Singtel has long had a reputation as a trailblazer among operators. Group CTO Jorge Fernandes talked to Contributing Editor Annie Turner about network slicing, network sharing, network APIs and AI’s future in the network.

Singtel is Singapore’s largest fixed line operator and one of its four mobile carriers. Recently it became the first to achieve 5G coverage throughout the Mass Rapid Transit network, including the six underground lines, plus 27 stations of the new Thomson-East Coast Line.

Its speed of innovation and willingness to partner mean Singtel has considerable presence in south-east Asia and beyond, aside from its ownership of Optus, Australia’s second largest telco, and a substantial stake in Bharti Airtel, India’s second biggest carrier.

Jorge Fernandes became Singtel’s Group CTO in June 2023, having previously been Chief Technology Information Officer at Canada’s Rogers Communications and before that CTO at Vodafone UK. He characterises this time in Singtel as one of “major building blocks” slotting into place. “There is a lot going on, but it’s a very exciting time,” he says.

5G Standalone moves 5G on

“We’re beginning to see green shoots of 5G benefits particularly with the deployment of 5G [Standalone] SA. They are still as relevant as we hoped they would be, but perhaps haven’t been monetised as quickly as we hoped,” Fernandes says. Now Singtel is working closely with some very large enterprise partners, particularly in Singapore, to co-create opportunities to leverage 5G SA and “the industry ecosystem [which] is starting to catch up [and] is somewhat connected with 5G SA.”

For example, he expects Reduced Capability (RedCap) devices to usher in “interesting, lower cost, simple solutions that will enable a much broader availability of IoT devices at the right price point”. Singtel ran Singapore’s first RedCap trial with partners MediaTek and Ericsson late last year and signed an MoU with Qualcomm to work on other 5G trials and use cases in February.

Fernandes adds, “Because we have the 5G SA network we can do [network] slicing. The opportunity is there, but you need the network, to invest in the infrastructure, and that ecosystem of partnerships.” Singtel pioneered the world’s first app-based network slicing in February, with Ericsson and Samsung.

Not every means of monetising 5G is about leading-edge technology. Another option is to expand into new territories. For example, in August, Malaysian operator Maxis signed up for Singtel’s all-in-one platform which will deliver 5G network, edge computing, cloud and services orchestration.

Another option is infrastructure sharing. In April, Optus signed a network-sharing agreement with TPG Telecom. Under this regional Multi-Operator Core Network (MOCN) agreement, Optus will provide TPG Telecom with access to its RAN in areas of low population and Optus will license some of TPG’s spectrum.

Fernandes says, “This arrangement leverages our scale and sharing is a tool to improve return on investment. In regional Australia we have very sparsely populated areas, yet we need to provide critical communication to those Australians…With this agreement, we’re able to deploy 5G in regional Australia probably two years earlier than we initially thought.”

Opportunities abound with AI

The next major building block identified by Fernandes is AI and of course especially GenAI, but he emphasises, “We do not build technology for the sake of building technology. I always start by asking my teams questions like, ‘What problem are we trying to solve and what is the benefit to our customers?’.”

He sees there is plenty of opportunity around AI but stresses it’s early days, especially for what Singtel sees as the all-important, telco-specific large language models (LLMs). The foundation for telco-specific LLMs is telcos’ data. “Infrastructure vendors and even cloud providers don’t necessarily have access to that data”, he says. “We need it to be available in environments that have the ability and the tools to manage vast amounts of data and train these LLM models.”

Fernandes adds, “This is another [area] for partnership, where you have data science engineers from the hyperscalers who know how to build and train LLM models and their agents, working with telco engineers who understand telco language configurations, logs and incident logs, and can use this data to train new LLMs. We are very early in this phase but we need very strong collaboration in this ecosystem.”

