Fireside chat on 5G, platforms, and beyond

The industry is vastly undervalued, and there is ample value hidden in it. Every ten years a new mobile generation arrives and consumes an enormous investment cycle, and every time the same question goes unanswered: what gets monetised beyond capacity and efficiency. Alongside it copper revenue declines toward zero while its cost stays, and fibre demands capital precisely when money has stopped being cheap. Europe is fragmented by regulation into players too small to use the assets they own, and an underutilised mobile network makes no sense to anyone. Consolidation is the structural answer, and private equity returning in 2024 will help, because scale is the game.

Europe is not generally behind on 5G. Two things explain the pattern better than culture. If you were early on 4G and the return never came, the incentive to jump again is weak, though 5G still helps wherever capacity is constrained. And nothing happens before a licence exists. In Sweden ours came very late, so until it arrived we could run proofs of concept and nothing at scale. Once awarded, we took good spectrum across the footprint and led.

Focus decides the rest. Our mission as connectivity providers is ubiquitous, safe, reliable and affordable connectivity, and there is real risk in fragmenting into territory that is not our own. Plenty of operators tried media and television; few succeeded. Growth sits close to the core, and in B2B most of all: secure connectivity into the data centre, SD-WAN, security services, enterprise 5G networks across manufacturing and transportation. The valuations we see in the tech industry, and now in AI, all rest on secure digital infrastructure that we build, and we capture too little of it.

Greenfield and brownfield look like opposites and run on the same principles. Starting new lets you work from first principles and pick best of breed. At Reliance Jio the discipline that mattered was refusing to build for the past: we launched without circuit-switched fallback and without national roaming, accepting a smaller addressable base to avoid carrying yesterday forward. Three things transfer. Experimentation and fast learning, against an industry used to cycles of half a year and longer. Building forward rather than backward, because backward compatibility is comfortable and it drags. And people, because technology does not drive transformation, people applying technology do, which takes a mindset shift, hard training and tolerance for failure that is not repeated twice. Greenfield becomes brownfield quickly, so continuous improvement applies to everyone, hyperscalers included.

Network and IT converge, with care. It is all software, and containerisation, DevOps practice and APIs apply on both sides. Network workloads still carry latency, reliability and scalability requirements that classical IT does not, and parts of the industry jumped the gun putting them into fully containerised environments and got burned, which made the whole transition slower than a planned approach would have been. Convergence arrives, and until it does the two domains need their own expertise.

Two practical points close it. Processes need an owner by name. Most order fallouts trace back to a free-text field at acquisition, an address that cannot be fulfilled later, rather than to any system. Give the acquisition process an accountable owner and the entry gets fixed, and with a digital identity at the front there is no manual entry at all. Second, AI needs a D in front of the As: digitise end to end, then automate, then add analytics and intelligence. A brownfield estate hides far more manual work than anyone expects, and AI over broken data returns confident nonsense. On build against buy, AI is becoming the centre of operations, so invest in your own capability and prefer a platform to a point solution. The point solution is cheaper first and more expensive at scale, because it never connects to anything.