is fully documented and shared with them. So they know whenever they ask, "How's it going?" We can share that, make sure that they have the assurances that they need to roll out a single global model for this edge deployment. For us, EUR 40 million of incremental revenue in six months time. Again, reinforcing that strong position of Legrand as the key partner in critical infrastructure. We put all this together. AI, not a shocking thing to say, it's transforming the market. It's reshaping what infrastructure looks like. Scale, it's much bigger. It's much more complex.
We know we're going to be going through power transitions. We know we're going to have to deal with hybrid infrastructures and differences on projects and differences that each one of our customers, whether it's a hyperscaler or a co-location company, are going to demand. But our expansion of our portfolio puts us in a position to address all the major elements of that critical infrastructure. We take that strong reputation that we've built, the know-how in terms of construction and project management, and apply it across that critical infrastructure. That, for us, is a strategic differentiator. That's the unique part about Legrand, and I think it really positions us in a very unique and powerful way for AI infrastructure and beyond. That is the data center strategy for now.
I think by the clock, I'm a couple minutes ahead of schedule, so have a couple of important announcements. First thing, we are going to go to a lunch break, so I think we have everything close by. We're going to come back, and you want to be back at 2:00 P.M. because we're going to start our technology keynote. Up on stage, and I'll introduce them here, we will have Blandine Antoine, our Executive Vice President of Products and Technology, and Rebecca Gilstrap, who is Senior Director of Data Center Strategy and Products. With that's a wrap. All right. Thank you everybody.
Good afternoon, good morning, good evening to those who are following us online. We have the toughest slot in this day, which is right after lunch. I thought I will make it easy for you and give you the takeaways right now. There are three of them. First, Legrand has a broad portfolio of highly specialized, high-performing solutions that serves the needs of our customers throughout the power chain into cooling. From low voltage to medium voltage, from cooling to powertrains, from hyperscalers to enterprise data centers, we have all the solutions, and we are getting ready for more.
Second, our open architecture design philosophy and our worldwide integration capabilities allow us to provide customization at scale, thereby enabling customer choice, and thereby winning customer preference. Third point, as you have seen on the booth, as you have seen on our slides, we are ready for the transition to direct current.
We are building a reference design that supports this transition and will allow our customers to pick the degree to which they want to move into direct current. With that said, Brian has explained to you that core to our business and core to our success is our ability to solve our customers toughest problems. What is our toughest problem today? It is to get compute online as quickly as possible in a world where permitting, where power are becoming scarcer and more difficult to get. How do we do that? We must simplify complexity. Those are complex projects. We are talking gigawatt scale design centers. Two years ago, 1 GW project would only mean a nuclear power plant. Even the biggest airports are, at best, 200, 250 MW of power. Outside of China, it takes 6 - 15 years to build 1 GW nuclear power plant.
Our customers want to build 1 GW systems in less than two years. This is a very complex project, and our goal is to help them by simplifying it. Second piece is when they start executing, they want to reduce risk, and they want to go as fast as possible. This is where our integration capabilities, the prefabricated solutions, and the containerized solutions that you have seen on the booth come into play. Finally, when the facility is ready to go live, we want to make sure that it runs as a clock with high-end energy efficiency, 24/7, 365 days a week. We do that by supporting them not only through design and installation, but also through testing, commissioning, and maintenance services. Finally, we know that every megawatt matters. Sustainability has been core to Legrand's portfolio design for years.
We are onto our seventh CSR roadmap launching next year, which means we have been in this business of sustainability for 20 years. We know that we need to make our systems more efficient in terms of energy, water, but also materials, in particular copper. How do we do that? We believe that providing best of breed systems is a really good place to start. You know that our data center position has been built on acquiring very successful businesses with very specialized solutions all over the world. We start with that performance, but we know it is not enough. We complement it with application engineering and co-design so that we create those unique solutions that our customers need because they have unique needs, unique ways of working, unique operating systems. To facilitate this integration, we adopt what we call the open architecture. We are integrators.
We have our own systems, we have our own solutions, but we don't mind pulling into our systems a good and efficient product from somebody else. That's what our panel building business has been doing for years. We believe that is the best way to serve our customers because we don't lock them into a particular architecture, a particular system. We give them this freedom to operate, which is why they trust us and continue to work with us. A lot of people talk about grid to chip. I'd like for you to think that it's better to talk about chip to grid. Because where does the change come from? The change comes from the chip technology evolving very rapidly.
We'll be showcasing in this keynote, first focusing on our current portfolio, then looking at how it will be evolving, on how those changes in the chip technology are pushing changes through the whole power architecture. We like to think of our business as five pieces. In the previous presentation, we were bundling together the physical compute infrastructure and the monitoring systems. Here, we'll be splitting them in two because we want you to understand really what makes this solution special. We'll start from the chip, physical computer infrastructure, your racks, your PDUs, et cetera. Talking about how we design them so that you can have monitoring through different sensors and integration in an intelligent platforms. Moving into our cooling technologies, talking about critical power, then finally explaining how our service business is supporting the performance of our customers throughout this power train and cooling technology.
Rebecca will present to you a quick overview of our portfolio. It's not an easy thing to do. As we mentioned, it's roughly 140,000 SKUs all together, so we'll try to keep it compact, but if you need more details, we'll be very happy to answer any question you have. Then I'll take the mic back to talk about how we see the transition to low voltage direct current. With that, Rebecca, it's up to you.
Thank you, Blandine. One of the most important points that I want you to take away is that Legrand has been in the data center business, and we have built that business and that portfolio, that product portfolio, based on physical infrastructure. Down in the booth, you saw all of these components pulled together into a full data center architecture. Here we have the physical infrastructure pieces that are closest to compute and that are supporting the most valuable assets within the data center space. Starting with racks, cabinets, and enclosures, we're actually housing as a first line of security and protection to those valuable compute assets. You have pathways, taking power and connectivity, and connecting those cabinets. Then we actually have the connectivity pieces that are running the data, the bits, over those pathways.
You saw the high amperage power Busway that we have that is pulling all of that together. We are protecting, we are creating the pathways, and then we are also powering and pulling it together, again, for our most valuable assets. One of the things that our customers rely on us most for is our intimate knowledge of airflow, power densities, cable management. Again, if you were downstairs and you were in the booth, you actually saw the rigidity and the density that we are able to get within our cabinet platforms. We took all of this and we combined it with the OCP, the Open Compute Project, ORV3 rack so that we could have the best of breed from the Legrand systems and combine it with open systems and an open platform for the next generation of AI mass compute.
Additional pieces that can be added to this, you could see in the OCP and LVDC infrastructure that we had, power shelves, busbar, liquid-cooled busbar, and also the application of our ZPE serial console. Also taking all of this together and adding intelligence on top of it with our Smart Rack Controller. Again, all of the physical infrastructure to protect those most valuable assets within the data center, while also adding on the intelligence on top of that. The other piece that I will add to this is that it is not only the cabinet showing up and being there and then being loaded. Legrand has impeccable capabilities in terms of rack and stack. The density that can be deployed in dynamic loads. If you think about these cabinets being fully loaded, you can actually see on the bottom here, casters.
They can be rolled into the data center and still keep the structural integrity as if they were a static load. Physical compute infrastructure is expanding beyond just the typical pieces that you have seen, and it is becoming the backbone of AI factories. Structural racks are these backbones. Again, if you saw downstairs, there was actual data center structure set up. What you have is this structural rack is supporting Busway, it is supporting cable and power pathways, it is supporting liquid cooling. Heavy, heavy manifolds that are coming out and coming down into the racks. It is also supporting, again, that separation of those critical systems. One of the other pieces that is important is that you can take these pieces as standalone parts, but you can also put them together. That is one of the unique pieces for Legrand is that we know this space.
We understand the criticality of the space as a whole, and we understand the need for the durability and security. The other piece that I will add to it is, it is the energy efficiency that we are able to do. You are getting a consistent airflow within the structural rack. You are also getting the structural integrity of the actual backbone, and you are also getting faster deployment speed, and safety and serviceability. As we separate out those systems, so you can actually see the structural rack coming out, we are creating designs so that each of those individual critical systems can be serviced and sustained through the lifetime. Moving on to monitoring, management, and control. I will take you back, reminder on business school, you have to measure to manage. We take that to heart in the data center.
In order to be able to manage better efficiency and optimization, you need to be able to monitor and manage your platforms. Our intelligent rack PDUs, our serial console servers, and then also the additional intelligence that we have taken and applied to other products. With these visualizations, we actually enable data center owners and operators to have revenue-grade metering capabilities. We give insights so that data center owners can make proactive PUE optimization decisions, and we give the command and the control to data center operators so that they can deploy faster and then they can keep operational resilience within their portfolio. By doing so and taking all of that and extending it outside of the rack, we have created the most extensive communications protocol on the Busway market.
Rack PDUs are an excellent example of how our engineering teams are working hand in hand, where we are taking market leading and first to market technology from a hardware perspective and coupling that with the firmware that we have from our Xerus Technology Platforms. Again, coupling the engineering expertise in both hardware and again, also in firmware and software. By doing this, it has given us the largest PDU patent portfolio in the industry. The other attributes that we are able to get from this are visualization of harmonic distortion. While we have these capabilities at the rack level, it will become more important as Blandine gets into her section as in we are looking at other product portfolios, how we take this technology and apply it across our full product portfolio.
