So two things, same as yesterday (satellite constellations and enterprise networks)… Firstly, Sulagna has delivered again; her piece on Starlink’s rise says Musk’s outfit is no longer a stopgap for rural broadband, but a strategic, potentially dominant, new global layer of telco infrastructure. The standard telco framing – that it is a complementary black-spot service; per the Deutsche Telekom tie-up last week – feels temporary. Musk’s play is scale (always scale), as we know; and Starlink is a vertically-integrated space-comms setup that rewrites geographic constraints and cost structures. In a capex-constrained industry, its disruption (always disruption) is totally strategic, where networks aren’t built ground-up, between trenches and rooftops, but projected down from free space – from up in the heavens, where, even as it gets crowded, the view is clear. 

It is total telecoms – like the networking equivalent of a Dutch footballing ideal. Secondly, physical AI (always AI), and how to make AI work at the edge. Factories want total connectivity – like the Industry 4.0 equivalent of the same – for sub-10ms inference workloads. Which means high levels of control and performance, and, most crucially, of predictability and reliability. Plus, these venues, microcoms of Industry 4.0, require deep OT integration just to keep their machines on the track. Which is infrastructure – servers and networks – as choreography; as total football, again. Without localized private compute systems, connectivity networks, and data fabrics, the system collapses – and the AI stops, the ROI fails, and the dream dies. As humanoid robots go from pilot to production (post-2028, surely, in time for a private 5G uptick?), the constraint isn’t the AI model, but the system. 

With apologies to Johan Cruyff.