Ericsson retreats from hardware with 5G Advanced launch

For years, mobile phone networks were a collection of relatively dumb boxes, wires and chips that applied the laws of physics to connect calls and transport data. But they have started to resemble living organisms more responsive to external stimuli. Prod one in the future and it may automatically react. A human controller should be able to issue specific commands and watch the artificially intelligent network carry them out.

Ericsson calls it the "programmable" network, an idea the company has made central to its launch of 5G Advanced, the latest version of the newest generation. In July, the vendor showed off an "interference sensing" feature as part of its 5G Advanced portfolio. Piloted by Optus in Australia, it detects interference between cells and takes remedial action to boost performance. "It does up to 40 billion calculations per second," said Sibel Tombaz, the head of Ericsson's cloud and purpose-built 5G radio access network (RAN) product line. Optus measured the average increase in user throughput at 22%.

A host of new features was added to the lineup this week. With "outdoor positioning," a telco operating the "standalone" variant of 5G, which brings a new core network, could pinpoint the location of a user device to within 20 meters, down from hundreds on 4G, said Tombaz. "Device battery performance" introduces algorithms that promise to lengthen battery life for smartphone or VR headset users. Other features talk up energy savings, performance gains and differentiated types of network service.

The big pitch is that network operators do not have to buy new equipment to take advantage of these 5G Advanced features. That follows heavy cuts by telcos to capital expenditure in the last couple of years. None wants to rip out the 5G hardware it has recently paid so much to install. Calls for "software-upgradability" have grown louder. "We don't need any new hardware, no," emphasized Tombaz.

Software eats Ericsson

No hardware might sound like a problem for Ericsson, which has traditionally sold expensive boxes to upgrade networks. But it needn't be if each 5G Advanced feature can generate a profitable and recurring stream of subscription revenues, and that seems to be the plan. In the future, headline sales might be lower. Yet margins could improve.

This is certainly what Börje Ekholm, Ericsson's CEO, led analysts to expect on his company's recent earnings call for the third quarter. "We will be selling more software in the future and the software content of our products will go up," he said. "That will structurally help gross margins, but it may structurally hurt the top line a bit and we're in that transition. It is a bit hard to be specific now." There are already signs this is happening. Reported third-quarter sales dropped 4% year-over-year at the mobile networks business group. Yet its gross margin rose 9.4 percentage points, to 48.3%.

Convincing telcos to spend additional money on 5G, hardware or no hardware, is the challenge. Average revenue per user (ARPU) has tended to fall in many countries despite network advances, and complaints about the 5G return on investment have become routine. Ericsson's sales patter is likely to be more straightforward when there is an obvious cost saving it can highlight. "We feel operators are becoming much more opex-constrained," said Tombaz.

But what telcos really want are network features they can charge their own customers – be they enterprise clients or consumers – to use. There is, then, some alignment between the 5G Advanced launch and the mission to produce industry-standard APIs (application programming interfaces) that would expose network features to developers. Omdia, a Light Reading sister company, expects these API transactions alone to generate revenues of $8.7 billion in 2029, up from just $161 million in 2023.

That's a big deal for the likes of Vonage, the company Ericsson bought for about $6.2 billion in 2022 to function as the software platform intermediary between developers and networks. Last year, Vonage generated total revenues for Ericsson of about $1.5 billion. A sizeable share of that market for network API transactions would bring dramatic growth.

But $9 billion is not such a big deal for a global telco market generating annual revenues of about $1 trillion. The hope for the service provider community is that new applications based on these more advanced network features will stimulate extra, monetizable usage. Picture your millennial on her way to a Taylor Swift concert, terrified that a plain-vanilla 5G service will stop her from uploading selfies and dance footage to her social media channels. With a few clicks on the smartphone screen, she could enjoy a "RAN differentiated connectivity" feature that night for a small additional fee and ensure her Swiftie followers avoid disappointment.

This is a somewhat hackneyed example that might never take off, and RAN differentiated connectivity could have greater potential among enterprise customers seeking service level agreements and guarantees. But it illustrates the basic idea. "The fact the network is programmable means you have the required capabilities in your radio access network that give you the flexibility to be able to change and adapt and deliver the required intents," said Tombaz.

Covering all the bases

The standalone variant of 5G is not a prerequisite for all the new features. Ericsson also insists the underlying hardware is irrelevant. Essentially, that means features should work the same on both purpose-built 5G and virtual or cloud RAN infrastructure. The necessary software is hosted in what the industry calls a distributed unit (DU), a box responsible for RAN compute. In purpose-built 5G, that would be an Ericsson appliance featuring application-specific integrated circuits. Virtual RAN replaces this with a common, off-the-shelf server and general-purpose processors, typically made by Intel.

Even so, Ericsson admits software is not going to be identical due to some degree of hardware dependency at Layer 1, the slice of RAN software hungriest for IT resources. "Layer 1 will always be different and we will need to write the Layer 1 code for our own accelerator, and we will also need to do that for other accelerators for the cloud RAN," said Tombaz. She is optimistic that Ericsson can do this efficiently while guaranteeing feature parity between the purpose-built and virtual RAN portfolios. "We know how to do it, and we are prepared to do it," she said.

Outside Ericsson, there is likely to be some concern that new features implemented in DUs will shrink the opportunity for third parties developing rApps, hosted in bigger and more centralized facilities. Features such as "energy efficiency and management" seem to obviate the need for many of the rApps advertised so far.

"Our vision is that we will solve whatever problem we can solve in the RAN, because that is where you access the real-time information," said Tombaz. "You cannot wait for seconds, minutes or hours to be able to have a closed loop." Ericsson, of course, still resists the parallel concept of xApps with near-real-time functionality.  But that should not be a great surprise. As software becomes more critical to its future, it will naturally want to retain as much of the business as it can.