For example, we take our customized intelligent edge and power and eHouse modules, and we are able to apply the intelligence to these new construction models. They are customizable, they are prefabricated, and they are helping customers deploy compute faster to the degree of 50%. The other piece that we do from a differentiation is that we are able to be the design and engineering pieces through the entire life cycle. Our monitoring and management gives us visibility to the health of these systems, and we are able to provide that proactive participation within the data center ecosystem and help data center owners understand what the environment looks like and how they can improve their operations from an efficiency perspective.
The second example that I want to give you from a monitoring and management control perspective is a research project that we are doing where we are driving to reduce operational expenses by 50%. The other key metric with this is that we are driving and striving to reduce the carbon footprint by 60%. These are the types of R&D projects that we are going and we are partnering with and we are going out into the market, and then we are also making public so that we can continue to drive that innovation and bring improvements to our product portfolio, but also to the industry. Legrand's advanced cooling portfolio is a heterogeneous and pragmatic approach to cooling. Again, we went over the hot aisle containment and the importance of airflow from a cabinet level, but also from a structural rack perspective.
The next stepping stone is the rear door heat exchanger, where we move into active airflow, and we can handle up to 140 kW per rack. The next step is that we have single phase direct to chip and also two phase direct to chip. One of the unique abilities that we have is being able to understand that compute and then also couple technologies together so that we can provide, again, heterogeneous approaches to our customers. Let's dig in a little bit deeper on this. Our rear door heat exchangers have been in the market for over 20 years. They have been supplement and supportive of HPC, high performance computing environments. By taking this, and by taking the densities that we have coming from AI, we have a tried-and-true technology that we are able to deploy and provide to our customers.
Earlier, Brian shared with you one of the achievements that we had for a leading AI company in the U.S. We had the customer ask of taking our technology that we had and deploying a 48-volt rear door heat exchanger. Again, we went from customer ask to prototype in less than four months. We went from customer ask to production in less than five months. This is the speed and the customer intimacy that we are driving, again, across our portfolio. The other piece that we are looking at from a cooling perspective is, again, I mentioned it before, how are we combining technologies to get the highest efficiency possible? In 2027, we are launching a combination between the strengths that we have within the rear door heat exchangers alongside direct to chip immersion level thermal performance.
This is something that is ideal for multiple platforms and also incorporates many of the features that you saw within our physical compute infrastructure, including the ORV3 rack. Across the critical power infrastructure, you saw it down in the booth. Brian had multiple customer success stories. Legrand is playing across the full portfolio, from generation to power distribution, to power protection and power quality. Our strength is not only in the product portfolio itself, but also the local depth that we have and the global support that is behind that. One of these examples is from our Keor FLEX AI team. The Keor FLEX UPS team that developed the highest efficiency UPS with a modular design in the industry came and put together a solution for us to have an AI simulator.
This AI simulator is able to replicate the high fluctuations that we've been talking about from an AI workload perspective. By doing so, it is able to look at the peak loads and make sure that the UPSs and other technology are able to filter and decouple these fluctuations. This comes in 300 kW modules and is a scalable solution. Again, by having solutions like this, it gives us the testing capabilities to make sure that we are protecting the compute and that AI workloads will be protected. Across the critical power infrastructure, we have moved from being a panel builder to extending out to global engineering and integration for our containerized solutions.
So whether it is modular, whether it is prefabricated, we have power solutions that are being pulled together that are helping our customers move faster, deploy faster, have safer environments, and also have better certifications and lower risk within the data center deployment.
Not only is it the deployment pieces that we are helping from a critical infrastructure, but we are also helping with the life cycle services of these offerings. Again, this is where our monitoring portfolio comes in and gives us that visibility so that we are able to proactively, again, work with our customers from a preventative maintenance perspective, provide health checks, be proactive with spare parts, and make sure that our customers, we are working with them hand in hand on that component management. Again, we are reducing risk, we are shortening time to token, and we are making a safer environment for our customers to operate.
Across these testing and life cycle services, we are working hand in hand with our customers to design and install, to commission and do the acceptance testing, and also, as I have said multiple times, to be proactive in the preventative and corrective maintenance across our portfolio. This is building trust, it is building partnership, and again, it is working with our customers to make sure that the most valuable assets within the data center are protected.
An example of our testing and life cycle services extension comes with the acquisition of Avtron. We extended the power and thermal load testing capabilities, and as Brian mentioned earlier in the customer success story, we are able to test liquid cooling with our Avtron liquid cooling load banks. By doing so, we are able to take and replicate the loads of AI liquid-cooled servers, but we are also able to stress test those loads.
It is not only mitigating the risk before commissioning, but it is also the optimization of those systems so that they can be running as efficiently as possible. In all of that, we are driving, again, the performance and the reliability across our portfolio and for our customers. Regardless of what the construction modality is, Legrand is helping our customers move faster, move safely, and also deploy at a speed and in a density that we have not seen before. As we head into 2030, there are market reports saying that the construction model can change up to 50%. Again, Legrand is rising to this challenge and is here to work with customers. As Benoît said, regardless of what is going into these pieces, we take an open and agnostic approach and work hand-in-hand with our customers to make sure that we meet their application needs.
With that, I am going to hand it over to Blandine to go over how we are powering the D.C. architecture revolution.
Thank you very much, Rebecca. I think you've understood how all the solutions illustrate my three takeaways. One, through acquisitions and organic development, we have built a portfolio of highly specialized, industry-leading positions that we want to continue to leverage. Second, our open architecture and our integration capabilities allow us to do customization at scale, which is really a premium offering today in the market. And third, as illustrated by some of the novelties that Rebecca highlighted, particularly in the cooling technology field, we are ready for what's coming. I think what we're trying to demonstrate in the next chapter is how we'll be leveraging those three strengths to take a really strong position in the direct current transition of architectures. You all know a lot about direct current. Direct current is nothing new.
It's actually the first form of electricity that we ever discovered back in the 17th century when people, and Mr. Volta in particular, were jolting frogs with current. It was direct current at the time. But since then, we've moved to AC for a lot of many good reasons. Now the industry is trying to move back to direct current for two reasons, space and power constraints. Space. Why? If you want to have a very efficient compute system, you want your chips to be very close to one another so they can work in sync. The farther away they will be, the more the chances that they will disconnect. So you want them very close. So you want to put as many chips as you can in one rack. To do that, you need to take out of the rack anything that's not a chip.
Start with your conversion modules. Second, power. As we're growing rack densities, I think it was either Brian or Benoît who showcased that very fast-trending density in the racks. We need to bring power to that rack. I'm sure you've seen on the various booth at the exhibit today that we already have fairly big cables coming into the racks. So if we're multiplying power density by 10, 20, maybe 50, we would need, with a similar architecture, to multiply the number of cables by 10, 20, 50, which would become completely not practical. How do we keep growing the power without growing the number of cables? Well, you have two solutions. Either you grow intensity, but that's not practical because to grow intensity, you need to put more pipes for power, you need to put more cables or you grow voltage.
The engineering solution is to grow voltage. When you do that, the benefit is you reduce your losses. So you can either reduce the amount of material that you put in your system, have pipes twice as large as before, or reduce the losses and hence improve your PUE, have more effective operations. And operators will make different trade-offs. They'll probably meet somewhere in the middle on those two benefits. The second thing you want to do is not just increase voltage, but also shift to direct current, because when you do that, you can decrease the material resistance, and more importantly, as Brian showcased, you will reduce the number of conversion steps, hence growing the effectiveness of your whole powertrain.
Now that you have decided to take the conversion modules out of your IT servers to get more space in your rack, it is very easy to say, "Well, if I am taking them out of the rack, I can move them pretty much anywhere in my powertrain." The first solution is to centralize them inside the rack, and that is what the OCP ORV3 architecture does. Then you can move it into a sidecar, and then you can move it further upstream, closer to the grid, so that you win on this conversion, the number of conversion steps. Space and power are the two driving factors behind the transition to direct current. How will that transition take place? We believe, and I think it is an industry consensus, that this will not take place overnight. It will be a phased transition.
Phase one is pretty much where we are today in the most advanced deployments, where the racks are not just standard racks, they are OCP ORV3 racks or versions thereof, where you pull the AC-DC converters outside of the servers, centralize them in the power shelf in your rack. Use a direct current power busbar at the back of your rack to power those servers. So you gain a little bit on the conversion because you centralize everything, you gain in terms of spatial density into the rack. The next phase, and we have been showing on this model here what will change, with blue representing the direct current technology. The next phase is to take those power shelves outside of the rack and bundle them next to the rack in a sidecar. We have demonstrated our prototype downstairs in the booth.
The power sidecar will allow the rack to be fully direct current with a 48-volt, that is the current server technology. The benefit of the sidecar is that it is retrofit friendly. You can use the same AC powertrain as you have in current buildings and convert your compute to direct current just by putting a sidecar and changing your IT rack. In that sense, while we call it phase two, we believe it could actually be quite a permanent solution. There are many facilities that will likely choose to retrofit and just put the sidecar and the IT rack next to it, rather than upholding the whole investment and changing the whole powertrain to direct current.
Phase three is where you get all the benefits from eliminating the conversion steps in the powertrain, moving the direct current conversion from the chip closer to the grid, closer to your transformer. You will hear about SSTs. I will talk a bit more about them. We believe the TRU, the rectifier unit, has a bit more potential, at least in the short term. Low voltage direct current distribution, low voltage direct current busways. Obviously, your protection systems need to move to direct current. We are introducing the I-Load Stabilizer, I will talk a bit more about it later. Obviously, your IT rack will also evolve, with ultimately the conversion happening from 800-volt DC to 48-volt DC directly in the servers.
