Orange to productise Olympics 5G offering – and reveals LLM approach

Orange Group has said that it will draw on its experience of providing connectivity for this year’s Paris Olympics to scale its 5G stand alone (SA) private network offering, with major trials afoot in Belgium, Spain and France.

Throughout the Paris Games, Orange ran a private 5G network throughout the city, providing capacity across 32 spots and 120 official sites across the region.

Speaking at a media event in London last week Bruno Zerbib, the company’s executive vice president and CTIO, and Laurent Leboucher, the firm’s group technology officer and senior VP of Orange’s innovation division, said they now had a “unique experience” on how to make workable, scalable 5G private networks a reality.

“We’re going to spend a lot of time sharing over the next few months how we made the games in Paris work,  and how we became the telco that has the most unique experience in making 5G at scale a reality and how we are going to turn it into a product-based offering,” said Zerbib.

“We had to secure broadcast quality service within the 5G network, and we used all those advanced capabilities to make that work with tons of visitors who were using the network at the same time while protecting these very important streams,” he added.

Orange used a “standalone” version of 5G (5G SA) during the Paris Games which enabled it to leverage network slicing capabilities for this dedicated use case, guaranteeing performance on this virtual slice.

Leboucher added that the French telco was now able to offer customers “different flavours” of 5G private networks – to provide more flexibility and cost efficiency for the end user.

“Up to now we used to deploy dedicated 5G networks for stadiums and so on. But because we’ve invested in 5G SA now we can reduce the network cost  and slice it and to create efficiency. We  are starting to do this in Belgium, and Spain and soon we will do it in France,” he added.

According to Leboucher, Orange will soon be able to not just build a dedicated solution for enterprise customers on a project basis, but it will be able to transform this capacity into “a platform that can be leveraged as a service capable of combining both the scale of the public network and the ability to use the private infrastructure and architecture. The best of both worlds,” he added.

Agile all the way

Since Zerbib’s appointment last year, the French telco has consciously moved away from the traditional linear progression of the mobile generational model (4G, 5G, 6G etc) choosing instead to  act more like  hyperscalers such as Amazon or Google, offering clients capabilities, products and services through platforms or APIs.

In part, this is a reflection of what is happening within the wider industry, with mobile providers collaborating with the developer community on API integrations through projects such as CAMERA , Open Gateway,  and, more recently, last month’s joint venture spearheaded by Ericsson and a dozen or so telecom operators (including Orange) to scale their offerings and sell their APIs globally.

Zerbib, who worked in Silicon Valley for 25 years at Big Tech firms including HP, Cisco and Yahoo –  reiterated this approach, adding that it was essential that the telco rolled out new capabilities on the software side more frequently.

“Through the power of [the Ericsson] joint venture we can enable the system in a unified way with  one API and code to scale globally.

“It [the Ericsson JV] will also enable other developer platforms and hyperscaler developers by exposing the capabilities of our network. We want to build solutions that add value.”

Zerbib added that the industry needed to avoid fragmentation when it came to APIs, to make sure they were universal “so that we don’t create a lock up situation for customers.”

While Orange’s CTIO said that the company would take a use case-led approach, this translated into fewer actual examples than you might think,  although Leboucher did highlight fraud prevention as one popular application of telco APIs- which is backed up by GSMA research.

Augmenting networks with AI

The rise of generative AI and automation is something that Zerbib believes will compliment this agile approach. Internally, Zerbib revealed that AI has made Orange a better company. The firm has trained 40,000 employees to use its AI capabilities to increase operational efficiencies, he revealed.

One use case, the French exec added, would be using LLMs to map the layout of an enterprise’s entire network. The model could understand and anticipate customer support requirements and effectively become an engineer offering appropriate solutions.

Bringing it back to 5G private networks, Zerbib talked about fully automated and virtualised 5G networks that operate through AI – and hinted that the transformational 5G experience that telcos have long been espousing, will finally be realised through AI augmentation.

How the firm best goes about creating these LLMs is where Orange finds itself today, the CTIO admitted, adding that the telco was currently reviewing its options.

Zerbib and Leboucher do not currently think that Orange will need to build an LLM from scratch – nor one that was telco specific – especially when a series of generic “Vanilla AIs” were working well for some applications.

Generic LLMs – such as Chat GPT (which French exec Zerbib referred to as a “she”), Google’s Gemini, Facebook’s Llama and France’s Mistral have all  been good enough so far for network applications such as summarising tickets and alerts that technical staff investigate, Zerbib claimed.

The Orange execs added that the costs of training LLMs from scratch were currently too high -but a more cost friendly and equally effective option might be “fine tuning”  some of these existing LLMs.

Zerbib also hinted that processing power to train LLMs may require some telco expertise.

“There is going to be a need for some edge computing, where you have something in the middle,” Zerbib said. “Processing would happen not on the device or in a huge data centre but in smaller facilities of the kind Orange uses for its network