Ray Le Maistre, TelecomTV (00:05):
So we're in London for future net world 2025. I'm here with Bruno Zerbib, he's group CTIO and EVP at Orange. Bruno, thank you very much for joining us today. Really good to see you. See you now, you are on the opening keynote panel here at the event this morning, and one of the topics that was discussed was how there's a shift now, or you were suggesting there's a shift within the telecom operator community towards truly considering the needs of customers rather than just supplying services and application. So is that really happening across the whole industry or maybe is it only happening at orange? And how is that manifesting itself?

Bruno Zerbib, Orange (00:51):
So it's happening, I don't want to oversell it. It has, it's not a solved problem, but I would say that there's an awakening at the scale of the industry that the notion of build it and they will come, doesn't work. That's what we did for 5G for instance. We hope that by providing better technology, we will have emerging use cases didn't work like this. And so as we are transforming our infrastructure to be less about hardware upgrade, but becoming more about continuous updates that are software defined. And as you know, we're becoming more and more essentially an operator of a cloud infrastructure. And with continuous innovation, we're going to have to become agile. And when you are agile, it means you need to focus on user stories. You need to try to figure out what are those use cases that you're going to be building towards. So as we put an end to this generational upgrade paradigm to some extent not entirely, and become more obsessed with continuous innovation, then the business case for each continuous innovation is going to be about the demand side. And so that's changing the culture. We are now re-imagining what is the function of a product manager. And it's not about just a 20 euros a month plan for 5G. It's really about enabling this use case for those customers that need low latency because that's the killer app that they have in mind.

Ray Le Maistre, TelecomTV (02:19):
Is that kind of model going to be possible to implement, though? That sounds like it might be quite complex. And the thing that customers are very often complained about is, I don't understand what I'm getting for my money from my service provider, whether that's the enterprise or the consumer. So is it possible to make this simple to meet user demands and make the proposition simple?

Bruno Zerbib, Orange (02:46):
We live in a world of paradox. I think that on the one hand you have people that say, Hey, I would love to get a very cheap plan for two euros a month. At the same time, you have people that say, I'm willing to spend three, four euros a month if you protect me from scammers. I don't want to get those undesired calls. So you're saying you're valuing protection from an accessory calls more than the connectivity itself to some people. Did you? It's very weird because from a CapEx perspective, one is much simpler to address than the other, and unfortunately is the one that is the most expensive, that sometimes is the hottest one to explain. So that's the situation we're in. I think as we are focusing more on the needs and the perception of those needs and the criticality of those needs for our customers, then we can assign some monetary value in a much more effective way. If it's all you can eat connectivity, then it's becoming a commodity. A commodity remains essential, but it's a commodity. It's not differentiated. If you want to monetize more and capture more value, you have to move away from a commoditized product.

Ray Le Maistre, TelecomTV (03:59):
Okay. And does this mean that over time, or maybe this is already happening, that the key relationships for companies like are changing from the hardware infrastructure providers to the cloud platform operators that will enable this kind of agility?

Bruno Zerbib, Orange (04:17):
So five years from now, so we'll be, I mean we believe in infrastructure at Orange. Essentially if you compare us against other telcos, we've made the commitment of investing in infrastructure, owning infrastructure. That's part of our

(04:32):
Strategy and that won't change. But on top of that, what we're really going to be is as automated as the AWS Azure or Google Cloud of the world, we have no choice. We are going to embrace the same kind of technology, the same kind of agility, the same kind of culture. We're going to have our N becoming eventually a dark N because that's the way it works in the cloud environment. That's the natural evolution. Once you use the same technology, then you have to also adapt the same culture, otherwise it breaks. So that's the evolution path that we're on.

Ray Le Maistre, TelecomTV (05:14):
And that cultural change is happening now. Are you seeing it within your own and other organizations?

Bruno Zerbib, Orange (05:19):
Yeah, yeah, it is. It is. Because first of all, I mean clearly we are moving away from those appliance that you have to upgrade every six months to a year with software upgrade to something that's much more dynamic. Now everyone at Orange that is in tech on the technology side of Orange is very comfortable with the notion of Kubernetes containers orchestration. It's kind of obvious and it's interesting because it's not necessarily people who are in their twenties, could be people in their forties and fifties that did not grow up with those technologies that have fully embraced that. So it makes me very optimistic. The thing we underestimate is that telecommunication engineers are very educated, very well-rounded, very smart, and they can really adapt to new emerging technologies much more. And maybe sometime we might give them credit for.