Going back to my thesis, it is not grid to chip, it is really chip to grid, and you see that the move of direct current coming closer and closer to the grid is driven by the increased density in the rack. This chart summarizes this transition. This is your traditional AC architecture, where you have your servers, where the conversion from AC to DC takes place in the PSUs. Then moving forward to the grid, your IT racks, PDUs, the low voltage AC busbars and Busway, your secondary protection, the UPS, the main protection, the transformers, the medium voltage switchgear, and ultimately, your substation and the grid. Legrand has solutions for every single box on that line, from the rack to the medium voltage switchgear. The architecture that are being deployed today in the most advanced data centers are the OCP ORV3, which we just talked about.
Rebecca showed you what the rack looks like. In that rack, we are pulling the conversion outside of the servers into this power shelf. Everything stays the same. Here again, we have solutions for every single box on that drawing. Next phase, which is a bit of a bridge architecture, where we are allowing retrofits from existing AC architectures to a direct current rack by introducing the sidecar, which will bundle all the various converters into one place. Same thing. This does not change, and Legrand has solutions for every single box in that line. What we see coming next is a conversion happening further upstream, and we believe it can take two versions. One is keeping your very robust, highly reliable electromechanical transformer. You have seen some of them at our booth, those big, sturdy cast resin transformers with a big magnetic core, aluminum windings, cast resin around them.
Those are highly reliable, industry-proven technologies. They do not fail. You can easily make them redundant, and we know that our customers are very risk averse. I like to give this as an example. As you have seen, one of our jewels of the crown are the rear door heat exchangers. We have iterated on this solution. More energy efficiency, hot swappable doors, lower weight, and yet customers still buy the old version, because the old version has been on the market for 20 years. It is reliable. It does not fail, and if it does fail, we know how to fix it. They are willing to let go of the other benefits because of that reliability. So they will not get rid of a medium voltage, low voltage electromechanical transformer anytime soon.
We are introducing the rectifier, which is basically a power electronic system that brings that voltage from AC to DC, and then getting the rest of the powertrain to convert to direct current technology. There is a little bit of a risk here because you have power electronics in there. We believe this is a risk that we can control because this rectifier is basically half a UPS. As you have seen in the previous presentation, we have quite a strong engineering background in UPSs. We have the highest efficiency UPS in the market. We know how to deal with those matters. So here again, if we think about this phase of transition to direct current, Legrand is developing or already has solutions for every single box on that line. Finally, when the technology is ready, solid-state transformers.
Whole bunch of questions about how we get there, but it is likely that semiconductor technologies could provide benefit at this stage. All of that will be done in the Legrand way, the way that we know how to create value for us and for our customers. Those pieces are being developed as modular, customizable systems, so that we can tailor them to the exact needs of our customers. We are adding a data control and management layer, the famous Xerus Technology Platform that you have heard about quite a bit today, and that I will go into more details too.
This is our sidecar. It is on the booth. It will be ready for commercialization in 2027. It combines a lot of our strength with customization, capabilities, intelligent redundancy, integrating the backup units to a certain level. This is a 660 kW unit, going down to 550 kW if you want some redundancy.
Advanced monitoring, seamless serviceability, easy to maintain, easy to install, and obviously sustainability core to our offering. We are aiming for industry benchmarks. Efficiency was up to 98% conversion efficiency. Going back to my transition description, we are showing here what we believe is the most likely end state in the next five years, which is your transformer, your old electronic and mechanical transformer, combined with a TRU. If I go again from chip to grid, we have a new LVDC I-rack. We are introducing this AI load stabilizer, which will basically buffer the fluctuations that AI is imposing into your power system. Your LVDC protection, the rectifier, that transformer. I have not pictured heat here, but you would have your medium voltage switchgear and then the substation. What do each of the things do? This is your rack.
We are connecting everything with a busway, easier to install, easier to maintain with flexibility in where you put the tap of boxes. You can integrate protection in those boxes. The AI load stabilizer, the protection unit. We design protection units. We have panel building businesses across the world. It is just a matter of integrating solid-state circuit breakers in those designs, and we will be able to create those offers. Across that line, we have solutions for each box. Now I would like to get a little technical because I think it is fun. This is our chips going to the grid over there. You can imagine your rack as being a piano of various GPUs.
All being put in the same place with your DC to DC conversion, bringing the 800-volt DC to the 48-volt that is used in the servers, connected to your protection system, backup generation, which could be solar panels or more likely battery energy storage systems. Your AC DC rectifier, potentially AC power generators like those that we sell with our new acquisition, Girtz, and then your substation over there. What does this thing do? It is connected to that busway, and it is basically creating a reservoir of power so that any excess demand here does not pull from the grid but pulls from that bank of energy. Some people out there, very respectable people, are trying to convince you that to protect the grid, you need to put a big system here to absorb those fluctuations.
Control theory 101 in engineering classes say that you need to control the perturbation as close to the source as possible. That is true for noise, that is true for power. That is true for anything that could transfer energy into your system, because if you do not quench it here, then you get oscillation throughout. You are moving the trouble upstream, and that can create quite a bit of damage. Why? Because those things will go up and down in power intensity quite a bit at very fast frequency. If you combine several of those data halls, sending those high frequency changes throughout the system, best case, you will have to oversize your system, so putting a lot of CapEx to make sure that you can cater to the peak of demand.
Worst case, you get into an oscillation system, and you can look up on the web the Tacoma Bridge accident in 1940, where the winds put in resonance a bridge and the bridge just crumbled. That could happen in such a system if you ended up having all your data halls coming into resonance. To make it simpler, I like this hydraulic metaphor. Current is water flow, pressure is voltage. So your high voltage dam here, pushing the current into the expansion tank, which would be your transformer, bringing the voltage to a lower level. Having your valve here to cut the power if everything goes wrong. That is your protection system. Moving into this distribution of taps. Each of those taps, each of those faucets is one GPU.
If one is off, one is on, one is off, one is on, you have somewhat of an average system. As we said in the introduction, the benefit of putting all those things together is to get synchronized compute, because that is where you get efficiency in your training. All the taps going on at the same time, off at the same time, on at the same time, off at the same time. If anyone has done any gardening, you know that if you suddenly open the tap of your gardening hose, you will find it jolting because you get a lot of high pressure suddenly, and you do not want that jolt to percolate through your system.
We are introducing this double tank system with one small super rapid transient response reservoir, i.e. a super capacitor, and one larger, slightly slower reservoir that can pick up the demand once the capacitor has already shaped it, and that is your BBU, your battery backup unit. The combination of those two systems is the AI load stabilizer, and depending on your load profile, how big the swings are, how fast the changes are, you can choose to have more supercapacitor, more expensive, or more batteries, less expensive but not as rapid.
Each customer will be able to customize their system based on their own needs. That is the I-Load Stabilizer. As I mentioned, the challenge is this ultra-fast power load fluctuation. The solution is to absorb it by putting these CBUs and BBUs on the busway so that they can absorb what is coming from the various racks in the data hall. It performs three functions.
The first one is to be the energy buffer for this powertrain. The second one is to shape the load. If you size those correctly, instead of seeing very high swings of demand, you will see a much smoother, more stable load curve that will not go as high. It is basically an average of your peaks, which makes it much easier to get a permit from a utility company. They do not like those swings, but if you can commit to a very smooth load profile, they will be happy to give you the permit, and you will not need to oversize your upstream powertrain to match that very high intensity. You can just put it at the level that you have chosen based on your design.
Finally, it also performs a ride through function, which is your typical UPS function of preventing any perturbation from the grid to making a damage to your IT rack. All of that will be enabled, monitored, and controlled by our Xerus Technology Platform. So lots of sensors in the rack, lots of sensors in the I-Load Stabilizer to check power harmonics, to check voltage, to check current, but also temperature, water leakage, intrusion, motion. We have developed this completely in-house on an open platform. Even the sensors are in-house. So we have our own PCBAs and CPUs, embedded operating system, which is called the Legrand OS, if you want, based on the Xerus, and then a variety of application and APIs which allow the systems to communicate to the DCIM of the customers.
Legrand is not super famous for software, but we believe that with this particular one, we need to make a bit of a buzz. It is already deployed in millions of devices across the world, has been on the field for more than 13 years. My three takeaways, by now you should be able to repeat them after me. First, we have a very robust portfolio of high-performing solution that has been built over the years through acquisitions and organic development. Those are specialty solutions recognized as leading solutions by the market. Second, our open architecture design philosophy, coupled to our integration capabilities, allows customization at scale, driving customer preference by enabling their choice. Finally, we are able to cater to rapidly changing needs.
Our history from the first racks we built to our current direct current system design shows that we can understand customer needs and very rapidly bring a solution to market. You heard about our rear door example system. In less than eight weeks, we were able to put a new design on the market. We have many of those stories, and this transition to direct current is another one. Direct current will not be one size fits all. There will be a variety of architectures coexisting in the market, from OCP ORV3 to full direct current, and we have solutions for each of those phases, from low voltage to medium voltage, for hyperscalers to enterprise systems, for AC to DC. With that, it is my pleasure to welcome on stage Achraf Hegazy, the CEO of our Malaysia business, and Jason Lim, Managing Director of Buss Power Systems.
