The weak telco appetite for a 6G hardware splurge is already prompting some big shifts at Ericsson and Nokia.
Every ten years or so, governments charge telcos billions of dollars for the rights to send mobile signals over a crisp new slice of electromagnetic spectrum. The birth of another generation of mobile technology is an exciting and potentially lucrative time for the suppliers to those telcos, too. New spectrum requires new radios. Ericsson, Huawei, Nokia, Samsung and ZTE, the world’s biggest radio access network (RAN) vendors, look forward to the launch of a G like a toymaker awaiting Christmas.
But the forthcoming 6G standard, expected to tumble down the chimney of the telecom sector around 2030, could seem more of a thieving Grinch than a gift-bearing Santa. 5G, the latest generation currently in use, has been a major disappointment to most of its operators, costing billions to deploy without delivering any meaningful sales growth. No “killer app,” persuading consumers or businesses to spend more on connectivity, has materialized. The only compliment paid to 5G is that it has provided additional capacity for traffic-clogged networks. All this has sapped telcos’ enthusiasm for the next G.
The message from the Next Generation Mobile Networks Alliance (NGMN), a club of big operators, has been emphatic. “6G must not inherently trigger a hardware refresh of 5G RAN infrastructure,” said the organization (PDF) in a previous white paper. 6G’s introduction should be possible “through software-based feature upgrades of existing network elements,” added the authors. Howard Watson, the soon-to-retire chief security and networks officer of BT, voiced sentiments shared by others when previously asked if 6G would necessitate another investment splurge. “We’re not doing that again,” he said.
Ten years after
Consequently, the ten-year upgrade cycle, which started with 3G, may be coming to an end. The implications for the big vendors are huge. Future revenues might be lower, but without their historical peaks and troughs. Software looks set to become an even more important part of the product mix. Some hardware activities will probably be relinquished altogether. Vendors will be leaner, with a much smaller number of employees. Profitability might improve.
Nothing out of the 6G standardization workshops that have happened this year suggests the industry should expect otherwise. 6G will almost certainly be based on orthogonal frequency division multiplexing (OFDM), the same waveform technology that underpinned 4G and 5G, making it easier to introduce alongside the older standards. At first, it will have to be deployed in much higher spectrum bands, simply because lower and more valuable frequencies are already occupied. Even if Nokia is right in thinking advanced antenna systems will be able to extend the range of a 7GHz signal to that of 3.5GHz one, telcos would struggle to justify a major investment.
Minus a killer app that boosts their own sales, the only reason for investing would be to provide extra capacity on congested networks. This might be more economical than adding 5G equipment to current sites. Yet the rate of traffic growth has been slowing on mobile networks, data from multiple sources now shows. William Webb, an analyst, author and former regulatory executive, reckons growth will stop before 6G is due to be launched. Even if it doesn’t, AI will deliver massive improvements in spectral efficiency, according to vendors and telcos. That should enable 5G hardware to cope.
It’s the end of the Gs as we know it
Recognition that the Gs cycle has finished came from none other than Börje Ekholm, the CEO of Ericsson, at the very start of this year. On a call with analysts about annual results, he urged the industry to think of 6G as an “evolution of 5G” and “not as a new type of generation.” With that, the peaks and troughs of the historical investment cycle will flatten out. Ericsson’s sales will owe more to software upgrades than hardware components.
Before 6G arrives, or in its early years, Ericsson’s full retreat from the development of custom silicon for RAN compute is no longer inconceivable. For several years, it has dabbled in virtual RAN, substituting Intel’s general-purpose central processing units (CPUs) for its own silicon products. Custom has long trumped general purpose on efficiency measures. But the performance gap between them appears to have narrowed significantly with the release of Granite Rapids, Intel’s latest CPU, as acknowledged by Ericsson and various telcos.
Funding the development of semiconductors is an expensive business, too, and Ericsson’s R&D budget has grown sharply as a percentage of sales under Ekholm’s tenure. In 2016, before he took charge, R&D spending accounted for about 14% of revenues. Last year’s figure was almost 22%. If the company leans more heavily on Intel’s CPUs in the future, it could fatten profits or divert R&D spending into radio technologies and software instead.
