All telecom is going to the cloud, right? Well, maybe not so right. Microsoft, who has probably accumulated the best software technology for telco applications, is pulling out of the telco application space in favor of supplying only infrastructure, which means that they’ll sell Azure hosting to telcos if they want it—surprise, surprise. This article from Light Reading outlines what’s happening.
Microsoft’s partners in Azure for Operators, Affirmed and Metaswitch, have customers before and through their acquisition by Microsoft, but it doesn’t seem to be nearly enough for Microsoft to stay committed. I had comments on the Microsoft move from 18 operators, and they all said that while they may well host some elements of service infrastructure, mostly but not exclusively 5G, they’re not interested in using public clouds as the basis for the majority of their mobile deployments. This is a significant shift from 2022, when most said they were considering that very step. What’s changed?
The biggest factor, according to the 18 operators, is recognition that 5G won’t really create positive ROI. You probably recall that I never believed that was likely, but operators hoped and were encouraged in their hope by media, analysts, and vendors. When hope was finally dashed, operators recognized that buying into cloud provider service for 5G posed a risk that their whole 5G deployment would be seen by the Street as a loss. So…cut costs by cutting margins.
Related to this is the loss of credibility for the Open RAN paradigm. Open RAN was a justifier for cloud provider telco services, both because it promised a rich set of hosted features to drive revenue and because it required some hosting resources as open as it was. Given that Open RAN has failed to produce new features/revenues, and that operators have been pulling back from openness in mobile infrastructure generally, this hurts cloud hosting of 5G.
The third point is that telcos are seeing the cloud network increasingly as competition, and don’t want to boost its breadth and credibility. As cloud-hosted front-end components take more and more of a role in applications, and as SD-WAN and SASE provide cloud-to-VPN connectivity, operators see the cloud’s own network threatening their VPN business, which on the average is a bit more than a quarter of their revenue and nearly a third of their profit.
The big question, of course, is what this means to telco services and 5G/6G overall. Answer: A lot, but the industry may not see or admit to it all.
The first point is that traditional mobile service vendors are the only ones who could expect to profit much from 5G. Microsoft’s Azure for Operators was an integrated open offering, and neither Microsoft nor any other cloud provider has made that role successful. That almost surely means that any open-model 5G will require the operator to take integration responsibility, which operators have almost universally told me they don’t want to do. I think the mobile infrastructure giants recognized this from the first, which is why they jumped on the open-model bandwagon. Openness is a cure for a recognized risk, not a vaccine against a general one.
The second point is that private 5G is another delusional step and not a real opportunity. Yes, there will be some of it, just as there was with 4G LTE, but it’s not a big enough space to move the needle. The cloud providers, as Microsoft has shown, already realize that there is little point to trying to sustain a 5G service position justified by private 5G alone.
The third point is that hype hurts, even players who should know better. There’s nothing wrong with riding a hype wave; it’s smart PR and smart from an investor perspective. If the media is telling an attractive fable, erect a giant billboard that buys into the story, but don’t spend your money to buy technology that relies on the fable becoming reality unless you know how to drive that transformation and are prepared to do the heavy lifting. Microsoft and the other cloud providers just didn’t read the telco market properly.
Point four might be the most important point: the telco market, because of its dependence on long-cycle standards, is inherently more susceptible to hype. Look at both 5G and (now) 6G. We have standards activity almost a decade ahead of any actual product or service. The media and analyst communities (increasingly related these days) are always looking for something new, and the fact that they find it so far in advance means that there’s plenty of time for exaggeration to grow into wild speculation, and yet not enough information for any player in the space to make realistic assessments.
If we look at 5G based on what we know today, we might well conclude that the biggest impact on services it would have is FWA. Fixed broadband serves devices and applications that actually require the higher bandwidth 5G supports, but at the same time they don’t require the mobility features. Might we have focused on establishing a two-model 5G model, one where the service offers full mobility support and one where it assumes fixed endpoints? Even as a replacement for WiFi using public spectrum, an application already suggested for 5G, would make more sense if it were based on a simplified operating mode.
What does this mean for cloud providers themselves? Potentially a lot, but with stress on the qualifying term. In theory the telecom market could be huge for them, and in theory telco reliance on public cloud could reduce the risk they’d enter the cloud/edge hosting market themselves. Yes, the former hope was unrealistic and the latter unnecessary given telco reluctance to stray out of basic connectivity services, but certainly Wall Street may voice displeasure at this news. But Microsoft’s Street creds come from AI, and it may be that the Azure for Operators thing is just a distraction.
And that may be what’s behind all of this. If selling to telcos was in part to build Street cred, and if AI does that better, then why bother with the telco story, particularly if there are solid reasons why it might never actually generate much revenue. Telco sales is demonstrably aren’t going to do much for cloud provider bottom lines. Maybe AI won’t either, in the long run, but that’s not demonstrated yet. Better an unproven asset than a proven failure, I guess.