Good afternoon, everybody. Good morning, good evening. Let me now take you to Malaysia, a place that has become very strategic recently. A lot of players are investing heavily in Malaysia right now. For Legrand, for us, it is not only a place where we want to catch opportunities and continue to grow, it is also a place where we are building strong capabilities to support the data center industry. Let us start with first why and how Malaysia is leading the way in the data center development in the region in Asia Pacific. Of course, there is the diversification of the hyperscaler from U.S. Of course, there is the spillover from Singapore a few years ago, where operators are looking to places not too far to build capacities in IT and megawatt. There is also the fact that in Malaysia, there is affordable and available land, power, and water.
A part of this power is already renewable. There is a strong regulatory framework such as the Data Center Task Force, the Green Lane Pathway, that help build data center faster and connect it to the grid. There is a vision, there is a strong governmental policy that aim to transform Malaysia as a digital-driven and high-income nation by 2030, and a regional digital player in the region. All of this has proven successful because we jumped from a few years ago from less than 100 MW -1 ,300 MW now. The projections are giving us 2,500 MW by 2030. Important point to note, 70% of this demand will be AI-driven. So what has started as a spillover from Singapore a few years ago has now gained its own momentum. It is a momentum built around scalable power, industrial depth, execution speed, connectivity, and hyperscale capital.
On our side, what have we done, and what is the journey of Legrand in Malaysia? We have been in the country for a long time, for four decades now. Before the data center boom, we were relying a lot on the commercial and residential segments with the acquisition of Megapower, the leader in PVC conduit in 2011. At that time, 85% of our sales was coming from electrical essential. A few years after, we acquired our first capabilities in the white space with the acquisition of SJ Manufacturing. But it is really the past two years that we grew a lot and increased our position with the acquisition of Linkk Busway Systems, the leader in bus duct and Busway in Malaysia and in the region. The acquisition of SRS Power Engineering, the leader in critical power infrastructure. The shift has happened for us in Malaysia.
We jumped from a $20 million business to a $300 million business with 65% of our sales coming from data center. 30% coming from energy transition, followed by electrical essential. Behind every adventure, behind every success, behind every story, there are teams, and we are lucky to have great teams in Malaysia. You see here the different teams that we have and all the talents that we have in Malaysia. We work together, we move in sync to bring more value to our customer and to support the industry. More than products, we want, of course, to build a solution. What makes our platform unique in Malaysia is that we leverage both on local capabilities, local engineering, and on the Legrand global portfolio with all the brands that you have seen, such as CRS or Avtron.
And we are capable to leverage on both local capabilities on manufacturing, engineering, or execution, and on regional one. This help us to tailor according to the needs and to tailor according to the local requirement. This takes me to the platform that we have built in Malaysia. We have invested a lot in three different dimension. The first one being the fact that we can do ETO or CTO. One third of our sales today come from solution that are assembled and already tested in our factories and facilities before being deployed to the customer site. Supply chain, today we are capable to accompany an hyperscale player and deliver product on a different site in the same time on the span of a few months or few years. On the opposite, we are capable to sprint and deliver needs for a specific project of our customer.
Last point, seamless execution. From engineering validation through the FAT to commissioning, testing, and installation, we have teams that are dedicated to follow up the full life cycle of the project. This being done, it's the three different ingredient of our platform, and it's helping us to build strong partnership with our customer. Few word on sales, and this, I think, was also mentioned by Brian before. The whole game on the sales, it's because the data center buying journey is already a complex journey and it's becoming more and more complex. The game that we are playing or trying to play is to make Legrand global agreement and framework work locally. For that, we leverage, of course, on the different agreement that we have, and we leverage on our local team. The game is to make the local works in our region.
We support, of course, also regional and local player. By playing on the different dimension and engaging at different level, we can engage early the specification, influence the specification, and make execution easier locally. All of this is helping us to build long-lasting relationship with key players of the industry, and I believe you can illustrate some of it.
Thank you, Achraf. Good afternoon, everyone. Good morning, good afternoon, and good evening, I guess, to everybody at home. I'm Jason Lim, Managing Director at Linkk Busway Systems Malaysia. Malaysia is indeed a very vibrant market right now in the DC landscape, and we love our position in it because we have built a very strong track record with some of the most demanding data center customers in the region. Let's look at some of these successes. Firstly, on the left, you see an international global co-location customer. We participated in their project last year. That first phase of that multi-phase project was 100 MW. But more than the scale of the project, I'd like to focus on the breadth of the project.
We participated in it and we delivered many products throughout the Legrand portfolio in this project, from cable management, racks, power distribution units, busways, busbars, all the way from the gray space throughout the data center to the white space. For us, this proved a very important point. Through global credibility and local execution, we are able to not just supply one or two products, we are able to supply a fleet of products. Through that, we push up the value pool here. Secondly, we go to the middle. Slightly different case here, a very large local colo. What is very striking here is how these local colos have also grown very rapidly. What was our first purchase order from them was for 36 MW of busbars a few years back.
It was now 80 MW last year, and this year it looks like we are going to double that throughout a fleet of products. This colo incidentally themselves have made a public announcement to say that they are going to build out 1.5 GW in the lead up to 2030. That's pretty crazy. For us, we are fully embedded in this ecosystem. We know the local contractors. We have long-standing relationships with consultants. We navigate tight timelines with them. We anticipate execution risks. Because we are able to do all of this, these data center colos, they love us. Through that, through problem-solving with them, we are heavily involved in their projects. As you can see here, we've supplied busbars, busways, skids, low voltage switchgear. We even do the servicing and commissioning with them.
More than that, we feature very strongly in their roadmap and their pipelines coming up. Third of all, hyperscalers. Hyperscalers have very strong demands in custom solutions and repeatability. Thanks to our global key account team, which has had plenty of traction with these hyperscalers. Here in Malaysia, what is tangible for us is that we see many more opportunities, many more allocations, and that's happening month to month. That number is going up. This track record is very important to us. Why? Because Malaysia, as I said, is a vibrant market. From our position, we don't want just to be counting the megawatts that are in Malaysia. We want to maximize and capitalize on Legrand content in every single megawatt in the build-outs in Malaysia. Scaling up. The market is scaling up. Legrand is scaling up as well in its capabilities.
They have acquired these two companies, Linkk Busway Systems, where I'm from, and SRS Power Engineering this past couple of years. Linkk Busway Systems is a market leader in power busbars in Malaysia, and SRS Power Engineering is a market leader as well in LV and medium-voltage power protection solutions. These two acquisitions have brought EUR 135 million of combined revenue at time of acquisition. Obviously, these numbers have gone up a lot since. I see a smile from Benoît, so he's pretty happy about it. Apart from this, it's not just the numbers, it's the synergies that have been created. Just now, Brian mentioned data center build-outs have gone from 18 months- 9 months. It's pretty crazy. In Malaysia, we see plenty of that. We haven't seen the Meta tens yet, but it might be a sign of things to come.
The focus here in Malaysia has been global credibility, but definitely local manufacturing, local involvement, local proximity. This glocal idea is very appealing, and that is totally in our Legrand wheelhouse. We have that through these acquisitions, and through this we see a lot of synergies, and these synergies provide multiplier effects for more and more business going ahead. Synergies require good docking, as I learned. Linkk, I hope, is a very good example of this. From very humble industrial roots in 1990 to over 30 years, we have built manufacturing capability. We have developed epoxy technology. We have expanded internationally. By 2024, we were market leaders in power busbars, sitting on two factories in Kuala Lumpur in Malaysia. In 2025, when Legrand acquired us, it was a proven industrial business. What is very interesting here is what happened in the ensuing 12 months.
Integration took very short time. In 12 months, now today, we are sitting in factories three and four, with factory five coming up on the horizon in coming months in anticipation of the rising demands in the market. It has been truly an accelerator in all essence of the word, and we see that in a few areas. Firstly, the team. To cope with these kind of crazy demands, obviously, there is a burgeoning workforce. This workforce with Legrand has been able to tap into global resources, global knowledge, and applied it locally and regionally. On top of that, whilst doing that, employee satisfaction has been very high, at 90% or more. Beyond that, there has been a lot of value created. Just now, we talked about how we win projects with the local colos and the international colos.
On top of that, we do have interaction between Linkk and Group R&D. We create an accelerated product roadmap, more customized solutions locally and for the region. Going forward, what do we want to do? We want to continue to win and win more. From there, we want to, as well, build more in Malaysia, develop in Malaysia, and serve Legrand customers around the world. Thanks. Back to you, Achraf.
Thank you, Jason. To conclude, three points. Outlook look good in Malaysia. In the next few years, the digital economy should contribute by 1.2% every year to the GDP. There is a clear vision, and we see a projection at 2.6 GW by 2030. Second message, as Jason already mentioned, we are not starting from scratch. We are already i n it, we have already built our platform, and this is putting us in a good position to continue to catch opportunity and to continue to bring value to our customer. With this being said, with the extended portfolio that we have, capabilities that we have built locally, we are confident to double the sales in Malaysia by 2030. This being said, thank you, and I believe it is time to welcome our CFO, Franck Lemery. Thank you.
Good afternoon, good morning, good night, probably for some of the people online. It is now time to share a few key financial figures. As an introduction of 2030 ambition, I just would like to recap a few of the major achievements and major financial achievements of the group recently. First achievement is that in a, let us say, challenging economy, we have been able to grow. We grew 43% between 2025 and 2019.
Obviously, in a retreating building market, the volume growth was very modest and was supported only by energy and digital transition. Two other growth engine worked very well. The first one is the pricing, which was at +22% on the period. Pricing, as you know, is a strategic asset of the group. It can be also a tactical asset in an inflationary backdrop. The second engine definitely has been M&A acquisition, bringing +21% at scope on the period. The second achievement is the margin evolution and the value creation. As I said, sales were up 43%, operating profit 48%, with adjusted operating margin improving by 70 basis points. It improved by 70 basis points on three factors. The first one is the organic, or let us call that an operational leverage of 180 basis points, mainly driven by productivity.
