Accomplice Article
When you’ve got attended any Cellular World Congress occasion in recent times, you’ll by no means have been distant from a robotic. From mechanical canine prowling the halls at commerce reveals to robotic baristas serving espresso, robots have lengthy been a novel option to reveal the facility of cell know-how.
With the arrival of AI, nonetheless, the embodied AI is being springboarded in direction of sensible deployments, with autonomous operations changing into more and more viable. From Honor’s humanoid robotic ‘Lightning’, which broke the human world file for a half-marathon earlier this yr, to robotic canine serving to present safety on the FIFA World Cup, the robotic period is virtually upon us. AI that had as soon as been confined to a cellphone or laptop computer display screen will quickly be making the leap to the bodily world.
What does that increase in bodily AI imply for networks?
At MWC Shanghai 2026’s 5G-A Trade Evolution Summit, discussions round 5G-Superior (5G-A) had been now not focussed on merely larger velocity and capability, as a substitute presenting the know-how as a foundational layer upon which the rising bodily AI ecosystem could be constructed.
However totally supporting multi-modal brokers, real-time digital twins, and autonomous humanoid robotics will depend on greater than a easy improve. Operators will be required to radically re-engineer the underlying 5G community, prioritising low latency, uplink and effectivity greater than ever earlier than.
This paradigm shift might be a serious problem for the cell trade, however it may provide an enormous reward: the creation of a token-based enterprise mannequin that would lead a path to progress.
Constructing symmetrical networks for comfortable robots
Maybe the most vital change represented by the creation of embodied AI is the larger demand for uplink.
For a few years, cell networks have been designed for a downlink-heavy world dominated by shopper video streaming and net looking. With the speedy rise of AI, nonetheless, this architectural norm is being overthrown.
Humanoid robotics, autonomous industrial autos, and multi-modal AI terminals will all rely on evaluating massive quantities of information – usually from quite a few sources in various media – in real-time. This can require speedy compute capabilities to make sure the near-instant response occasions essential for autonomous exercise.
Essentially the most fundamental resolution for this might be to merely place the required compute capabilities on the gadget itself, whether or not that could be a buyer smartphone or a robotic sentry canine. The issue, nonetheless, is that working power-hungry GPUs straight on these units destroys their battery life and business viability.
“Excessive power consumption and the ensuing quick battery life is a limiting issue,” mentioned Chen Qi, president of AI product line at TD Tech, an organization she described as “a robotic mind enterprise”. “Utilizing a robotic mind [in the device] takes round 20-times extra power throughout autonomous exercise than working it remotely. We shouldn’t be placing that stress on terminals – we should always use the cloud and put that stress on the networks.”
Networks will subsequently be required to balancing downlink and uplink, guaranteeing {that a} minimal degree of uplink capability is delivered to all related AI terminal units. World operators are progressively reaching a consensus that 20Mbps uplink will grow to be the baseline technical requirement to maintain real-time AI modelling, situational consciousness, and digital twins.
In a world stuffed with AI terminals – 15 billion by 2035, in accordance with Huawei’s Clever World 2035 report – 5G-A might be important to making sure that degree of uplink at scale and sustaining cloud-edge synergy.
“Scaling autonomous intelligence places quite a lot of stress on our networks,” mentioned Yang Lifan, Deputy Common Supervisor of China Unicom Beijing. “We are able to deal with two cameras per robotic, however what about eight? We are able to help 5 robots on the website, however what a few hundred working concurrently? We have to extremely optimise our 5G-A networks for these situations and meaning a a lot larger deal with uplink.”
It’s no coincidence that Huawei launched its GigaUplink resolution on the occasion, utilizing multi-antenna know-how upgrades and new algorithms to ship a five-fold enhance in uplink capability.
Past adjustments to throughput demand, the latency necessities of embodied AI are essentially completely different from shopper web use. When a robotic or autonomous car interacts with human environments, it requires human-like response latencies – round 650ms – to make sure security and precision. In consequence, best-effort community supply will quickly be out of date for B2B industrial use circumstances, with deterministic efficiency changing into a vital community characteristic.
“Massive bandwidth, uplink enlargement, and consumer expertise assure. These would be the key community options that allow the cell AI period,” mentioned Eric Yang, President of Huawei Service Enterprise.
A name for Higher 6GHz spectrum
Shifting community structure technique is barely half of the battle for delivering steady protection for a quickly AI ecosystem. Spectrum bottlenecks are a serious concern, with further capability required to make sure ubiquitous easy service.
On the Summit, securing steady midband spectrum was seen as foundational for delivering multidimensional experiences, with the higher 6GHz (U6G, 6.425–7.125 GHz) band positioned as a key useful resource. It gives a robust mixture of each protection and capability, complementing current mid-band 5G spectrum and bridging the hole to 6G.
This name for entry to U6G comes throughout an ongoing world debate about the way forward for the band. U6G is extremely coveted by the Wi-Fi trade to alleviate stress on the crowded 2.4 GHz and 5 GHz bands. Nonetheless, as Tim Hatt, Head of Analysis and Consulting at GSMA Intelligence, factors out “cell is more likely to be capability constrained than Wi-Fi.”
“We must always actively promote U6G and align it with C-band, whereas refarming decrease bands for much more capability,” argued David Li, President of Huawei’s TDD Product Line. “U6G is the second-best spectrum for widespread 5G-A deployment after C-band (3.4–4.0 GHz). With enhancements to our know-how, we will quickly have the ability to make the U6G protection pretty much as good as C-band.”
In tandem with U6G entry, refarming spectrum within the legacy 2G and 4G bands will even be a precedence. By pooling these frequencies by means of superior provider aggregation, they will ship the ultra-wide bands that 5G-A calls for, creating a strong basis for cell AI use.
Tokens: A manner out of the ‘quantity entice’?
Monetising 5G usually seems to be an evergreen problem for the cell trade. Regardless of widespread 5G deployment and protection reaching over 99% in premium testbeds like Hong Kong, world ARPU has persistently stagnated. The increase in AI terminals, nonetheless, is about to show a basic financial disparity between uncooked information transmission and AI computational workloads.
Below the standard volume-based enterprise mannequin, operators generate minimal income from a gigabyte of information, although transmitting the thousands and thousands of AI tokens inside that information requires immense community sources and drives up computational electrical energy prices. By re-engineering community pipelines round token transmission slightly than bytes, telcos can bundle, resell, route, safe, and invoice for AI capability in ways in which cut back friction for patrons and create new recurring income.
“The trade is transferring in direction of token monetisation fashions,” mentioned Yang, noting that Community-as-a-Service (NaaS) frameworks would permit operators to provide tier-based, deterministic service ensures primarily based on consumer location, utility profiles, and exact latency necessities.
On this manner, Huawei argues that operators must evolve past the pure connectivity layer, changing into an orchestrator of not solely information site visitors however of compute energy.
“A byte-plus-token technique will redefine business worth for operators. Sooner or later, the distinction between information site visitors and tokens will proceed to develop. We have to be able to embrace that, each with how we construct networks and the way we monetise them,” mentioned Li.
Utilizing 5G-A to embrace the long run
The consensus from MWC Shanghai 2026 means that an AI-native ecosystem requires a basic realignment of the cell ecosystem, requiring each infrastructure upgrades and a shift to new business fashions. The further velocity, capability, flexibility, and reliability of 5G-A – supported by further spectrum within the U6G band – will present an excellent basis for the monetisation of the token economic system.
By performing because the unified orchestrators of each spectrum and computational energy, telcos can step out of the quantity entice and safe their place because the indispensable spine of the bodily AI revolution.
