You’ll know lots of this already, but it is interesting to see it written down. Open RAN is becoming more about “interface standardization rather than supplier diversification” – writes Dell’Oro Group. That was not really the original pitch. Break up the RAN, standardise the interfaces, and let operators mix and match components – for cost and performance, to loosen lock-in; to let the light in so smaller vendors can innovate, legacy carriers can flex, and the whole industry wins. Work by the O-RAN Alliance on ‘open fronthaul’ has provided that flex between the radio unit (RU) and distributed unit (DU). But opening interfaces does not automatically make networks easier or cheaper to run. A multi-vendor array across 20,000 sites, say, creates integration work – which tilts complexity, certainly, and cost, possibly, the other way. And so operators have been less inclined to change suppliers – lock, stock, and barrel.

Which is one reason the big RAN vendors have retained so much of their positions. Operators may want an open architecture; but they definitely do not want to be the systems integrator for such rangey estates. Dell’Oro Group says open fronthaul adoption “continues to gain momentum”, albeit “more gradually than anticipated”, and will be the “preferred interface” for “next-gen RAN platforms” among non-Chinese RAN suppliers. But “most deployments will continue to use radios and basebands from the same vendor”, it says. Cloud RAN, about where the RAN runs, is another matter, it says. RAN is not an ordinary IT workload, and purpose-built hardware is good with all the performance, latency, and synchronization. Move that onto general-purpose CPU/GPUs and you might also pay in power and compute overheads. Hence Dell’Oro’s downward revision to its cloud RAN forecast.

It is not whether the RAN can be virtualized, but whether it can be virtualized economically. Of course, AI-RAN might change the calculation – if operators are going to deploy accelerated compute for network optimization and edge inference services, then utilization looks better. But 6G does not necessarily change it. Operators have sprawling estates, always in flux, besides any new-fangled RAN: spectrum, fiber transport, operational back-ends, power systems. Choices about the open RAN mix must be taken in-the-round. And open RAN has had a high bar, anyway. Ericsson, Huawei, and Nokia kit is generally good; operators know how to run it. If it costs more to integrate, consumes more power, or introduces other headaches, then the business case gets difficult. Which doesn’t mean open RAN has failed, of course; it has just been most successful as an interfaces-and-standards exercise.