Inflation balance was not that favorable at that time, but productivity has been very strong. I said 4% of volume, like-for-like headcount -12%. That is huge productivity at the border of the group. The second factor is that we have invested a little bit more than usually in restructuring. Just to give you one figure, you know the millions, but we closed 40 sites during that period. The last factor is a very usual one. The one you know very well is the dilution coming from our acquisitions. Third acquisition, it is more about cash. Third achievement, sorry. It is about cash. First, the free cash flow. Very robust free cash flow generation, EUR 7.2 billion over the six years, at an average ratio of 15.1% of sales. What did we do with that money? Capital allocation in the period. Top priority was M&A, EUR 4.3 billion.
Dividend comes second, EUR 2.8 billion, and some share buyback for EUR 8 billion. Capital allocated is slightly above the free cash flow, meaning the debt slightly increased, but the balance sheet remained very solid, very robust. At the end of 2025, net debt to EBITDA is 1.9. This is for looking back. Now let us look ahead to 2030 ambition and start with the market conditions. We are expecting the three focus area of the group to be positive for the upcoming years. Starting with data center, 32% of our current exposure, we are expecting the market to grow mid to high teens. Energy transition, 22% of the current sales should grow mid-single digit. Essential infrastructure, 46% of our sale, is expected to grow low single digit with a progressive recovery. Now, what does it mean for Legrand in terms of top line?
2030 growth has been revised up as far as organic is concerned. We aim to achieve an organic CAGR of +6% to 0.8%. M&A, through acquisition, the scope should be around 5% per year in average. As it has been introduced by Benoît, there should be some divestment comprised by EUR 0.5 billion- EUR 1 billion of sales. In terms of margin and cash, once again, we have upgraded our adjusted operating margin outlooks. Shooting now for 21%-22% of adjusted EBIT margin, obviously always including restructuring and acquisition. Free cash flow should remain strong at 13%-15%, and cash conversion above 100%. As far as cash is concerned, that's it for capital allocation. Capital allocation will still be very clear and the policy will still be very disciplined. Top priority remains M&A, where we should dedicate roughly 60% of our free cash flow.
Dividend remains attractive at 50% of net income payout. Some share buyback, selective. What does it mean, share selective? It means at least compensate the dilution coming from LTI and employee shareholder plan. If relevant, a little bit more. According to that, embedding the divestment, the 5% of scope, we want our balance sheet to remain solid with the leverage of net debt to EBITDA comprised between 1.5 and 2.5. That's it. My last slide is a wrap-up of this financial model, of those 2030 upgraded ambitions. As you can see, it's very clear. As you can see, it's highly value creative. It's about stronger growth profile. It's about improved adjusted EBIT margin. It's about cash generation and disciplined capital allocation policy. With that, I'm on tower. Back to Benoît.
Thank you, Franck. I feel the level of energy has gone down a little bit. No? I think you need another technical keynote, maybe. I'm going to be very joking, Blandine. No, we can't. I'm going to be very fast, and then we'll open the Q&A. A lot has been said. I'd like to emphasize the fact that all what we've been telling you will position ideally Legrand for the next decade. If we look at the percentage of our sales made in what we call energy and digital transition, so data center, energy transition, and digital lifestyle, it used to be 40% of our sales back in 2021. Last year, it was 53% of our sales, and by 2030, if we do what we intend to do, should be 70%. Our objective is not to shoot for 2035 targets yet.
We are going to have an exciting decade, given the repositioning we have made. As a conclusion, we've been successfully repositioning Legrand, and I believe we have strengthened a lot the group's growth profile, as illustrated by 2025, 2026, and the new guidance. We, I believe, have a unique business model supported by best-in-class profitability, cash generation, and M&A, and we have upgraded our ambitions with value accretive ambitions for 2030. This being said, now I think it's time to open for questions and answers. Thank you.
Okay. Okay. We'll start the Q&A session, of course, giving priority to the room here, and I see already many hands that are raised. Please, when you ask your question, stand up, tell us your name and the institution you are working for, please, before asking your question. Ideally, as I see that we have many questions, if you could have one question with one follow-up, and then we'll hand over the mic to someone else. Maybe we'll start here with Ben. You can stand up and you'll get a mic.
Thank you very much. Ben Uglow, Oxcap. Thank you very much for the comprehensive presentation throughout the day. I'm sure there's going to be a lot of questions about 800-volt DC. The number that really stuck out, and frankly continues to blow me away, is your acquisition track record in this space. You did eight acquisitions, acquired EUR 1 billion of sales at 12 x EBIT. My question is, how? How sustainable? The environment that we're in seems to be absolutely insane. Everything is expensive. How can Legrand do it, and how are you doing it differently from competitors, which are allocating massive amounts of capital at far higher multiples?
Well, I cannot talk on the behalf of. Can I take that one, yeah? I cannot speak on the behalf of my competitors. How are we doing it? Well, we try to entertain a close relationship with targets. We pay fair prices. Sometimes it's 13x or 14 x EBIT, so it's not always 11 x or 12x. I think we are a good home for companies. We respect our words. We are fast in terms of negotiating a deal. Companies who join Legrand know that they're going to grow. We also have ways and means to incentivize an aligned interest. For example, we are doing a lot of earn-outs or even JVs sometimes, so that the owner can stay and participate to the growth of the company. We have a pretty good reputation. We've been able to have reasonable multiples.
I still believe that in years to come, I cannot commit on a precise number, but I am confident in our ability to continue to make deals at reasonable prices. The key criteria is not that much multiple. It is really the return on invested capital. We can accept to be at 6% or 7% on year one, provided within a reasonable timeframe, we exceed the WACC. We should be at 9% or 10% within three, four years. I believe it is a reasonable framework. Maybe we can have the testimony of Jason, because Sorry, Jason. Well, Jason, he didn't mention that, but he is the founder of Linkk. So he created the company, what, 30 years back?
Yeah.
He is the one who negotiated with Legrand.
Yes.
Maybe you can, from the target side, let's say, tell us, tell our friends, how did it work out?
I think for us as a founder, especially going through this crazy data center exponential curve, it is new to somebody like us, new to an organization like us. Obviously our heart goes to the people around us. We do not want short-term gains only. We want the company to be in a better place, the people to be in a better place. I think that's why Legrand is a beautiful fit for a company like us and for hopefully many more acquisition targets. We feel like it's the best place that provides growth and provides a safe ground, safe learning, growing ground for the people that are there. That's what I feel.
Thank you. One quick follow-up is just on, I guess, I do not know if it is for Benoît or Franck, but on the margin upgrade to 21%-22%, is that more to do with productivity or is it to do with operating leverage on the data center side?
Well, it will cover the whole. Globally speaking, I think that there are three types of levers that we can, organic levers, that we can unlock. The first one is just the growth coming from synergy of acquisition, coming from better absorption of fixed cost. Then we have the second type of lever, which has been the usual one of Legrand. The Legrand way, restructuring productivity, staying lean. The third one is about some new levers like AI. AI is a new. We have ambition in productivity in AI, more IT, more digitalization. All that will contribute. Obviously it is embedding under 21% - 22%, is embedding also the dilution, potential dilution of our acquisition. One never knows what it will be, but probably it will be the dilutive. There are plenty of other things that can be improved in the company.
The world is different when you are growing 6% - 8%. No, but then when you are growing 3%, 4%, makes your life a little bit easier. You have more SG&A to absorb. You have more leverage on your production cost. So yeah, we believe that it is consistent with the fact that we are upgrading our sales target.
Wow. That is. Maybe Daniela, the first row.
Here. Yeah. Sorry, sir.
First row.
Yep. Sir.
First row.
Yeah.
Hi, good afternoon. Daniela Costa from Goldman Sachs. Two questions. A quick one for Franck on the targets, then a more general one. Starting with the one on the targets. You have upgraded the margin guidance, but you did not change the free cash flow range. Is what you are pivoting to just a more capital-intensive business? How should we think about CapEx or working capital?
No. Well, yes. Free cash flow to sale is not upgraded. The value, of course, is upgraded because the top line is growing. But as far as the percentage of free cash flow to sale is concerned, well, it remains stable because we have more growth ambitions and two types of growth. Organic growth has to be financed as it is consuming some cash. Second, additional M&A. The previous M&A target was 3%-5% of scope. Now it is around 5%, and you know that the acquisition are dilutive on the working cap.
Okay. Thank you. Then just more general.
But in terms of CapEx to sales.
Yeah
we should be approximately the same level as today. We are shooting for 3%-3.5% of CapEx.
Yes, exactly. Typical working Cap to sales on typical CapEx to sale shouldn't move.
We will have more CapEx dedicated to capacity for data centers, but it will be compensated by less CapEx elsewhere. The total level of CapEx will be unchanged compared to the past.
But one of the two is increasing slightly in percentage of sales, I guess.
Yes, but-
Working capital.
Working capital mainly on behalf of acquisition and accelerated acquisition.
Yeah. Got it, clear. Just more in terms of the structure. I think when Legrand was more essential infrastructure, it used to be a very decentralized organization. That was sort of how you used to explain the greater margins than you had versus peers and others. The data center business is a much more customer-concentrated, more global. You are talking about bundling products. How have you adapted the organization and their need? Have you changed incentive systems? How does the whole decentralized model versus a more centralized customer and opportunity?
