Telcos in AI infrastructure: Repurposing networks for growth

Listerine was conceived as a surgical antiseptic before it became the answer for bad breath. Bubble wrap was designed as a new home decor wallpaper before it became a material for the packaging and transport industry. Vaseline was an annoying ‘rod wax’ before it became a staple in almost every household as the magic healing jar. Similarly, INS Vikrant lives on after its glorious days as India’s first aircraft carrier in the curves of Bajaj’s V15 bike. At the same time, Bombay’s then-cotton mills left the space for Mumbai’s lofty and swanky shopping malls to rise and shine.

Several iconic movies continue to reap all the effort and scale invested in their shooting sets by repackaging them as amusement parks. Dr Joseph Lawrence, Alfred W Fielding, Marc Chavannes, and Robert Chesebrough did something ingenious: their products solved problems different from the original ones they had intended. Products that live on even today.

Get the drift? Yes, you guessed it. No better time than now for telcos to utilize all their legacy muscle, infrastructure, and capacity in powering the gigantic compute farms that are super-busy growing Artificial Intelligence (AI). This is not only because the AI industry needs all this to grow and glow, but also because it is a good business opportunity for telcos who are disoriented and deflated with knocks from all sides; dwindling margins, falling revenues, new transition pressures, and hard-to-fight new species of competition. All the unused dark fibre aside, their experience and Capex of many years can come in very handy for pumping all the infrastructure oil into these new AI tanks.

Telco Brawn: AI Brownie Points?

AI is a creature that is growing like a hormone-charged teenager right now. It is gaining height, stubble, pimples, cravings, and the glimpses of adulthood as it sleeps every night in those grids and factories. Its appetite is ever-expanding. That explains all the massive investments. Remember Trump-Elon’s USD 500 billion AI infrastructure project, announced in January of this year? Stargate or Microsoft or Meta or Google, the need for massive AI build-outs is not ceasing significantly, despite conjectures that the AI bubble is finally dipping in enthusiasm.

A McKinsey report on AI infrastructure highlights how telcos possess a vast geographical reach and the ability to manage large-scale networks and variable demand, enabling them to meet the growing appetite for high-performance compute and connectivity driven by Generative AI (GenAI) and agentic AI applications.

AI data centres can gain a lot from all the fibre lying with telcos. Even if hyperscalers and AI giants build those colossal data centres needed for AI, they will have to turn to someone for connecting these tanks to scale and deliver their fruits to users and markets. This can easily translate into a USD 30–50 billion global market opportunity. The telco capacities can connect these massive AI capacities or end users, along with the space and power to host them. With configurable network solutions—the next generation of software-defined networks (SDNs)—telcos can also efficiently manage requirements for different AI workloads running over the cloud, McKinsey observed.

Telcos can provide managed services around AI, particularly distributed and edge AI services, to meet specific industry and user needs.

Examples like Verizon’s agreements with Google Cloud and Meta for network infrastructure for AI workloads, Lumen’s handshake with Microsoft for dark fibre, Verizon’s edge-computing deal with AWS, Indosat Ooredoo Hutchinson, Singtel’s moves in tapping the Graphics Processing Unit-as-a-Service (GPUaaS) market, or Telenor’s launch of a sovereign AI for the Nordics through NVIDIA prove that the opportunity is a real one.

What is worth remembering here is that many telcos missed the Cloud bus, and that is where AI grids and factories can be the trains they should make sure of catching up, by building and sharing the railroads they need.

John Strand, CEO of Strand Consult, confirms that many telecom operators opted out of the opportunity to enter the cloud market. In several countries, telecom operators chose to divest their hosting centres, which were the first generation of cloud solutions. Today, it is market-dominated by hyperscalers and players who often market themselves as domestic players that host local and often-sensitive data for governments, the financial industry, and others.

“I believe that AI will not be a separate product, but a product that is integrated into IT solutions that hyperscalers, including Microsoft, Google, and AWS, deliver to their customers. In India, there will be a market for large IT providers that provide solutions where AI is integrated. Some of these players will also sell their solutions outside India. The question is how much AI they will develop and how many solutions will be delivered in collaboration with some major AI suppliers.”

Ask Naresh Singh, Senior Director Analyst at Gartner, who avers that growing tech sovereignty requirements and compliance needs for data privacy and security have opened opportunities for new cloud and AI service providers. “When it comes to AI infrastructure, there is also a supply-demand gap, which has led to the emergence of new players like GPUaaS providers. Some telcos have also decided to invest in this market, but most continue to wait and watch.”

Telcos must focus on specific opportunities, such as cost-optimised AI inferencing services, and try to differentiate through an end-to-end offering.

Global AI infrastructure is still at a nascent stage, which provides a real opportunity for this worldwide, points out Singh. “Also, the technology is relatively new, compared to traditional cloud technologies. This means the gap between players is still not so wide, and hence, a new player like a telco could catch up by making smart investments by rapidly developing capabilities or acquiring interesting startups in various layers of the AI stack to become more competitive.”

