16
From the newsletter (September 9; sign-up or follow if you want it sooner): Telecoms is moving deeper into AI infrastructure as networks become programmable, distributed and intelligent – but the opportunity comes with awkward economics, new competitors and a blurring of the boundaries around what telecoms actually is.
Has there been a more dynamic – or dicier – time in telecoms? To whit: take a look at today’s stories, from news coverage and sundry press announcements. Because telecoms is moving closer to the centre of the AI action – more programmable, distributed, intelligent, indispensable – and yet most of the money and power (as we know very well) sits elsewhere. Let us present the corporate bullet points…
Infrastructure beneath infrastructure: Nokia’s “scale-across” argument is that AI clusters are spilling beyond individual data centers, making long-distance optical networks part of the compute architecture. Glenn Lurie makes the same argument from the commercial side: connectivity can no longer be an afterthought to compute, power and data-center construction. The network is the business – to get AI to work, at work. Like we said in our piece yesterday about exec migration between sectors, telecoms is not telecoms anymore – it is an interleaved layer, part of a programmable cloud-controlled stack, called AI, and leadership must be apprised of everything. Nokia had led.
More invisible, and more valuable: AT&T/Leo is a good example: fiber, 5G, and satellite are being presented as one connectivity platform. The customer doesn’t care about the network mechanics underneath – so long as the data finds the model, the model returns the insight, and the action is taken. (It’s an IoT story.) Equinix is pushing the same abstraction in a different part of the stack, with Fabric One automating connectivity across clouds, networks and AI infrastructure. We talked yesterday about Verizon’s big fiber build-out in the US (80m new miles of Corning-made glass fiber) to underpin a “single, converged architecture” to “unite mobile connectivity and broadband”.
Exposing new network capabilities: Vodafone’s quality-on-demand API lets businesses dynamically request particular network-performance profiles rather than simply buying connectivity. Lurie points to Open Gateway as the emerging common interface between applications and networks. Which looks like an important shift, ostensibly, from selling access to selling network behaviour. But what capabilities? Well, Deutsche Telekom is using pre-ISAC passive RF sensors on towers for live drone detection; 5G/6G will build on the service with ISAC. Again, a very different proposition from traditional connectivity – the network becomes a distributed sensing platform, as the story goes.
Intelligence creates a trust problem: The UK’s Digital Catapult is being funded to build a vendor-neutral assurance capability for private 5G and an auditing framework for AI-native telecom networks, including trustworthiness, explainability, auditability, and continuous monitoring (etc). In other words, the industry is moving beyond standards and capabilities, and beyond test and deployment logistics, to finally address serious-minded questions about performance and trust. It shows, at once, what a smart and mature technology 5G is for critical industries, especially in its private form – and just how hard it is to box-shift technology to an impossibly-diverse global economy.
Where does all of this leave us? Well, old network boundaries are blurring, and starting to dissolve. Satellite is another access layer; optical transport is part of AI infrastructure; metro interconnection is part of AI inference; mobile networks are becoming APIs and sensors. The old distinctions – between terrestrial networking, cloud computing, AI modelling, satellite systems – are increasingly difficult to maintain. And the economics, as ever, are awkward. Lurie points out the mismatch again: hyperscalers will spend $760bn on capital projects this year while global telco cap-ex will decline about 2%. Apples and oranges, but the meaning is clear.
Show me the money – say telcos; and that cap-ex will jump again. But the industry is being asked, urgently, to build critical infrastructure and capabilities for the first- and second-generation of cloud-era new-money makers to ride on top of. Big opportunities, but real and present risks as well. Application interfaces, priority services, private networks, integrated sensing, optical comms, satellite complements – all offer ways for telcos to move higher up the value chain, as most have promised since the early days of 5G. But each lever opens doors for new market competitors, new cyber risks, new infrastructure owners – and the traditional telco domain is already changed forever.