Well, we used to be, decades ago, a very decentralized company. But for the past 10 years, we have made a journey of being a little bit more centralized on a number of topic. When it comes to data center, the globalization is mainly on two things, product development and customer management. On product development, we have a central team led by Blandine, who is combining the capacities or capabilities of many different companies we acquired, and making sure that the work is done consistently, and maybe we will say a word on the way it is organized through tiger teams. Make sure that the specialist in power management is talking to the specialist of rack and that everything works as a system. It is now organized and processed so that the teams can work together.
As far as the customer fronting is concerned, that is what Brian said. We had a team of key account managers handling the international key accounts, which are hosted in the regions, and we work together and manage that. The fact that we are mostly a country-based organization when it comes to front office is not an obstacle, neither in terms of product development, which has always been organized quite globally, and what we call the PNT department, nor for the customer fronting, which is organized with key account managers. As far as incentivization is concerned, no, but of course, we have adjusted the targets. The data center guys are expected to do strong growth. You can grow 20% and not get 100% of your bonus because sometimes we feel that it is not enough. We have adjusted.
But the basics of the bonus, relying for country managers mostly on organic growth and profitability improvement, and also CSR has not changed. Do you want to say a word on-
Sure. Very briefly.
The way the product development is organized?
So we've recently implemented a new position, which is the one that Rebecca is leading as a global data center product portfolio, product management position. So her role is not easy because you know that even though we have global customers, there still are local norms, local installation habits, and so it's making sense of all of that to develop a unique data center product roadmap that takes into account those specificities. We have massive opportunities in that realm. You were mentioning how we're getting productivity. We believe that platforming some of our offers, which were developed over time through acquisitions, is one big opportunity for getting better cost. Because we know that even though our customers are currently racing to install power, at some point, they will also try to put some pressure on the margin.
We're getting ahead of that to make sure that we continue to have a very efficient portfolio and that we're able to cross-sell. Those local teams, such as Achraf's and Jason, have that knowledge of their market. They have that knowledge of their customers. It's our responsibility to make sure that they know about our whole offer, which is a complex and very broad offer, 140,000 SKUs across various product families. So this central position of product management will be particularly helpful. It also helps us, Rebecca is based in the U.S., it helps us bridge with our U.S. teams, who are very familiar with the hyperscalers and their requirements, and make sure we understand them and bring this knowledge to all the markets in which they're deploying capabilities.
We move to that bit. Yeah.
Hi, it's Max from Morgan Stanley. Could you just walk us through, for your data center growth target of mid to high teens, how have you actually gone about formulating that? I realize you can look at kind of gigawatts, dollars per megawatt market, but maybe just so we understand it from our side, what's actually gone into that assumption, and is there any market share gain as well?
Well, we are not very sophisticated, Legrand guys. So we took the gigawatts. We said the market going to move somewhere between the 17%, 18% in gigawatt, and we'll do approximately the same. But again, if we can do more, we'll be very happy to do more. But for us, the best proxy of the underlying market growth remains the gigawatt.
Okay. And just as a follow-up. You've previously said that you'd like the data center business at 40% of group, or maybe that's kind of how you thought about it longer term, but you didn't want to go too much. You'll probably get there easily organically. Just when we think about kind of future acquisitions, should we think about maybe you're done, or we see more of a balance of acquisitions and where they've been quite data center-heavy. How should we think about that?
Well, it's difficult to say because, of course, to make an acquisition, you need to be two. I think you are likely to see more data center acquisitions coming, but not at the pace we have had so far. What was the number you said? 15 acquisitions in two years.
Yeah.
Which are data center-related. Well, it's a lot. I don't believe we'll see that pace. You're likely to see a number of acquisitions coming in energy transition, because we really intend to build energy transition as a second pillar for the growth. Then it will depend on opportunities. We'll really see, and even in traditional essentials, if we find something which is highly complementary, a little bit like the two deals we made in, or the four deals we made in Australia and New Zealand, we'll consider. So there's no taboo, provided it makes sense, and it fits into the portfolio. And we'll have the financial means, because not only we intend to dedicate, as Franck said, 60% of our free cash flow to acquisitions, but we will also have this divestment that will provide us additional capital to reallocate to those acquisitions. Gaël, yeah. Gaël?
Thank you. Gaël de-Bray from Deutsche Bank. Could you provide the breakdown of the $2.5 million per megawatt by category, between compute and critical?
Well, we are already giving a lot more information than anybody else in this market. We have, of course, a precise breakdown. If I can give you a number, approximately half is probably critical power, let's say, and the other half is the rest, so compute, load management, cooling, and so on and so forth. This is the only order of magnitude that I cannot give a lot more granularity, because then we need to get into the individual product families. But half critical power, half the rest, approximately.
Why do you expect the physical compute infrastructure segment, the market potential for that to increase over time, and the low voltage direct current architectures, the final one, does not look very intuitive knowing that the value of the PDUs will certainly get lower?
Yeah, but the PDUs represent a very, very small value of the compute part, and then you will have AI load stabilizer, you have a high-density IT rack, you have a number of components that will either remain or be added. Then again, if you take the EUR 3 million or the EUR 2.5 million of existing PDUs represent a very small part. But we happen to have a very strong market share in PDUs because it is a historical business. We made two acquisitions, but it remains quite a small part of the EUR 2.5 million.
If I may.
Yeah, sure.
We already see it today with the ORV3 racks that we are selling. The value of the power shelf and the busbar is equal to or greater than the PDUs. Everybody focuses on that, but there is other infrastructure that replaces it.
Yeah. So. Move to James.
You go, James. You stood up already.
Okay.
Thanks , James. I'll stand up. Andre from UBS. I have a quick question on the growth targets. I just ran some maths here, and with data centers growing mid to high teens, that gives you 5% - 6%, I think, already. If we throw in a percent for the energy transition, then it leaves very not much growth for the rest of the group at the low end of your target. Is that how you see it, or is that your kind of traditional conservativeness?
Oh, traditional conservatism. The very day where you upgrade our number, you consider us as conservative. No. We want to have clear evidence that the building market is going to rebound. Again, without repeating what I've said, it is not expected to be supportive in the U.S., and the rebound in Europe is expected to be quite slow. Yes, the majority of the growth should come from data center and to a lesser extent, energy transition.
Thank you. If I may ask second question on the portfolio pruning that you announced. Yeah of EUR 0.5 billion- EUR 1 billion. Yeah, c ould you just talk about what provoked it, and what kind of businesses we are looking at? Is that geographies? Is that product categories?
I cannot be too specific in terms of what has been identified, of course. It is not a number we are throwing away. We know precisely what kind of asset we intend to divest. Those are good businesses, as profitable as the rest of the group. So it is not a matter of boosting our margin by selling a failing business, a non-profitable business. It is just a matter of capital allocation. We believe that our capital would be better allocated to data center energy transition topics than on those essential assets, which have some value, and that is why we think that we can easily sell them, but which have probably more value for other owners than for Legrand.
It is purely a matter of freeing some capital and reinvesting this capital into assets that have more strategic interest, more growth potential, more synergy potential with Legrand than what we intend to sell. We will be more specific the day we are going to sell them.
Thank you.
I have to add that, yes, it is new because Legrand has never announced such a portfolio pruning, but we have been doing the exercise every year. It is not something we discovered. Every year, we have a series of meeting internally where we try to identify assets that are less strategic to Legrand. I have met some of you in the roadshows, and some of you asked me whether this process was in place. I said, "Yes. We have not found so far, but it may happen that we find something which is of interest of selling." What we have finally found. The process is not new, but we think that again, we have a better use of our capital than being invested in those less strategic assets. Over.
Sir James Moore from Rothschild & Co. Redburn, thanks for the time. Two questions if I could. Benoit, this is not meant as a critical question, actually quite the opposite. Delighted with your defense of 800-volt DC today. It is great to see the ship set going up. But for the cynics who are going to come back at me, how can I answer when there is probably a plus side of the equation and a minus side of the equation? Presumably that increase from 2.5 - 3.3 or 3.8 with solid state, if we ignore solid state.
Yeah
That increase of however much that is, what, $0.80 a megawatt, is presumably made up of a decline on some products, PD or others
Yeah
an increase. I just wondered if the magnitude of the decline, is it like $0.50 of decline, $2 of decline offset on the other side? How significant is that piece of the jigsaw?
Well, I cannot easily answer this question, but the importance is that net is a plus and a clear plus. At the end, what really matter is how much of those. It is not that much whether it is 3.1 or 3.3 or 2.9. What really matter is how much of this value pool you can capture. Today, the reality is that when we have 1 MW , we almost never sell $2.5 million. When we sell $0.5 or $0.6 or $0.7 million, we are happy enough, because we are able to get some product families.
99% of our time is spent not on computing whether it is going to be 3.1 or 3.3, it is spent on trying to increase the 0.5 - 0.7, and then to 1 and 1.5 by providing our customers with higher added value, by bundling products together, by coming with container or skid-mounted solutions, by adding capacity through acquisition. That is what we do rather than really scratching our head for too long on whether it is going to be 3.3 or 3.4. I do not know if you want Brian to add something.
Yeah. The other part of it, the PDU market is different. There is where we lead significantly, which is in intelligent PDUs, and then there is the traditional, we will say, dumber, more basic PDUs. If you look at that across the segmentation of the data center market, our intelligence is valued more at the edge. Think about smaller sites, remote management. The big core compute sites, generally not using Legrand PDUs. When we talk about the AI training loads, it does matter because they are going to become a larger and larger percentage of the market. The reality is we are not nearly as impacted in the core where you are doing AI training, when you are doing the heavy loads. Those are more dumb PDUs, quite frankly, from competitors.
