Monday (telco diary) | Network platforms and an AI hill of beans

0
1
Monday (telco diary) | Network platforms and an AI hill of beans


From the newsletter (sign-up if you want it sooner): Different moves from Liberty Global and Vodafone, the NTIA and AT&T, NTT Docomo and Ericsson, plus TIM, show how telcos are reorganising around AI. Wrap-around commentary from Verizon and Nokia, and you have a future view of the market – about smarter network platforms for the AI era. 

What you need to know today, in terms of headlines, and more…

Liberty Global has completed its acquisition of Vodafone’s 50% shareholding in VodafoneZiggo for €1bn in cash, plus a 10% equity interest in the newly-formed Ziggo Group, which will hold Liberty Global’s interests in VodafoneZiggo in the Netherlands and in Telenet in Belgium and Luxembourg. Which chimes, opposite-ways, with last week’s discussions about the UK firm, which has sewn-up the takeover of JV-partner Three UK in its home market, and is a prime example of a tier-one telco in the late stages of ‘getting it house in order’ for the new AI era – of stripping back non-core operations to scale-up its core ones.

Word from the NTIA is that the US has put a big chunk of the ‘4.4 GHz band’ (4400-4940 MHz), currently used by the military, up for review for future 6G services; the 1.6 GHz band, 2.7 GHz band, and 7 GHz bands are being reviewed for the same in the US. Which might be taken with news last week of AT&T closing its $23bn purchase of 30 MHz of 3.45 GHz and 20 MHz of 600 MHz spectrum from EchoStar, and in the context of speculation about SpaceX/Starlink building a terrestrial mobile network to go with its D2D services – if anything related to the firm can be dismissed so easily as speculation. Richard Haas doesn’t think so.

Meanwhile, NTT Docomo has picked a baseband server platform from Ericsson to manage its 4G/5G and AI RAN workloads in Japan – for optimizing network traffic and radio performance, predicting faults, lowering energy usage, even (eventually) running agentic RAN apps. Ericsson claims two-times the processing capacity and less than half the energy consumption. It uses Ericsson’s in-house ASICs. The deployment started last month (July). Docomo has been an Ericsson customer for years, but it is a shouty piece of business nevertheless – and evidence that operators are starting to invest in hardware capable of supporting AI-native RAN.

What else? Well, if these are all signposts on the road – Liberty and Vodafone reorganizing portfolios; AT&T getting a headstart on a new glut of spectrum; Docomo upgrading RAN machinery – then Sean Kinney has a couple of write-ups from interviews and events that are far more interesting, actually, about how telcos go further on their journey, to weave themselves into the AI economy as inference moves to the edge. Verizon says the real asset for telcos is not access to commoditized AI models, but their operational intelligence built up over decades – the data, context and expertise needed to make AI agents useful as they cascade across distributed AI infrastructure.

Nokia, Red Hat and CoreWeave reckon the next challenge is coordination – connecting inference workloads and edge resources across domains. Meanwhile, the other story today (from last week) is that TIM has explained away its recommendation to shareholders to accept a takeover by state-owned Poste Italiane, on the grounds the offer is fair, and it would do well to combine with the postal provider’s digital platform – to pursue all the big-ticket telco opportunities with AI services and industrial customers. Go figure. But it is a reminder that smarter networks don’t amount to a hill of beans in this crazy [AI] world.

It takes trusted platforms and good relationships to make money from them.