Singtel lays out its AI progress, ambitions

• The operator has long been investing in AI software and datacentres
• Its business model around its AI assets offers short- and long-term returns 
• Its widespread use of AI across its operations make it a leading AI-native telco

Singtel clearly laid out its AI progress and ambitions during its latest investor day, which highlighted how advanced its AI portfolio has become and where it is heading next. 

The operator might be most often associated with being the leading telco in its home market, the island state of Singapore, but it has long been one of the largest regional operators – it owns Australia’s Optus and holds significant stakes in Telkomsel in Indonesia (30.1%), Globe Telecom in the Philippines (46.6%), AIS in Thailand (26.8%), and Bharti Airtel in India (26.9%), with the latter holding also giving it an ownership foothold in Africa courtesy of Airtel Africa. 

It has also been revamping its structure in recent years, and developing its own technology, to position itself to be a significant player in the AI infrastructure and services sectors, initially in Singapore (naturally) but also in other markets – and not just those where it has existing economic shareholdings.

At the heart of its AI infrastructure and services strategy is its Digital InfraCo division, which comprises its datacentre, subsea and satellite assets, as well as the already well-established Paragon multi-network, multi-cloud orchestration platform – it emerged as long ago as early 2022 – that provides a single management system for communications, cloud, edge, network APIs and AI workloads.

Within Digital InfraCo, Singtel boasts its Nxera AI datacentre operation, which has facilities in Singapore, Malaysia, Indonesia and Thailand, and which aims to have 230 megawatts (MWs) of operational capacity – 120 MWs in Singapore and 110 MWs in other markets – by the end of its current FY27 fiscal year (which ends in March 2027), up from just 62MW in 2024. 

Nxera has also established a joint venture with Hitachi with a view to entering the Japanese AI infrastructure market and is set to broaden its portfolio into many other global markets by taking full ownership, in partnership with private equity firm KKR, of international datacentre operator ST Telemedia Global Data Centres (STT GDC) in a deal valued at $5.1bn and which is expected to be completed any day now. STT GDC has facilities in 12 markets, including Italy, Germany and the UK in Europe.

Importantly, Digital InfraCo also has a strong relationship with Nvidia, with which it announced the formation of an AI centre of excellence (CoE) earlier this year. 

Digital InfraCo is also home to RE:AI, a sovereign AI cloud platform that offers a range of GPU-as-a-Service, Model-as-a-Service, Token-as-a-Service and AI-as-a-Service products to enterprise and government users in Singapore, guaranteeing that data doesn’t leave the island state, large language model (LLM) processing happens locally and all transactions are audited, with Singtel claiming that its sovereign services cost about half of services that require cross-border transactions and workflows.

In addition, RE:AI it is forging key partnerships with the likes of France’s Mistral AI to develop services for specific vertical sectors in multiple markets, and AI data and memory infrastructure provider WEKA for the development and provision of sovereign AI services in Singapore and the broader ASEAN (Association of Southeast Asian Nations) region.  

RE:AI also works hand-in-hand with Singtel’s Artificial Intelligence and Data Analytics (AIDA) business unit, which centralises its AI R&D initiatives, and NCS, Singtel’s enterprise IT and systems integration services subsidiary that develops AI solutions for its customers.  

 Singtel says the RE:AI services are proving popular and that it has secured 12MW of power capacity to build what it claims will be Singapore's largest GPU cluster. The initial 6MW of capacity, for which 300m Singapore dollars (US$236m) of capex has been assigned, has already been sold and has a “time to revenue” period of between six to nine months. That deployment will have paid for itself by the start of 2030, after which the current contracted revenue streams will deliver about S$100m (US$79m) of EBITDA. 

Singtel regards this as a “fast payback” on its investment, noting that it is already negotiating long-term customer contracts for the second tranche of 6MW capacity, deployment of which will commence only once the operator has secured contracts that account for more than 80% of the capacity. 

By contrast, Nxera offers a “complementary [investment] return profile” as, by the nature and scale of its deployments, it takes about three years from the start of any deployment to get to revenue generation, while the payback (return on investment) period is between seven and ten years. 

However, RE:AI requires a tech refresh every four to five years, while Nxera’s tech refresh timeframe is about 13 years.    

Across the telco’s operations, Singtel says it is benefiting from a broad deployment of AI.

In Singapore it has reduced churn and improved customer satisfaction ratings as a result of deploying AI-enabled personalised customer engagement systems, while it is also claiming swifter recovery from network issues and a more efficient deployment of new applications thanks to the use of IT and network automation systems and “intelligent operations”. 

In Australia, it says Optus has improved workforce productivity as a result of making AI available to all employees, as well as claiming improved customer experience ratings and faster network fault identification as a result of AI processes. 

The NCS enterprise services subsidiary, meanwhile, has developed its own Sunshine.AI platform to improve productivity and develop specific AI applications for customers. 

- Ray Le Maistre, Editorial Director, TelecomTV