There is an incremental opportunity if you look at us now having power solutions in cabinet, in rack, in those core sites where we do not do a lot of that today, and still being very relevant at the edge because you are not going to have those massive DC power loads when I am doing basic inference type work or cloud and networking work.
I think-
Yeah. There is an unequal cost benefit based on where we actually play. Again, as you are thinking about it, that is why we cannot just sort of spout a number because there are two dimensions of change. One is product A for product B, but the other one is the application piece, which again, think AI training versus everything else.
If we zoom on UPS, because UPS is the second application where we should disappear in, I think, the hybrid and then the full LVDC application. It is part of the EUR 2.5 million, but our market share in UPS is today very low. So we see a lot more opportunities behind the BBU, the AI load stabilizer, which incorporate part of the UPS function than risk in losing UPSs in AC architecture that we do not really sell today.
Great answer, thanks. When you think about your growth plans for AI in general over the next four or five years, how much, it is easy for all of us to look at the megawatts announced and the additions, and the supply side of the equation is abundantly clear. But do you look at all at the demand side of the equation and the degree to which AI is actually going to be useful, and the degree to which model companies are actually going to make free cash flow, and the sufficient degree of free cash flow to justify heading towards EUR 2 trillion of CapEx at the end of the decade? Or do you just leave that as a problem that is frankly all of ours and too big to solve?
Well, I am not sure we have a say in it.
Well, I think Benoît is right. We do not have a crystal ball, but it has been interesting to see that in the past few months, the gross margin of some of those players have moved from 35% - 80% based on new pricing schemes. So it looks like they are finding a way to make money out of those big investments, and we will see how it all pans out. But the news seems to be positive so far.
We've seen a lot of, you had a couple of years back, this DeepSeek topic. You had, a few weeks back, the fact that AI could kill 10% of humanity within 10 years. The key question is, are the CapEx slowing down? The answer is no. On the contrary. If you look at the past couple of quarters, there's been an upgrade or an increase in the CapEx forecast of a lot of those guys. So up to now, there's absolutely no slowing down in CapEx investments.
Yeah. You take your question. Yeah.
Thank you. It's Phil Buller from JP Morgan. It sounds like you don't have any real gaps in the data center portfolio, but you also have quite an aggressive 5% M&A target. How should we think about the focus areas technology-wise? Is this DCIM, BMS, more cooling solutions?
Well, we cannot be too specific because, again, it's confidential information, and I don't want the whole market to go after the same targets as us. We still have gaps, either product families or geographies, because sometimes we are relevant in one product category in the U.S., but not yet in Asia or the other way. Second comment, not all the 5% going to be dedicated to data center. We have a number of gaps in energy transition that we intend to-
Fill
to fill. We still have a lot of ideas. I told you that we have a pipeline about 400 targets. I didn't make the exact count, but out of the 400 targets, you probably have, I don't know, 50, 60, 70 targets in data centers. A lot of discussions going on. But again, statistically, you will not see the five or six acquisitions a year dedicated to data center. You're probably likely to see less, but a lot of acquisitions in other complementary fields of activities.
As a follow-up to that,
Yeah
you touched on competition. The rate of change towards this LVDC is very quick. Are there any specific categories within the data center that you think are most easily addressed by the competition, perhaps that's solid-state transformers or something else?
Well, solid-state transformer is going to be a highly competitive area. This is one of the reasons why we are still wondering whether it makes sense for Legrand to have its product offering. But again, it's one piece of the powertrain, and the powertrain is one piece of the whole data center architecture. Trust me, even in LVDC, there are going to be a lot of pockets of growth and pockets of profitability. Go down and you will see the AC market is already super competitive. You have a hundred and hundred of players providing rear door, busbar, medium voltage breakers, and so on and so forth. Now, I hope we have demonstrated that despite it's a competitive market, we have all the assets we need to continue to grow. It's true in the AC world, it's going to be true in DC world.
DC will not be a game changer in terms of competitive landscape. All the more as all the different architecture will survive and coexist. I had a discussion at break time with some of you. Take the hybrid. A year ago, there was this concept that the hybrid architecture could probably last two or three years, and then everything will switch to LVDC. We believe that it will probably last much longer than that, because as was said, this is a perfect solution to retrofit existing AC-powered data center, which will not be willing to go through the cost and pain of moving to full DC. A lot of different architecture will coexist. A lot of competitors will continue to be tough, but we have the assets to win.
Thank you.
Yeah.
There's a question from-
Maybe it's just next.
Okay. Yes.
Thank you. It's George Featherstone from Barclays. I'd just like to start maybe on the AI load stabilizer. Perhaps a bit niche, but is the right way to think about this product as a viable, competitive solution to the medium voltage UPS that might sit outside the data center that we hear from some of your competitors?
I'll be direct. I think a medium voltage UPS is way too far from the AI load to provide the protection it's claimed to do. In any system, you want to quench the perturbation as close to the source as you can. They might say it provides the same function. I don't believe it'll be as effective, in part because it means that the perturbation has to travel all the way through the powertrain to that medium voltage UPS to provide this buffering capability. You lose all the benefit of downsizing your architecture because you need to have this high fluctuating, intense load go through your powertrain. We do think that the AI load stabilizer, which is our version of what you see called ESR in many of the reference architectures, has a real advantage over that particular solution.
Now you'll have space for everybody. No, but I mean-
Well, the other-
They can, yeah.
Yeah, the other advantage is if there is a problem, you are in a much more isolated area of impact, right? So if something goes wrong at that rack or in that row, and I've got my load stabilizer there, it's isolated to one subset. If I've got it back in the gray space, everything downstream can be impacted from that, which again, it's a much more serious impact to the reliability and the performance of the site. Again, whether it's wave perturbation or even fault protection in any architecture, the more you can be close to the point of use, there's a lot of advantages to it.
That being said, it doesn't mean that a medium voltage UPS would not add value, right? It would have a different function. You still need to provide backup power and transient protection to your ancillary systems. Think about your cooling system, think about all the ventilation systems. But in terms of supporting buffering for the AI loads, it's not the right solution from a technical perspective.
Okay, thanks for that. Then maybe just a general question now on the data center business. You are clearly doing a little bit more direct to customers, and this is different from your legacy business. So I just wondered, within the contractual terms of that, now you are doing a bit more commissioning as well of the data center products. Are there any extra warranties or anything like that that we need to think about on an ongoing basis in terms of what you commit to the customer?
Globally speaking, no. It is a new business, but you shouldn't think about a riskier business from a legal contract management point of view. We are getting organized on that. By the way, this is one of the most centralized contract management business that we have. We have trained many people. We are implementing AI-supported contract management system in order to be fast and to go very deep. So, no. Looking at is the balance sheet riskier or the business globally riskier on that front? No.
Okay. Thank you.
Yeah, thank you. It is Martin Wilkie from Citi. The question was really on the hyperscalers and how their technology roadmap has defined the strategy for you. Because when we look at some of the stats on how big full DC could be by 2030, there is a big wide range of estimates from different companies, different consultants, and so forth. But do you get a two-year roadmap from your customers to inform you of what you need to develop for 2027, 2028? Or how much of that is that you have to sense what is needed, and how much of it is that you are guided by your customer?
It's actually NVIDIA that sets that, because they have the roadmap on the GPUs. I would say the hyperscalers reluctantly comply. They like to tell us NVIDIA doesn't know anything about building data centers, yet they dictate all of this different infrastructure. That's context. Relative to your questions, it really depends on which one. They all have general ideas of how they want to design. I would tell you the reality is never exactly what that is. We do sit down, and we'll work on product specifications. I mentioned the example I went through where I said we're an extension of their infrastructure, and they're putting a lot of functionality into those switch panels. That's probably a two-year to three-year roadmap on that product as it fits into their powertrain.
The other components in that powertrain that are less dependent on the functions might not be as long reaching. It will vary by both product and the particulars of the customer. As much as they know we'll be involved in those conversations. I would just say, we value it. It's important to make sure we're going to have product ready. The volume unit forecast has been shifting a lot. These first couple of years in this AI era have been really dynamic. I'd point to, go back to, and again, back to NVIDIA, the Blackwell rollout. Yes, they did start producing them, but the demand that happened, there wasn't sufficient supply, so they reverted back to Hopper production, which had slightly different architecture. Projects got reprioritized based on chip availability.
All the forecasts that had been put in place were put aside, and then we did what we had to do to support the market. It's an and both kind of dynamic. The other thing depends on the product. Talk about gen sets. We have Girtz now. Gen sets, because the lead times are so long, they're going to be at least planning high-level capacity and trying to get in line with orders two to three years on actual hard demand, not just a general roadmap and forecast, but get in line for capacity. That works back. Load banks are around a year, and then the rest of the infrastructure products are anywhere from four to six weeks to about six months. All of those things will shape how specific they are with that. Again, each one has a slightly different approach.
Some are global. Others will be global for 50%, regional for another 50%. There's a lot of variability in that.
But there are two consensus today in the industry that the numbers we gave, so probably less than 5% of new IT load by 2030 are going to be full LVDC and 35% going to be hybrid. So total 40%. It is more or less industry consensus today, number one. The other industry consensus is that it will also depend on the purpose of the data center. We told you that out of the 30 GW of additional capacity that should be built in 2030, 70%, our estimate, will be AI and 30% non-AI. Out of the 30%, 70% are going to be AI. It is probably, I do not know, one quarter training, three quarters inference. The training piece is more likely to have full LVDC because that is where the benefits in terms of energy consumption is going to be very significant.