Ray Le Maistre, TelecomTV (06:09):
Okay. Now of course underpinning driving a lot of this change is AI and the arrival only two and a bit years ago of generative ai. And we are hearing a lot more about how AI enablement is changing network operations and the way that telcos run on a day-to-day basis. But is there a way to measure the actual impact and gains of productivity gains or any kind of gains within a company like Orage against the use of ai? Is it possible to put your finger and say, we have made a 10% gain here, or we are now 30% more productive in that team? Is that possible?

Bruno Zerbib, Orange (06:50):
Yeah, it's more than impossible. It's mandatory sometime the question is, is that cost avoidance? Is it business as usual? Because as you know, over the last 10, 20 years, 25 years, we've been getting more efficient. So that efficiency is built into our own plan. We are being expected to become more efficiently over year. Like any company, we have to do more with less. And there's an expectation that we'll do that. AI is in support of that expectation. That expectation P precedes ai. We have to be more automated no matter what happened. And automation did not start with ai. So I would say as someone, one of the person I talked to a few months ago in the Bay Area told me, Bruno, if you really want to think clearly about ai, just understand AI is another word for software.

(07:47):
Once you accept that, then the discussion changes entirely. It's a newer form of software based software. So it is absolutely built into our plan. The way we work with our finance team, it's really about productivity, how we become more efficient from the way we support our customers, the way we react to outages, how we do all of those things. And we measure those things and we say, okay, that's the outcome that you're going to see. And we are communicating now in comfy to the financial community. We're now officially communicating the savings that we're doing with ai. It's actually something that's been tracked and very official because that's something that we are reporting back now. We did that for the first time this year.

Ray Le Maistre, TelecomTV (08:41):
Okay. Now of course it was the same with when cloud became a big thing. Everybody got really excited about the cloud, but some people forgot that you actually need the networks and the connectivity for the cloud to exist and to work. And with software, of course, we need the underlying infrastructure and the best underlying infrastructure to make it reach its capabilities for a company like Orange. What do you need in terms of, let's call it AI infrastructure to be able to meet your goals? Do you need to develop your own edge data centers? For example, do you need to source your own GPU based servers? What kind of infrastructure requirements do you need now?

Bruno Zerbib, Orange (09:30):
I think my answer is evolving. So first of all, the network API topic and the AI topic are the same topic because all those APIs have to be in support of those emerging use cases that are going to be AI defined. And to really be a network, API company, you have to be a cloud company. Now the question is, should we spend hundreds of millions, billions on buying GPUs? If we did this two years ago and we had a lot of pressure to do that, we would've made a big mistake. It would've been money that the return investment would not have been there. They were very expensive. The Allens were not very effective. And for me, it's like buying an Nvidia GPU. It's like buying a nice tube that is melting before I have the time to put in a cocktail, right? By the time I find the cocktail, which is the killer app or the app that can monetize it's water, so it has depreciated so quickly that it's a mistake.

(10:27):
So that is the thing that kept me up at night, to be honest with you, is the depreciation, the speed of depreciation of dataset, the GPU, we have the facilities, we have all the data centers, we have that. We've been doing this historically. So we have that. The real challenge to only build an AI factory, it's really the cost of those GPUs and the fact that we're talking about huge amount of money. And so I think with the models becoming more effective on the optimization that is happening and we're talking about one or two order of magnitudes better because really it's the cost of inferencing. Now, the fact that we have sovereignty issues emerging, the fact that we have latency requirements in this new hybrid AI world, you're going to see a set of use cases where AI at the edge will be important as much as you can do it on the device.

(11:17):
And as you know, an iPhone today is as powerful as an M1 PC Mac of four years ago, which means in four years, the phone will be as powerful as an M four pro, which means that a lot of things that you'll be able to run some basic chat GPT on the phone three, four years from now. The only issue is the battery. I don't fully know how it's going to play out, but three or four years from now, you'll be able to do 23, 24 use cases that are in the cloud on the device, but you'll have multimodal processing requirements, low latency. And so what we think will happen between that, between TY trust control points, customers saying, Hey, we want to use American technology but we want to make sure it's being managed by European company or maybe French technology, French LMS or American LMS or with some kind of control. All of those things are going to really force us to think very hard about what is the role of telco operators with those AI factories. And I think eventually we'll have to play a role that will be

Ray Le Maistre, TelecomTV (12:19):
Significant, but for the future it's all could go any way at the moment.

Bruno Zerbib, Orange (12:26):
Yeah. But it's evolving in that direction. I think. I believe in infrastructure and AI is just another infrastructure play. And at the end of the day, again, the network API play, the AI play is the same play.

Ray Le Maistre, TelecomTV (12:38):
Okay. Alright. Interesting. Bruno, thanks very much for your time today. Great to talk to you and good luck for the rest of 2025. Thank you very much when things are moving so quickly. Thank you.

Bruno Zerbib, Orange (12:49):
Thank you.