He stresses that the cost and the labour required to train a model for telco specific use is, “Something we don’t believe makes sense for any one company. Here is another example of where we’ve partnered with other talents to create the Global Telco AI Alliance (the GTAA) [in February 2024].” The other founding members are SK Telecom, Deutsche Telekom, e& Group and SoftBank Corp.

In June, the founding partners signed a further agreement to set up a joint venture to co-develop and launch a telco-specific, multilingual LLM in English, German, Arabic and Bahasa [Indonesia’s official language], starting with customer care, for instance, digital assistants and chatbots. Fernandes acknowledges, “When we get into the networking architectures, it’s going to take a bit longer but there too we will see incremental capabilities built over time.

“For example, we might start with the radio interface, looking at radio optimisation, then extend use cases into transport, into the core. We will see incremental results over time. I would like to have the first use cases for telco within the next 12 months, then build from there.”

Changing relationship with hyperscalers

Is AI changing telcos’ relationship with cloud companies in other ways? Fernandes expects GenAI will generate new content in the cloud, leveraging human knowledge, whereas previously it was created by consumers and enterprises. He says, “That will have an impact on our infrastructure. We’ll see an increase in demand in the infrastructure required to connect cloud environments and an increase in the level of content moving not just upstream, but downstream.

“That dynamic change will require much lower latency, potentially moving cloud workloads closer to subscribers. For inference purposes, you don’t need very large scale cloud and GPU environments, a lot of the inference can happen closer to the user. That will drive a much closer collaboration between service providers. We’re seeing that in some of the investments they’re making in the region.”

Some examples of the dizzying number of investments and partnerships can be gleaned by the fact that in September, the operator and GMI Cloud announced a strategic collaboration to expand GPU capacity by combining both companies’ GPU resources and infrastructure. This announcement followed Singtel’s signing of a strategic partnership with Nscale to “unlock both companies’ GPU capacity across Europe and Southeast Asia” in August and another with Bridge Alliance in the same month.

In March, Singtel announced it will launch GPU-as-a-Service (GPUaaS) in Singapore and Southeast Asia in the third quarter of 2024 to provide enterprises with access to NVIDIA’s AI computing power.

Opening up APIs

Another area that Singtel views as being critical to developing new revenues and models is network APIs. Again, partners are at the fore. Fernandes says, “We’ve announced a partnership with Ericsson and 10 of the largest operators in the world to enable a global scale aggregator for APIs. We have a number of services we’re enabling, both in Singapore and in Australia, and across a number of our associates in Indonesia and Thailand, the Philippines and in India, but this [move] opens up the API landscape at the global scale.

“When you take into account all these building blocks beginning to come together, and the API landscape that will enable us to expose capabilities to our customers.”

Why all the excitement about open APIs now? Fernandes explains, “The network APIs, especially exposing Network-as-a-Service [NaaS], is something that enterprise customers and financial institutions have only begun to understand in the last 12 to 18 months. Previously the industry hadn’t come together in the way it has with the work done by the GSMA. If you’re a large financial institution, you want to serve all your customers, regardless of the network they’re on.”

What about the knotty issue of getting app developers on board? Fernandes argues that the aggregated approach will attract the developer community because the commercial model is easy to understand: any customer, whether a hyperscaler, developer or enterprise can reach the entire industry, or any specific part of it, easily through standardized, open APIs.

“If you speak to the developer community, they don’t design the applications for a future that may come,” he says. “They design applications based on the capability that’s there.” In September CAMARA, which is aligned with the GSMA’s API initiative, published its first major release of network APIs and committed to two releases annually in future.

Five years’ time

How does Fernandes see the telecoms industry’s future in five years’ time? He thinks AI will be the biggest accelerator of change in telecoms over the next five years. He says, “In the 90s, we knew the internet was going to be big. We didn’t understand how big. I’m reminded of that time now with GenAI. For me, as CTO, it is about making sure the company and our associates are ready and have the core foundational capabilities – the network infrastructure and data – in place so we can leverage and deploy new use cases.”