The penetration rate of LVDC in inference data center should probably be much slower. This is also an industry consensus. Now, again, the importance for Legrand is not to. Of course, we are tracking that carefully, but everybody can have his own advice. Our objective is to be ready whatever happens. So we have to be ready for the AC, OCP, hybrid, full LVDC architecture with product that will answer our customers' needs. If it is not 5% but 15% by 2030, fine. We will be ready. It will be a good business opportunity for Legrand.
Thank you.
Can I also add, just from an engineering perspective, we are whiteboarding with our customers. While they are sharing their roadmaps with us, we are also sharing our roadmaps with them. We are also uncovering problems as we are at the table with them or as we are doing data center walkthroughs. Brian had several examples in those customer success stories where it was joint development that was found out through the relationship, through the customer intimacy. They told us about a problem. We stepped up, and we solved that problem for them. That was not necessarily on a roadmap.
Thank you.
We move to Delphine.
Right. Delphine.
Yes. Thank you. Delphine Brault, ODDO BHF. Can you disclose the share of service in your data center activity? At what pace is it growing, and is it currently more or less profitable than the solutions you provide?
Well, we told you that life cycle services represented about 10% of our sales. Part of that being product, mostly load banks. It is probably half of that being load banks and the other half being services, so commissioning, installation. So it is quite small. Well, it is growing nicely. Now, if you compare our service sales with one of our competitors, you have to have in mind that there are two products that have a lot of associated services, UPS and cooling.
Well, we are not selling a lot of UPS in data centers, and cooling, it is quite small. It is 5% of our sales. So if we had a bigger chunk of our sales in UPS and services, we will do a lot more services. But given what we currently have as a setup in terms of product, it is about, let us say, half of our 10%. 5% for the data center sales, approximately.
Thank you. Second question, you say that you are working on SST, and you will see whether you will launch it or not.
Yeah.
Can you be a bit more specific on what will drive your decision? Is it related to market acceptance, market penetration, or competition risk as you touch upon briefly?
Well, it is a very basic return on investment. Given the profitability of the SST, is it worth doing the investment? That is it. We are in a world where technology is, sorry ladies, but is pretty widely available. Either you can source the technology, you can license the technology, you can buy a company. It is not a technological issue, it is more will there be enough market potential, margin, to justify the investment and the effort. We have so many opportunities in the data center world that we need to make sure that it is profitable enough. Again, it is not like if SST sales was driving the rest. You can be a very good SST player and not selling any IT rack, not selling any direct current switchgear, not selling any busbar, not selling any AI stabilizer.
The other way, you can be very good on all the product families without having an SST. Since there is no connected sales between SST and the rest, it is a pure ROI decision. That is all.
Yeah. Alasdair.
Yeah. Thank you. Alasdair Leslie, Bernstein. I suppose a kind of follow on to that because it is good to see that the TRU is part of the technology roadmap. I guess, when could we potentially see that launched? If we were sort of thinking about the sidecar potentially comes into the market in late 2027, could the TRU soon follow after that? Are you more confident around the economics of the TRU versus the SST simply because obviously you are using perhaps more of your existing technology here? Thank you.
Do you want to take that one?
Sure. We think we're fairly close to the TRU just because of our current asset base in the UPS systems. It hasn't been our highest priority for development. I think that's one of the takeaways from this presentation is it's a very fast-moving market, and we have to reallocate our resources super fast to catch the right waves. We believe we could get one to market if we put all efforts against it within two to three years, which we believe would be timely to catch that wave.
You don't need to have a product ready for the LVDC within six months, huh? Because the market is not yet there, and what is really important is that we co-develop the product. We need to spend some time showing what we demonstrated in the bus. We have already started to some of our customers, making the adjustments, trying to get into their specs, sharing the technical difficulties. If we have everything ready by sort of hybrid 2027, full LVDC by 2028, which is a sort of roadmap we can have, it makes a lot of sense. It's very consistent with the way the market is moving.
Yeah. Sorry. I can't take it if you want.
Thank you. Dominic from Millennium. I had two questions, one on the capital allocation. Share buybacks are now a part of your capital allocation policy, and you say it's selective. Can you just elaborate a little bit on the criteria you will apply to those buybacks? Just back on the envelope, if you look at your leverage targets, and the other M&A and dividend allocation you're planning to do, it looks like you could be buying somewhere around EUR 8 billion worth of stock up to 2030. Does that sound like a sensible estimate? The second question is on the data center forecast you're applying, the 180 GW, which is significantly below other forecasts, such as the 250 from third party. Can you talk a little bit about how de-risked you feel those forecasts are? How much visibility you have at this stage?
Well, it may very much be 250. We just took what most of our peers, the ones who know the market, shot the numbers, and the average is 180. If it was to be 250, then great. It will provide us with even better growth opportunities. We thought it was reasonable to take the sort of our industry consensus rather than Omdia or third-party estimates. Frankly speaking, whether it's 180 or 250 doesn't make a huge difference at the end. It's what. I didn't do the math, but it's a 23%, 24% CAGR instead of 18% or 17%, something like that. Yes, it will provide with additional growth potential for Legrand, but it's not like doubling. It could be 250, but we decided to take the industry consensus, our industry consensus, which is more around 180. On the share buyback.
I can take the question on the share buyback. Definitely, it's not your number. Let's say 13% -1 5% of a free cash flow to EUR 16 million -EUR 18 million top line cannot bring EUR 8 billion of share buyback. Answering to your point, what does it mean, selective? At least the minimum will be to compensate employee distribution, LTI employee shareholder plans. That's the minimum. The rest, it will be a little bit opportunistic. Meaning if the pipeline of M&A were to be smaller, which it doesn't look like, there could be some share buyback. Or if we were to overperform the model, so being on a longer term above the 15% of free cash flow to sell, then we will consider additional share buyback.
Let's look at what we did two years back.
Yeah.
Our level of leverage was down to one net debt to EBITDA, so we decided to launch. It was not a big program. We decided to launch a EUR 500 million share buyback program. We executed 400, and then we started to have a pipeline of acquisition building up again, so we decided to stop the program. The reason why we launched this 500 was typically because we overperformed in terms of cash management, in cash generation. If we have a lot more cash, or if we have more cash than expected, we'll not do additional acquisitions for the sake of doing additional acquisition. Then we will consider giving back to the shareholders. This is the approach.
Yes, as we are reaching now close to the end of the Q&A session, I propose that we take one of the questions we had online.
Yeah
Which is pretty interesting because we are asked the questions about our customers, so the hyperscalers and so on. I do not know who will answer that question among the team, but what infrastructure decisions should customers from Legrand make today to avoid limiting their business in the next 10 years?
Oof.
Avoid limiting their business in next 10 years.
That is pretty interesting to get that question because at least for this time, we are asking questions about our customers. What we believe in terms of, I would say that why I am asking this one, because I got it maybe four or five times, which is asked in a different way. I have it in another way. What is the biggest challenge for your customers for the next few years? Is it power infrastructure, cooling software? Limiting to the next or integrating everything into one complete system?
Well, I am not sure I have advice to give to my customers. Everybody can react. I think the biggest challenge ahead is definitely the exact theme of this trade show. It is how can we use less and less megawatt to run the compute system? Not only it is a matter of energy efficiency, but it has become a matter of social acceptability.
Yeah.
Not in Europe, actually, but even in the U.S., you have a growing number of concerns. The fact that data center, especially hyperscalers, are consuming too much energy and too much water. I do not believe it is really a water issue, but in terms of energy, it may become a problem for local communities. So I think our customers will have to demonstrate that they do whatever it takes to lower their PUE down to the 1.0234.
That short term, yes, you can have fossil generation. But midterm, you should also have renewables and so on. That the products they are going to use will be increasingly using recycled materials, and I think it is a matter of social acceptability. So if we want the market to grow, I think it is going to be really important for those guys to demonstrate that they are increasingly taking that into account. But this is my takeaway. Maybe you guys are closer to the customers. You can give your own feedback.
Yeah. In addition to that, I think having flexibility with your design, so not being beholden to a single technology roadmap, but understanding that whether it is a plan A or a plan B, or thinking about where you can replace product A for product B. I think the transformer example is a good one. Do not bet on solid-state transformers that do not exist yet or have not been proven as part of your infrastructure because it is, quote unquote, "the most optimal way to do that." I think optionality is going to be key. Because a constraint in one area can be solved if you have kept the design flexible by changing out other parts and systems.
Thank you. So we reached the end of the CMD.
Any question?
Maybe one word of conclusion.
Do you want to ask a question?
Benoît, sorry.
Do you want to ask a follow-up question?
No. I think we reached the end.
He's the boss. Okay. Well, just wanted, of course, to thank the Legrand team who participated in this and contributed to this event. Ronan's team, the Southeast Asian team, Juan and all the people. It has been a lot of work to organize this event. But more importantly, I wanted to thank you for spending the time. It's what, two days spent on Legrand. It's a big investment from your side. I know that you have a lot to do with other companies, so thanks a lot for making the trip to Singapore. Should you have any follow-up questions, the whole team is at your disposal for further information. This being said, we're going to be pleased to invite you for a small drink. You will have other opportunities to interact with the rest of the team.
Yeah.
Thanks a lot for coming. Thank you very much.
Yeah.