Separating signal from noise: Data centres as growth drivers for telcos

Download Listen

STL attended Mobile World Congress last month, on a mission to evaluate the perceived role of the telecoms ecosystem in accelerating data centre capacity.

This article is an excerpt from our report: MWC 2025: A tech show, not a telecom one – feel free to contact me at joe.hurman@stlpartners.com if you’re interested in discussing this report, or our strategy consulting and market insights offerings, in more detail.

Key themes

Data centres were the dark horse of MWC25. While not traditionally an event where data centres receive significant coverage, this year’s congress marked a step change towards a more balanced split between innovation in connectivity and compute infrastructure – all fuelled by the excitement around how assets and capabilities can be monetised in pursuit of a slice of the AI pie.

MWC stands were demoing a wide variety of data centre solutions – from data centre and interconnect solutions for service providers, through to cross-connect and cooling for data centre operators.

While it is early days in the AI digital infrastructure investment race, it is clear a range of monetisation avenues are available to operators. These divergent monetisation pathways must be evaluated and prioritised imminently, to ensure telcos can step into the AI growth train before it leaves the station.

Key announcements

• Arthur Mensch, CEO of Europe’s AI model developer, Mistral AI, suggested that telecom operators should look to become data centre operators. Of course, he has a vested interest in saturating the market and driving down his future hosting prices, but it is notable nonetheless – especially when you consider the company’s strategic partnership with Orange.

• This is the same telco that was publicly doubling down on its data centre operations at MWC – both in analyst briefings and through the appointment of Nicolas Roy, former CEO of Totem Towers, to lead Orange’s data centre arm. You’d do well to argue that these are unrelated narratives.

• BT announced its Global Fabric platform has recently gone live with customer traffic – following in the footsteps of the likes of Megaport, Console Connect and Lightstorm with a network-as-a-service (NaaS) platform targeting enterprises with an increasingly complex hybrid cloud IT estate.

• SK Telecom displayed an array of solutions for data centres, including GPU-as-a-service (GPUaaS), modular data centres with Elice, cooling systems with Giga Computing and SK Enmove, as well as mechanical, electrical and plumbing (MEP) systems for data centres with Schneider Electric.

• Qualcomm unveiled it is testing wireless cross-connect, with the aim of enabling traditional tree-and-branch cross-connect to move towards a more flexible, dynamic mesh architecture, adding connectivity redundancy while enabling development in dynamic traffic optimisation.

Key takeaways

Operators scramble to understand how they can monetise sovereign AI

Many operators were evangelising their credentials and positioning to lead the charge in sovereign AI. However, exactly how many telcos plan to deliver this is unclear. Of course, leading a national technology market provides them with significant assets in connectivity, existing customers and brand reputation, yet operators have previously struggled to invest in the skills and the operating models which enable agile digital offerings.

There is an array of open avenues in sovereign AI beyond just connectivity for CSPs. This applies to those pursuing infraco and servco business models alike (e.g., operating data centres and GenAI content localisation, respectively), and the optimal path will depend on the assets, strategies and market dynamics at play for each telco. Truly techco operators will pursue both avenues – transforming towards being national market leaders in both AI compute infrastructure and services. Take the aforementioned SK Telecom as an example, as it laid out its ambitious service portfolio in full at MWC this year, ranging from data centre cooling to local language LLMs.

Monetise, partner or divest? Divergent strategies for operator data centre assets

Many telecom operators have sold off their data centre facilities in recent years, including the likes of Telefónica and Rogers. However, with a variety of operators (such as Deutsche Telekom, KDDI and Indosat) doubling down on data centres, as well as unprecedented investment into developing global data centre capacity, could now be the time when we start to see telecom operators successfully monetising their data centres assets for enterprise applications? Orange was certainly singing to this tune at MWC.

The answer will depend on two key factors – the market and the access to investment. Today’s prominent markets are expected to grow significantly. However, attention is quickly turning towards more nascent markets both on a global scale and within regions (e.g., FLAP-D consisting of Frankfurt, London, Amsterdam, Paris and Dublin; and the rise of Madrid, Milan and Warsaw). In these up-and-coming markets, operators may possess the advantage of existing relationships with prospective anchor tenants. By adding to the mix limited existing competition and potentially some degree of regulatory capture, they may be able to position themselves for first-mover advantage.

The consistent thread across most successful telecom operator data centre offerings is the importance of partnerships. Many telcos are entering the market through joint ventures with more established data centre players to ensure they are pairing their national knowledge of regulations, skills and partners with global data centre expertise – examples of this include Telekom Malaysia and Singtel’s Nxera; Jio, Brookfield Infrastructure and Digital Realty’s Digital Connexion, and Omantel and Equinix.

The growing dependency on flexible connectivity to enable hybrid cloud architectures

Enterprise connectivity providers combining a NaaS platform with a simple customer user interface, lifecycle customer support and a flexible, deterministic network continue to position themselves as relative success stories in the sector. Console Connect and Lightstorm are two prominent examples here and were happy to share the interest they’d had from the ecosystem into their services for enterprises. This desire for flexible enterprise cloud connect and interconnect will only increase with broader IT trends towards hybrid cloud and multicloud architectures.

This success, and the broader AI opportunity, is filtering down to the vendors to such operators. One prominent optical vendor, speaking off the record, signalled that it was strongly pivoting towards opportunities in data centre interconnect (DCI), in no small part due to the significant government funding available to companies supporting infrastructure development for AI.