Don't expect AT&T 'open RAN' to go multivendor anytime soon

It's 2026 and AT&T's switch from Nokia to Ericsson across the one third of its footprint Ericsson didn't already serve is well advanced. There's a smattering of Fujitsu radios to justify the open RAN label and a sprinkle of Intel-based Dell servers to support the virtual RAN claims. Otherwise, AT&T's network looks as Swedish as Abba.

Several years ago, this confluence of single vendor and open RAN would have been as awkward as a meeting between Biden and Trump. These days, people unashamedly lump them together, referring to "single vendor open RAN" as if it were the most natural thing in the world. Vendor mashups were never the objective, it seems. What operators wanted was the option of a vendor mashup, or perhaps just the option of a swap.

Wait, didn't they already have this, as evidenced by AT&T's early removal of Nokia equipment? Open RAN's new interfaces won't do anything to treat the pain of a big swap. At best, they will make a partial swap possible. AT&T might, for instance, decide to replace some Ericsson radios with new kit from a previously unused third party. Thanks to the new open fronthaul interface, it could synchronize these with Ericsson's distributed units and turn a shade more multivendor.

But how likely is that? Either AT&T or Ericsson (and possibly both) must already shoulder the costs of writing off Nokia gear that wasn't naturally due for replacement. Another early swap would add to the financial pressure on no-growth companies that between them have already cut tens of thousands of jobs to protect margins. For a reminder, since the start of 2020, AT&T has shrunk headcount by 97,300 jobs, nearly 40% of the total. With its 99,950 employees, Ericsson today employs 11,500 fewer people than it did when current CEO Börje Ekholm took charge in 2017.

On top of that, AT&T would then have to integrate and manage another vendor at some expense. It could always delegate those jobs to Ericsson. But this would presumably not be free of charge, and it would also leave AT&T just as reliant on the Swedish firm. A source who knows a thing or two about operating mobile networks says costs in the construction phase are split 50/50 between electronics and services. And if this were not just a tokenistic rollout covering a few sites, those costs could be substantial. Multivendor enthusiasts would potentially have to justify spending hundreds of millions of dollars to Pascal Desroches, AT&T's penny-pinching CFO.

One conductor for the orchestration

Shockingly, though, a hardware switch is as likely as it gets. In that department, Ericsson from the outset will enjoy the company of Fujitsu (radios), Intel (chips) and Dell (servers), even if they are frequently not around. In the realm of software, it is on its own. Besides providing the software that runs inside AT&T's central and distributed units (whether its own appliances or Dell servers), it is responsible for another potentially important layer called service and management orchestration (SMO).

Changing network software is not like abandoning a Disney+ service and subscribing to Netflix instead. It is typically licensed for a multi-year period, which could make early cancellation as expensive as dumping hardware. Swapping the whole central and distributed unit layer for another vendor in a huge nationwide network would be just as onerous as replacing a fully integrated network, according to Tommi Uitto, Nokia's head of mobile networks.

AT&T could, alternatively, introduce another software vendor alongside Ericsson for part of the footprint. But this would still leave it with the problem of additional licensing fees. And integrating a new software vendor would probably be harder than adding third-party radios. That's largely because so much of the network functionality and performance is determined by what happens inside those central and distributed units.

For this reason, the software vendor usually takes the lead on systems integration, whatever telcos might say about their own involvement in that process. Add radios and AT&T can just lean on Ericsson as an integrator. The reverse of this, with AT&T similarly dependent on the radio vendor after a software change, is probably not feasible.

The huge cost of developing systems-integration expertise – it accounts for one in three R&D dollars spent by Ericsson and Nokia, according to an AT&T estimate – may partly explain why there is a multiplicity of radio vendors in commercial deployments but just a handful of mainly well-resourced software players. Mavenir and Rakuten Symphony are perhaps the only new entrants, and neither's projects are in great shape.

Sticky mess

The other giant software obstacle to a multivendor future is that SMO layer. Ericsson's contribution here is the Intelligent Automation Platform it launched in November 2021. Anyone should theoretically be able to write a network app for the RAN Intelligent Controller (RIC) this platform houses.

There's just one problem. Ericsson's RIC caters only to non-real-time apps (rApps), cutting adrift software companies like Cohere Technologies that prefer the near real-time (xApps) variety. Cohere's Universal Spectrum Multiplier software has been lauded by numerous telcos as a potential capacity booster for less advanced radios already in the field, and trials have shown it does exactly what it says on the tin. Ericsson's number-one priority, though, is to sell more advanced (and more expensive) radios. We'll leave readers to ruminate on all that.

Other telcos aren't too impressed with efforts on SMO standardization, either. In a white paper published last October, Japan's NTT Docomo and UK-based Vodafone grumbled that various important interfaces between the SMO layer and other parts of the network remained immature.

The implication of network diagrams included in this paper is that the SMO, cloud platform and virtual network functions might all today have to come from the same vendor. Nokia's Uitto has alluded to the same issue. A vendor providing both the SMO layer and the software for central and distributed units has the opportunity to create "sticky functionality" between them, he previously told Light Reading.

If AT&T technology executives attending last month's Mobile World Congress in Barcelona were hoping to address the skepticism head on, and show why critics are wrong, they miserably failed, to judge by Light Reading's conversations with numerous senior executives at the event.

There are only two viable explanations for the AT&T move. One is that it was unhappy with Nokia, for whatever reason. In subsequent public statements, however, it has praised the Finnish company's technology and insisted its decision had nothing to do with that. The other explanation is that a perennially cost-cutting AT&T, heavily dependent on Ericsson already, saw the financial benefits and convenience of operating a mainly single-vendor network, and was egged on by generous Swedes. The argument this was about open RAN – as it was originally conceived – just doesn't hold water.