He further pointed out that telcos should try to seize the opportunities offered by emerging economies’ tech ambitions and global geopolitical changes.

Strand is chuffed about India’s prospects here: “India is a unique case. There are three large telecommunications companies and a forest of successful IT companies. One can expect several Indian partnerships in the AI space. Such partnerships are hardly seen in Europe,” he said, adding that India is a dark horse of AI. “The country has significantly better opportunities in AI than a country like China.”

Square Plug in a Round Socket?

Will it be attractive? Yes, for sure. Will it be easy? That is a big question.

There are many approaches to this new industry. As Singh suggests, telcos can leverage these markets by partnering with the key players in the market, like the hyperscalers and the GPUaaS providers, to provide connectivity to scale the AI services across the world as AI inferencing services emerge as a major engine of growth for AI in the enterprise. “They can also provide managed services around AI, particularly in the case of distributed and edge AI services to meet specific industry and user needs.”

Telcos may pursue these in a big-bang way, in a phased manner, or take the pilot approach. This depends on their risk appetite, resources like financial, technical, talent, and collaborations, competitive landscape, and future goals, says Dr Nityesh Bhatt, Director, Dean, and Professor of Information Systems at the Institute of Management, Nirma University. “While this AI space is evolving fast, many global best practices are available from within and outside the industry that can provide some direction.”

Prashant Ramesh Malkani, Head of Network Infrastructure Business at Nokia India, advises that telcos could partner with hyperscalers to offer cloud-native networking solutions and productise private cloud connectivity and hybrid cloud orchestration. “Telcos could deploy dark fibre and AI grids to power AI compute centres and offer AI-driven network automation and predictive analytics as a service,” he elaborates.

Telcos could deploy dark fibre and AI grids to power AI compute centres, offering AI-driven network automation and predictive analytics as a service.

He further explains that an AI Grid is a distributed network of AI compute resources connected via high-speed fibre and edge networks. AI grids distribute compute across fibre-connected edge nodes to reduce latency and cost. Telcos can monetise infrastructure by turning networks into AI-powered services.

Strand concurs with that route. “Partnerships, partnerships, and partnerships,” he says, highlighting that Indian operators do not have the skills to develop, market, and sell AI compared to large IT suppliers and hyperscalers.

This new path, though, involves many possible setbacks and challenging questions. It will not be as simple as changing tracks, especially when the tracks are made for two quite-not-so-similar trains.

There is also the angle of the GPU ecosystem to consider. GPU cloud providers are currently in a strong position. Demand for GPUs is high; some hyperscalers have issues sourcing them, and global economic conditions mean enterprises are desperate to squeeze costs. However, supply and demand can change rapidly. GPU cloud providers essentially sell access to a third-party product. “They are not differentiated by any specific features or capabilities,” points out Dr Owen Rogers, Senior Research Director of Cloud Computing at Uptime Intelligence. “These providers are trading in commodities that may be valuable now, but whose value is susceptible to change. In the longer term, the future of these providers is uncertain. Hyperscaler cloud providers are protected from changing AI supply and demand, as they have diverse product portfolios and economies of scale. Smaller, GPU-focused providers do not have such protection.” Telcos should consider these factors while mulling potential AI opportunities and partners.

One of the challenges is that AI investment in infrastructure (Capex) is not a one-time phenomenon but requires sustained Opex too. Dr Bhatt reminds that competency development is a recurring effort that needs top leadership and HR support.

As the McKinsey report cautions, delays or missteps may heighten operators’ risk of falling further behind hyperscalers and other new market entrants. These have been the primary beneficiaries of growing data consumption over the last decade, while telco revenues remained largely stagnant. GPUs could also run the risk of becoming easy commodities, and there could also be risks of excess capacity piling up in the industry.

Singh recommends that telcos that have entered the cloud and AI services arena build their offerings around specific regional needs that hyperscalers have challenges meeting. “Also, they need to focus on specific opportunities such as cost-optimised AI inferencing services and try to differentiate through an end-to-end offering including AI-relevant managed connectivity and security services.”

“For telcos, it is also about leveraging their strengths, such as network assets and legacy customer relationships, to solve complex customer problems in their AI journey, particularly in edge AI. They need to accelerate their investments in AI technologies and skillsets.” Singh notes.

Uncertainty and risks dominate this route. As in the dot-com bubble, some GPU cloud providers will thrive, some will survive, while others will perish, contends Dr Rogers. Telcos should ask themselves about the value they are adding by offering GPUs. If the only value is in matching supply to demand, it may not be a risk worth taking. If the value is in a better customer experience and a differentiated AI-integrated portfolio, offering GPU capabilities might provide a return on the substantial investment.”

Repurposing is great, but it has to be executed smartly, strategically, and well-calibrated. Let us not forget that our kitchen sink favourite, Lysol, was once marketed as a female hygiene product. Listerine or Lysol—it is up to the telcos now.