To avoid dependency on Intel, and de-risk its investments, Ericsson says it is working to make its software as hardware-agnostic as possible. The idea is to have a set of code it could deploy on various CPU platforms, including Intel’s x86 architecture and alternatives based on the rival Arm system, without having to make big changes. This full disaggregation of hardware and software, as Ericsson describes it, would suit the objectives of telcos and chime with Ericsson’s messaging about programmability. Yago Tenorio, Verizon’s chief technology officer, believes Granite Rapids will be “software-upgradeable” to 6G.
Marvell-less 6G?
Nokia seemingly decided to quit investing in custom silicon for baseband needs when it accepted funds of $1 billion from Nvidia in late October. Until now, it has partnered with Marvell Technology on the development of chips for its 4G and 5G network products. In the future, it will generate 5G Advanced and 6G software that runs on Nvidia’s graphics processing units (GPUs). Justin Hotard, who became Nokia’s CEO in April, said it marks a “shift from proprietary to general-purpose hardware” at the company’s recent capital markets day in New York.
Much like Ericsson, Nokia sounds keen to avoid overreliance on a single supplier. While independence from Nvidia as its second-biggest shareholder will be harder to realize, the Finnish company is putting as much as possible into what insiders describe as a bucket of common software. Nokia is also optimistic that it will be able to transfer software written for CUDA, Nvidia’s GPU platform, to other GPU platforms without the need for a major rewrite. The real challenge for Nokia is persuading telcos to invest in what some regard as a very expensive item of semiconductor hardware.
Samsung similarly appears to have retreated from the purpose-built 5G technology it has co-designed with Marvell. Recent deals with Vodafone highlight products based on Dell servers and Intel’s CPUs. Telus, a Canadian customer, favors this common, off-the-shelf equipment over the dedicated appliances in Samsung’s portfolio. Verizon, Samsung’s biggest 5G customer in North America, has repeatedly talked about the advantages of its virtual RAN.
None of this directly changes the outlook for radio units, where the three vendors may continue to have a big hardware role. But higher-band spectrum constraints and the costliness of equipment will limit the appeal of new 6G radios. Refarming 3G and other frequencies for 6G deployment might not require the “hardware refresh” some telcos sound eager to avoid.
Software breakthroughs could also boost spectral efficiency on existing hardware. Ericsson recently claimed to have seen a 10% gain in trials that used AI for link adaptation, a RAN algorithm. Nvidia boasts even bigger improvements via its GPUs. Software designed by Cohere Technologies, a startup, improved spectral efficiency on a deployed 5G network by up to 50% during recent trials in Spain, according to Vodafone.
TikTok and whatever
Not everyone is convinced 6G is the end of the story. On a call with financial analysts last week, SBA Communications, a US towerco, predicted that capital intensity – expenditure as a percentage of sales – will rise with 6G’s launch. “It’s the lowest it has ever been for the big operators,” said Marc Montagner, SBA’s chief financial officer, putting the figure today at about 14.5% against a historical range of between 15% and 25%.
“6G will come,” he continued, according to the Seeking Alpha transcript. “The FCC is going to auction up a C-band in 2027. And the wireless operators are going to have to roll out a new technology because the world is going wireless with fixed wireless access, with AI and TikTok and whatever.”
Whether Montagner is right or wrong, the big kit vendors on which telcos rely are shedding staff and abandoning earlier activities. Ericsson has cut more than 15,600 jobs since 2022, finishing September with around 89,900 employees. Nokia scrapped 9,500 jobs between 2022 and the end of last year, about 12% of the earlier total.
Nokia has already transferred cloud infrastructure assets to IBM-owned Red Hat, withdrawing from that market, and more recently signaled a retreat from microwave backhaul and parts of the private 5G sector. An Ericsson exit from baseband silicon would represent further shrinkage. Whatever guise 6G wears, its biggest vendors outside China look set to be a lot smaller than they once were.
https://www.lightreading.com/6g/ericsson-and-nokia-get-set-for-the-end-of-the-gs
