Orange accelerates plans to transcend traditional network limits

The rise of AI

Most significantly, what has really been accelerating is the impact of artificial intelligence (AI), said Zerbib.

“We are now using artificial intelligence to make Orange a better company across the board,” he added. “We’ve been training. We’re have around 40,000 people that have been using our AI capabilities as tools within the company, so that’s just an incredible amount investing in terms of training.

“The way we think about AI is not to replace humans, [but] to give them more capabilities, more tools – from the way they’re going to help in interacting with customers, to the way we are designing our network and network capacity, making sure we optimise the capacity of our network. So, with the capex we invest, we can extract more value while being smarter about the notion of network intelligence.”

Zerbib revealed that while the use of AI is nothing new, the use of it – including generative AI (GenAI) tools – over recently is having a “massive” impact on the way Orange is thinking about the evolution of automation to become more effective and increase its quality of service. 

In one key use case, Zerbib believes that large language models (LLMs) are evolving so that firms can capture the topology on an entire network, and that models need to be fine-tuned and not retrained – a massively expensive process – through regular use. For example, the model could understand evolving and growing customer support needs and become a smart customer support engineer, learning from past issues and offering the appropriate solutions.

But with added AI, there would be more demands placed on the Orange network infrastructure. Zerbib envisaged a world of supporting LLMs working on an exponential scale, with as many as 100 billion parameters running at the edge of the network, working similarly to a typical current content delivery network server caching traffic to make operations meaningful with quick response time – and without hallucinations that endanger user trust.  

Zerbib readily accepted that Orange did not have the complete answer, but said it is getting much clearer on the emerging problems compared with a year ago, which is already substantial for the company in terms of how it’s engaging.

“We’re trying to figure out how we’re going to deliver 100% availability with extremely low latency, with incredibly demanding expectation in terms of uploading traffic from the device all the way to the cloud,” he said.

“What’s the role we’re going to have? We want to make sure that our network uses AI to be much smarter, and we want to have the best network to deliver AI. We cannot do that alone. We don’t believe we can build something at Orange that the other telcos are not going to be using. It doesn’t work like that. There’s no scenario where open AI only works with Orange and then it has to figure out another solution from the other telcos.

“So, to some extent, we believe that the telco industry has to tackle this together. Everybody’s working…to make sure that we have a consistent approach towards solving that end problem. And it’s all about this target customer experience.”

Even though the company believed that designations such as 5G were coming to an end, the functionalities offered by the current generation of mobile technologies, in particular 5G standalone (5G SA), would be crucial and boosted by AI. Network slicing was a key example, said Zerbib, offering users real business benefit with demonstrable levels of quality in their use cases.

“When you look at capabilities of network slicing, you have ability to have an express lane of certain kinds of traffic for your customers,” he said. “We believe that this is not going to be a static thing, where you’re going to have a baseline for an entire user all the time. I don’t think that will scale.

“This notion of dedicated and secure quality of service has been promised for many years and many times, and it fell multiple times, because each time you have those express lanes, [they] get saturated. So, the question is how do you make that happen this time around, in a way that’s really going to deliver on that expectation? We think the only way we’re going to do that is to have advanced orchestration powered by AI.”

An Olympic experience

Going forward, Zerbib said that the company would draw on “the invaluable experience” of providing the underlying communications and broadcast networks for the Paris Olympic Games.

The infrastructure covered major stadiums such as the Stade de Marseille and the Stade de France, through to areas such as the Invalides, to airports, railway stations and training centres, as well as iconic and unusual locations such as the Marina de Marseille, hosting sailing events, and the French Polynesian island of Tahiti, where sports such as surfing were held.

This resulted, said Zerbib, in the company being the telco that now has “the most unique experience” on how to make 5G at scale in private network a reality. He added that there was a story to be told of how Orange made it work and how the company was going to turn that into a product-based offering.

And this would be a product based on 5G SA, added Laurent Leboucher, Orange group technology officer and senior vice-president of the Orange Innovation Networks division.  

“We have different flavours of 5G private networks. We used to deploy dedicated networks for a campus and for stadiums. Because we have invested [in] 5G SA, now we have become capable to reduce [cost of networks] in slices and to create highly [efficiency],” said Leboucher.

“We are moving from a situation where we are not only capable of building dedicated solutions for customers on a project basis, but also transforming this capacity into a platform that can be leveraged as a service [for enterprise customers]. And this is really the power of having invested in 5G SA.”

The result is that businesses can get the scale of a public network and the ability to use the product architecture of private infrastructure.

“And so now you get the best of both worlds,” Zerbid said. “People want to have their own personal connectivity, their own personal network, something that always works. In terms of applications requirements, [Orange is also offering] cyber security and availability. And we’re doing this with more capabilities that we’re rolling out on top of our network.

“And that’s [another] reason why you don’t want a generational paradigm anymore. You want something that is incremental, something that looks like what a hyperscaler would be doing.”

Abandoning generational thinking

This notion of discontinuing the generational paradigm also included 5G Advanced, a new standard that was being rolled out by the major Asian telco such as China Telecom, China Mobile and China Unicom. For Zerbib, 5G SA and 5G Advanced was “essentially the same thing”.

“When we say we’re not in this generational world, this is really about the rolling out of more advanced capabilities,” said Zerbib. “And so, that’s exactly what we’re talking about.”

Drawing to a conclusion, Leboucher emphasised that Orange was now able to change infrastructure on a grid without disrupting the network function or disrupting the service on top. That, he said, was something that really made a difference.

The way in which the company used to update infrastructure was to “bring a truck and do things like upgrades like before”, said Zerbib, which was “very painful”. The reason why navigation evolved slowly was, he said, because the company was not capable of making updates fast.

He also warned that there was a potential drawback due to reliance on essential enabling technology suppliers whose development roadmaps were not fully aligned with Orange’s, meaning the ability to run open software solutions became more important.

“We come from a world where you had essentially two vendors: Ericsson and Nokia,” said Zerbib. “I’m exaggerating things, but taking everything from them – hardware, software and all of their feature sets – [meant] we were stuck with their innovation velocity. And the rest of the software world doesn’t work like that.

“So, first, we have to [tell the vendors], ‘We only want functionality from you, but the ground infrastructure is going to be based on open source.’ And then to only have a feature functionality, it needs a container to be essentially shipped, and then it can be easily changed.

“The new world that Laurent is describing is a world where we can more easily have many more companies providing granular technology components. Today, we think about it in a legacy way, thinking about 4G and 5G architecture, but tomorrow, maybe we need capabilities powered by AI in all of the network. And I want to push that across the board.

“The landscape of innovation outside of Orange – for all the telcos – is getting more broader and more granular, and we can have much faster innovation as well.”