By Annette Ikponmwonba | November 10th, 2026
Key Points
· The GSMA says telecom operators can unlock new revenue by helping businesses and consumers manage, access and optimise AI services, moving beyond traditional connectivity
· Tokens are becoming the unit used to measure AI consumption, much as voice minutes and data gigabytes were for earlier services
· GSMA Intelligence identifies three areas of operator AI expansion: infrastructure, models and solutions, with solutions drawing the strongest participation from over 200 operators globally
· China is an early test market, with average daily token usage rising from about 100 billion in early 2024 to 140 trillion by March 2026
· The “control layer”, covering model aggregation, intelligent routing and workflow optimisation, is flagged as a more defensible revenue source than simply reselling tokens
· The analysis warns the window is narrowing as hyperscalers, cloud providers and AI-native companies move quickly
Main Story
Telecommunications operators could unlock new revenue streams by helping businesses and consumers manage, access and optimise artificial intelligence services, moving beyond traditional connectivity offerings, according to the Global System for Mobile Communications Association. The body said growing adoption of AI-powered applications, agents and automated workflows is creating an emerging market for operators to monetise AI consumption, with tokens increasingly becoming the unit used to measure and manage interactions with AI models.
In an analysis published on its website, Kenechi Okeleke, Director of Regional, Social and Policy Research at GSMA Intelligence, said operators have largely focused on using AI to improve internal efficiency, but the technology is opening up opportunities to earn revenue directly from customers. He explained that, just as operators once monetised voice minutes and mobile data, AI is introducing tokens as another measure of digital consumption. Unlike gigabytes, however, a token’s value depends on the AI model, the complexity of the task and whether it represents input, output, cached context or reasoning.
Okeleke said operators that can simplify these complexities and package services around customer needs will be better placed to benefit. The analysis found operators expanding at three levels. At the infrastructure level, they are deploying data centres, cloud services, fibre networks and edge infrastructure to support AI factories, sovereign AI platforms and GPU-as-a-service offerings. At the model level, some are building or partnering on foundation models suited to local languages and market needs. The solutions segment has the strongest participation, with more than 200 operators globally offering enterprise copilots, conversational AI, productivity tools and industry-specific applications. Asia-Pacific accounts for 43 per cent of operator participation at the infrastructure level and 38 per cent at the model level.
The Issues
China offers an early picture of what commercialisation could look like. Average daily token usage there rose from about 100 billion at the start of 2024 to 140 trillion by March 2026, driven partly by AI agents and coding assistants. The country’s three major operators are testing different models: China Telecom bundles AI model access with connectivity and cybersecurity, China Mobile combines tokens with cloud PCs and AI agent capabilities in regional packages, and China Unicom is exploring a credit-based model for managing token use across multiple models. Okeleke cautioned that China’s experience may not transfer directly, given its large cloud businesses, heavy AI infrastructure investment and broad ecosystem of open-weight models.
The deeper question is where operators can earn durable margins. Simply reselling tokens risks turning telcos into low-margin intermediaries. The analysis points instead to the control layer, which helps businesses manage AI across models and providers by selecting suitable models, routing workloads, controlling costs and meeting governance requirements. It describes three stages: model aggregation through a single interface, API and billing system; intelligent routing based on cost, quality, speed, availability and regulation; and workflow optimisation that turns AI adoption into measurable productivity gains.
Competition is the other pressure. Operators face growing rivalry from hyperscalers, cloud providers and AI-native companies already building platforms to control access to AI services. Operators do have advantages, including customer relationships, billing and identity systems, security and compliance expertise, local regulatory knowledge and distributed infrastructure, but these will not guarantee success on their own.
What’s Being Said
Okeleke argued that operators must find their position in the AI value chain, build relevant capabilities and form partnerships without becoming low-margin resellers. He urged them to prepare networks for the changing demands of AI traffic, including potentially higher uplink requirements and more distributed processing, and to use network APIs and edge capabilities to differentiate their offerings.
He cited examples of the control layer’s value. China Mobile’s MoMA platform can reduce token costs by 20 to 30 per cent through intelligent routing, caching and context reuse, while AT&T has reportedly saved up to 90 per cent through its internal AI gateway. His argument is that operators can prove the commercial value of such platforms internally before offering them to customers.
He also issued a warning. The opportunity to capture a meaningful share of the AI economy is narrowing as technology companies move quickly to establish positions, and operators that fail to act risk remaining connectivity providers while others controlling the intelligence and service layers capture more of the value.
What’s Next
Operators are expected to test different commercial models, from bundles to credit-based token systems, with China’s three major players serving as the main reference point. Others will need strategies shaped by their own infrastructure, customer base and competitive environment, Attention will turn to whether operators in markets such as Nigeria and the rest of Africa move into AI aggregation and routing services, and how quickly they can build partnerships before hyperscalers and AI-native firms consolidate the market.
Bottom Line
The GSMA is telling telcos that the next revenue frontier lies in managing and optimising AI consumption rather than just carrying the data, but with the window narrowing and rivals moving fast, operators that stay connectivity providers risk watching others capture the value.
