There’s something in here about the promise and peril of this whole AI story, and of quite how critical infrastructure, including telco networks, has become. The lead story – a long interview with Microsoft from MWC – is a good measure of the market’s progress: a handful of ‘frontier’ firms embedding hundreds of agents across their operations, and a killer use case with Vodafone about shortening RFI/RFP response times for enterprise clients from weeks to minutes. That is a lot of freed-time, right there – and a clear ROI, assuming the time savings can be measured in money terms. 

But all of this has to work, or the house comes down. Telcos need AI to work – for new growth, value, relevance. As we know. So do data center companies, and everyone associated with the AI build-out – or so they will, at some point, once the ‘factories’ are built and the models are trained, and AI does some proper work inside enterprises. But so does every industry, actually – for all the same reasons. In the end, AI needs AI to work – if its promise is not to go pop. There is a lot riding on this, as we see; it is a fine balancing act.

Meta is deepening its AI investments and cutting its workforce at the same time – which is about cost pressures, talent realignment, and probably total economic overhaul. External shocks, just reported today, show how politics and infrastructure are now inseparable from AI strategy – global geopolitics, zeroed in on the Middle East, have stalled the biggest submarine cable project in the world; France is preparing a nuclear power push for sovereign AI data centers; Telia has a big sovereign AI project of its own in Sweden. 

This is global and critical, and it all feels dicey. Even Microsoft’s hopeful AI hit parade with Vodafone et al is almost upended by its talk at the end about the challenge to make the tech scale. Enterprises are “horrified” right now, it says, that AI agents will swarm and riot. How can they be identified, managed, controlled? If not sorted, fragmented data, inconsistent semantics, and weak governance will trap some AI deployments in pilot purgatory (best case), and, at worst, erode trust in the whole AI economy – hurting those who built it, and also those who bought it.  

It starts to sound like IoT. Except there are two differences. One: the whole world is watching this one; more than that, knowingly or not, everyone is invested in it. AI is totally critical – in terms of top-line economic money-flow today, and in terms of bottom-line political and economic power tomorrow. Two: the hyperscalers are in charge; or at least their models, about scalable elasticity, represent the commercial pole, and architectural north star. Of course, the hyperscalers messed up (kind of) with IoT, and (some) telcos have failed with private 5G – critical enabling tech, too.

So, the momentum is undeniable, but so are the risks. But AI is bigger and badder (interpret as you please), and just much easier to know. Therefore, right now, writing this, logic seems to say that the governance and scale challenges will be managed – thanks also to open networking initiatives in telecoms. Data will be controlled and orchestrated between countries, enterprises, technologies, and AI will be used for… everything (have mercy). Happy Monday, everyone.