What a brilliant and complicated little market. There is a clear sense – for RCR, anyway – from (finally) reading the latest (long) list of private 5G triumphs, as periodically catalogued by SNS Telecom & Research, that this is the best version of telecoms out there. More than that, it is the sector, troubled as it is, that shows the rest of industry what it might be – and, indeed, what it wants to be: connectivity plus compute, connectivity plus AI, connectivity plus security, connectivity plus control, connectivity plus automation – connectivity plus outcomes, ultimately.

A digital change platform for the industrial economy, in other words – where the industry’s best ideas (5G-Advanced, open RAN, AI-RAN, agentic AI, network APIs, NTNs, ISAC) are put to work in flexible ways (private, hybrid, neutral-host) to connect clever tools (software, compute, analytics) in scattered infrastructure (edge to cloud, local and global) for crucial assets (workers, sensors, cameras, vehicles, machines, robots) to deliver critical gains (uptime, innovation, safety). It is telecoms in miniature – and telecoms as it talks (constantly) about wanting to be.

We will write more about this tomorrow; but know that all of these technologies and triumphs are contained within this market. Meanwhile, in the big market, Nokia (as reported here) and Ericsson (as written yesterday) are both trying to sell versions of this future – but from different ‘AI-era’ start points. In the red corner: Ericsson’s latest results showed the discipline required to survive the current cycle: strong margins, careful cost control, and some progress in enterprise and cloud software, but with mobile revenues still constrained by drawn-out 5G SA investments.

As discussed, its problem as well, to an extent, is that a short-term run on AI componentry from the AI build-out behind the access network is putting pressure on its own supply chain costs – while it waits for the AI investment cycle to catch-up to the mobile network. In the blue corner: Nokia’s AI-RAN announcement today – a 2027 commercial launch, a 2028 efficiency target – seeks to make the 5G radio access network a direct part of the AI infrastructure stack. It is a brand new vision, and Nokia (and Nvidia) basically own it.

So the slight criticism, that AI-RAN, and particularly GPU-accelerated AI-RAN, won’t actually swell the RAN market might not matter so much for the Finnish firm – if it is hoovering up lots of the early experiments. But big operators, unlike smaller enterprise-ones, have been cautious about every major 5G-era promise – from APIs to slicing to edge computing. So good luck; timing matters, and telco timing is sloooow – and justifiably so, given its returns and responsibilities; and, actually, like most mission-critical industries at scale.

Which is the contradiction at the heart of telecoms right now. The industry has never had a clearer vision of what it wants to become. The challenge is turning that vision into a business case that operators can actually afford to build – which is what the private 5G sector is doing (even without Nokia, perhaps) – one radio, one application, one enterprise, one industry at a time.