Singtel’s RE:AI joins the ‘tokenomics’ crew

• RE:AI, the sovereign AI cloud unit of Singtel’s Digital InfraCo, is aiming to meet the expanding digital services needs of business users
• It has launched an AI token-as-a-service (TaaS) offering with a strong sovereignty pitch
• It is one of the first telcos to expand its AI services portfolio into the TaaS domain
• The debate as to whether telcos can capitalise on the trend towards ‘tokenomics’ is polarising

Can telcos make money by billing customers for AI token consumption rather than data usage or voice minutes? The jury is out, but Singtel clearly thinks so, as it has become one of the first communications service providers to launch a token-as-a-service (TaaS) offering via RE:AI, the sovereign cloud unit of its Digital InfraCo division. 

The Singapore network and digital infrastructure operator isn’t the first telco to try driving incremental revenues from the generation of AI tokens (the basic units of text, code or data that an AI model processes or generates), China’s three main telcos are ahead of the game in that respect – see More than a token gesture? China Telecom launches AI token subscriptions.

But it is certainly an AI-native telco pioneer with its TaaS offering, which the operator describes as “a fully managed service which provides enterprises with access to AI models, scalable infrastructure, orchestration and governance capabilities through a flexible, token-based consumption model”.

According to Singtel, its TaaS “provides enterprises greater control over their AI usage and cost through flexible subscriptions that allow them to select the token volumes they need,” while its models-as-a-service offering “enables enterprises to use a range of AI models without having to buy and manage them individually”. The operator also boasts that an intelligent routing feature built into its TaaS platform “automatically matches workloads to the most suitable AI models to optimise cost and performance. Applications such as software development, content creation, and voice and video services can be handled cost-effectively using leading open-weight models such as Qwen, Mistral, GLM, Kimi and MiniMax as well as selected frontier models when required. Enterprises can also onboard their own models to meet specific workload and business demands.”

At the heart of Singtel’s pitch is sovereignty – customer data remains within Singapore to help enterprises meet sovereignty and regulatory requirements – as well as simplicity and cost efficiency.

“As enterprises scale AI across their operations to improve productivity and efficiency, rising token costs and concerns about sending sensitive business data to AI models hosted overseas can hinder wider adoption,” states Singtel. “These challenges become increasingly important as enterprises adopt agentic AI, which can increase token consumption and the complexity of managing multiple models and AI capabilities. RE:AI’s TaaS solves these problems by hosting AI models on sovereign infrastructure in Singapore, powered by Singtel’s award-winning RE:AI sovereign AI cloud and integrated technology stack spanning AI datacentres, robust terrestrial and subsea connectivity and Paragon orchestration platform. This enables enterprises to balance cost, performance and scalability while maintaining data sovereignty,” the telco adds. 

However, Singtel has also recognised that data sovereignty might not always be a requirement. “For less sensitive workloads, enterprises can also leverage non-sovereign AI clouds to optimise cost and performance. TaaS provides token-level visibility into AI usage, enabling them to implement governance policies, guardrails and monitoring controls for responsible AI deployment.”

Bill Chang, CEO of RE:AI and Digital InfraCo, stated: “The AI opportunity is about scale. Enterprises are seeking a simpler, more integrated way to deploy AI across their organisations rather than procuring, integrating and managing separate AI models. Having embarked on our own journey to become an AI-first organisation at Singtel, we understand the practical challenges enterprises face. Through TaaS, we bring together our integrated technology stack and expertise in a single managed service, giving enterprises greater flexibility, visibility and control to deploy AI securely while optimising cost and performance. We want to help enterprises move faster and with confidence from AI ambition to business impact, accelerating digital growth and strengthening Singapore’s position as a leading hub for AI innovation,” he added.

Singtel’s move will be watched closely by telcos all over the world, especially those that have developed their own AI factories and are already seeking to monetise their AI infrastructure with offerings, such as GPU-as-a-service (GPUaaS). 

But will many follow in its wake? Is so-called ‘tokenomics’ a potentially profitable model for telcos? 

As you might expect, Nvidia, the major provider of the enabling technology (processors and software) for AI factories, is adamant that this is a great model for telcos, as it explained in this blog. In addition, vendors such as Huawei (see this announcement) and Mavenir, in partnership with Red Hat, are pitching themselves as companies that can help telcos develop TaaS offerings. 

But as TelecomTV’s Guy Daniels noted during the recent AI-Native Telco Forum in Düsseldorf, tokenomics is “somewhat polarising at the moment, a bit controversial. You’ve got real advocates and supporters of the approach, and equally you’ve got people who are very sceptical.” 

The topic cropped up during a number of sessions at the forum, with Yue Wang, China Telecom’s chief technologist of next-gen ICT, discussing her company’s approach to token-based subscriptions in this session focused on generating revenues from AI, while Mavenir’s chief technology and strategy officer, Bejoy Pankajakshan, also talked extensively about TaaS models during this set of presentations. 

One of the most vocal sceptics is renowned industry analyst Dean Bubley, who, as ever, declined to sit on the fence when writing about the topic on LinkedIn. 

“The idea that telcos must become AI ‘token factories’ or offer token-as-a-service propositions is getting tedious. A kernel of reality is being coated in multiple layers of absurdity,” wrote Bubley. 

Let’s see in the coming months how the approach plays out for Singtel, which is one of the 56 operators around the world being tracked and analysed for TelecomTV’s AI-Native Telco Index reports, the first of which has just been published – see New report identifies the AI-native telco frontrunners.

This TaaS launch will no doubt feed into the next index report, which is due to be published during the first half of next year. 

 - Ray Le Maistre, Editorial Director, TelecomTV