Humanoid robots, smart glasses and other AI gadgets may send more traffic than they receive, says Ericsson. But persuading telcos to invest will be a challenge.
If 5G’s most notorious use cases that never happened were the self-driving car and remote-control surgery, 6G’s will probably cover some form of physical AI. Ericsson’s pre-MWC Barcelona presentation in London two weeks ago showed an industrial worker with an AI visor beneath his hardhat, tracing lines in the air like a sorcerer mid-incantation. On the next slide, a smiling robot sat next to a hospital bed, its fiberglass fingers gingerly stroking the patient’s arm.
The Swedish company is drawn to such examples of how 6G could be used because they emphasize the potential importance of the uplink in a world of physical AI. 4G and 5G have been largely about streaming content on the downlink, whether games, movies, music or something else. The main attraction of 6G might be its ability to handle massive uploads of data, as robots and other AI gadgets gather information and feed it to the cloud. Ericsson’s cheerful-looking bedside Terminator sends more than twice as much data as it receives, with a “traffic ratio” of 70% uplink and 30% downlink, according to Ericsson.
All this, of course, would require 6G to provide much better support for the uplink than its predecessors have done. Ericsson’s top executives talk about a decoupling of the uplink and the downlink, using language that normally describes their approach to hardware and software in mobile networks. “The things that are changing, especially on the radio side, have much more to do with flexibility – having almost a complete separation of uplink and downlink,” said Erik Ekudden, Ericsson’s chief technology officer.
Band aid
The problem is that initial 6G deployments, now expected to happen in 2029, are likely to use much higher spectrum spanning the 6GHz and adjacent “golden” bands – a range that starts at 5.925GHz and ends at about 8.4GHz before frequencies become relatively useless and uneconomical for a nationwide network. As expansive as this sounds, chunks of it are unavailable depending on the region. It also propagates badly, struggling to blanket wide areas, overcome obstacles or make it through the front door of a building.
Experts might be able to boost coverage through technologies such as beamforming, which allows radio signals to be steered to their targets like heat-seeking missiles. Massive MIMO, a related concept, crammed antenna elements into 5G radio units. Technologists are now talking about extreme MIMO and giga MIMO to describe 6G radios featuring even bigger numbers of those antenna elements.
But there is skepticism that golden band 6G will ever map to the 5G grid. “We know that some of these use cases, whether it’s uplink demands, whether it’s more consistent services, will require densification,” said Ekudden. “Best-effort networks could make do with the existing grid. Our recommendation for leading telcos would be not to stay with that, because then you will not get the full benefit of the upgrade to 6G.”
Densification, or the building of additional sites, obviously demands considerable investment. It did not happen widely in 5G, which relied on higher spectrum bands offering poorer coverage than 4G, and it seems even less likely in 6G unless the forthcoming generation brings sales growth for telcos.
Joe Madden, an analyst with Mobile Experts, expects vendors to sell only about half as many basestations for 6G as they did in the 4G and 5G cycles, according to research findings he recently published on Fierce Network. Instead, operators will look to AI for spectral efficiency improvements while leaning on massive MIMO (or its successor), satellite systems and software upgrades to handle capacity needs, he reckons.
What’s more, spectrum constraints could be an issue in the golden band partly because it uses a technique called time division duplex (TDD) to send data over the air. In TDD, uplink and downlink communications travel over the same spectrum channel. Collisions are avoided by closing the channel to the uplink when downlink data is on the move and vice versa, rather like a set of traffic lights controlling a single-lane road.
None of that is an outright barrier to using new TDD spectrum for uplink communications. But many operators are already sitting on unused uplink spectrum. Frequency division duplex (FDD), an alternative technique, works more like a normal two-lane road, reserving one spectrum channel for the uplink and the other for the downlink. Much of the pre-5G spectrum awarded to operators is suitable for FDD deployment. And the uplink channels, in many cases, are deserted stretches of highway.
If uplink takes off with the arrival of physical AI, operators might be able to rely on existing spectrum and radios to support it, rather than densifying networks and investing in new 6G technologies. Ericsson’s latest pitch also chimes with Madden’s research by emphasizing the 5G improvements telcos could realize through AI. Much of the software could seemingly be deployed in the radio access network (RAN) without the need for expensive new hardware.
Tough sell
It’s also unlikely the new standard will bring any huge and disruptive changes, such as the adoption of a new waveform, although Ekudden does not rule this out. “However, if you can’t prove that there is some real merit to your proposals, that’s going to be quite a tough discussion in the standards meetings,” he said. “You have to prove that you’re coming with something that’s significantly better.” But if 6G is an “evolution,” a word used by Ericsson CEO Börje Ekholm to describe it, the gains might be too marginal to justify a 6G deployment in the lower bands that currently support 4G and 5G.
That said, some big operators are starting to invest in FDD-based massive MIMO, a technology that until recently was focused on TDD, partly to support future uplink needs. US telco giant Verizon is one such company, notes Gabriel Brown, a principal analyst with Omdia. Ericsson, meanwhile, continues to highlight the importance of a multiband strategy as 6G draws closer. “I hope that is a learning that everyone has taken to heart, which is that it’s not enough to have 5G in only a single band, whether low band or mid-band, but to have a combination,” said Ekudden. As older technologies are retired and spectrum is freed up, operators will naturally invest in the latest G available.
Not everyone is convinced by the 6G arguments about uplink traffic. Ericsson’s own mobility reports show that the rate of traffic growth has recently slowed, even if the exabyte totals are much bigger. Smart glasses and headsets are unlikely to be transmitting data all the time. More problematically, they will often be used indoors where fiber-rooted Wi-Fi has become ubiquitous. Research shows that up to 80% of smartphone data is today sent over Wi-Fi rather than cellular technology.
A slide in Ericsson’s London presentation showed that uplink traffic is now growing faster than downlink traffic for 80% of operators. It could, says Ericsson, grow by three times every five years starting in 2025, according to current expectations. Yet there has famously been no correlation between downlink traffic growth and connectivity revenues earned by telcos.
Ericsson once argued that traffic growth would eventually force its customers to make network investments. But annual RAN product revenues have not recovered after dropping from $45 billion in 2022 to $35 billion in 2024 and 2025, according to Omdia. Without evidence those smiling robots and visor-wearing engineers will be a good thing for the networks, telcos may continue to show a limited appetite for 6G.
