AI will create large potential opportunities for telcos in connectivity and use case enablement. But telcos must manage expectations and prioritise their investments carefully.

AI is a major potential opportunity for telcos

AI, and GenAI in particular, has become probably the most widely discussed (and often hyped) focus of innovation activity and strategic thinking at telcos and in the telecom industry as a whole.

On the one hand, the interest relates to the role that AI could play (and is already playing, to some extent) in automating and optimising both the network and other customer- and non-customer-facing functions at telcos. On the other hand, AI in both enterprise and consumer markets looks set to drive a massive growth in demand for various types of connectivity, along with an expansion of data centre capacity, some of which could be provided from within the telco network.

Industry thinking about the first of these two potential benefits of AI has tended to focus on operating expenditure (opex) savings deriving from incremental automation, and streamlining of network and operational processes. However, optimised networks and enhanced customer service also have a role to play in driving top-line growth if they result in increased customer satisfaction, retention and wins.

If you are not a subscriber, enter your details below to download an extract of the report

The role of AI in driving demand for telecom infrastructure and services, on the other hand, relates more to the potential for revenue growth and service innovation, as opposed to cost savings. Essentially, this is because connectivity is a critical enabler for two types of AI functions:

• Foundation model and LLM building and training: where AI algorithms use machine learning (ML) techniques to analyse, understand, summarise and recognise patterns within vast datasets, as well as predict and generate new content and outcomes from them (hence, ‘generative’ AI). While LLMs are focused on human language (voice and text), other foundation models apply the same techniques to other types of more or less unstructured data including video, images, audio and network telemetry data. Small language models (SLMs) are also gaining greater importance within GenAI. These are essentially slimmed-down LLMs, with smaller datasets and/or fewer parameters, optimised for deployments at the edge and on device. The relative importance of SLMs versus LLMs is another factor that complicates the calculus as to where in the network AI-driven increases in traffic and workflows will mostly arise.
• Inferencing: where trained foundation models and LLMs are applied to understand new data deriving from ongoing events and processes of different types, and are used to make and enact decisions about what is occurring and how to respond to it. In the case of a security application analysing video feeds, for example, inferencing could result in the detection of a presumed intruder and activate the appropriate response. In many industrial processes such as automated manufacturing, inferencing can be used to analyse data from components and systems to optimise or adjust their functioning, and identify potential problems before they impact production activity.

The reason why these AI functions have the potential to drive revenue growth for telcos is that they involve the transfer and processing over networks of vast amounts of data. The opportunity for telcos is of three main types:

1. Providing connectivity infrastructure and services, and data centre capacity, for AI workloads. What we mean by data centre capacity here is providing the physical facilities (such as spare compute capacity in the RAN or in adapted central offices) in which AI workloads can run. In this case, the telco merely hosts them for third parties rather than providing the services enabled by those workloads.

2. Partnering to develop and support B2B/B2C AI use cases and services across multiple industry verticals, delivered across multiple networks and network domains. This could involve providing the associated compute, and AI services and functions (e.g., model building and inferencing) themselves instead of merely the hosting facilities for third parties to run their AI applications.

3. Co-developing more bespoke, specialised, AI-based use cases and services that involve both specialised LLMs and inferencing, and custom network infrastructure and service design.

These three types of opportunity relate to the trio of broad strategic models for telcos that STL Partners has elaborated on in its recent research: the infraco, servco and techco, respectively.

Investments in AI infrastructure, services and technologies by different types of telco

Table of contents

• Executive summary
  • AI is both a substantial opportunity and risk for telcos
  • Different AI investments are suitable for different types of telco
  • Recommendations: No-brainer versus risky but high-reward investments
• AI is a major potential opportunity for telcos
  • AI workloads will gradually migrate from central locations to distributed sites
  • Inferencing will eventually drive AI workloads to the edge
  • Some inferencing will still be carried out in centralised sites
  • It is too early to model accurately how AI will drive networking demand for telcos
  • Servicing real customer AI demand: Infraco, servco and techco plays
• Telco roles in AI: Focusing investments around internal capabilities and customer needs
  • AI connectivity and infrastructure provider
  • AI enabler and service provider
  • Techco – AI company
• Conclusion: Differentiated approaches are required to seize the AI opportunity

Related research

• The journey to a self-healing network: Intelligence, agents and complexity
• The telecom techco: The role of network cloud, automation and AI
• Finding value from AI, analytics and automation in the telco: Part 2
• Network innovation as an engine for growth: A manifesto

If you are not a subscriber, enter your details below to download an extract of the report

David Martin

David Martin

Senior Analyst

David Martin has specialised in telco cloud at STL Partners since 2016, writing numerous strategic reports on different aspects of the topic. He also originated STL’s Telco Cloud Deployment Tracker: a major database of commercial telco cloud deployments by leading operators worldwide. He is a telecoms analyst of around 25 years’ experience, having also worked for leading analyst firms including the now Analysys Mason and Omdia. David obtained a First Class degree in French and German at Trinity Hall, Cambridge, where he also pursued doctoral studies in French.