- China’s 350 km/h railway is becoming a testbed for AI-controlled 5G
- China Mobile says premium passengers can receive network performance two to three times better than customers on standard packages
- The bigger story is AI turning control of the network into a new revenue stream
For years, mobile operators have had a 5G problem that had little to do with the technology itself: customers were reluctant to pay much more for it. I recently described this as telecom's great AI dilemma: everybody wants the future, nobody knows how to monetize it.
China Mobile may have found an answer on one of the most demanding mobile networks in the world.
On the Beijing-Shanghai high-speed railway, where trains travel at up to 350 km/h (217-and-a-bit mph, for the Imperial Luddites in the room), China Mobile and Huawei are turning 5G-Advanced into something more interesting than another exercise in peak download speeds. The network can identify that a subscriber is traveling on the train, distinguish that passenger from users alongside the railway, recognize when particular applications need help and allocate network resources to prioritize them accordingly.
Then China Mobile can sell the difference in experience.
Its 5G-A HSR Premium Package effectively creates a premium class within the mobile network. China Mobile and Huawei say customers subscribing to it can receive data rates two to three times those available with standard packages, particularly for latency-sensitive, bandwidth-heavy applications such as video conferencing, livestreaming and gaming. That makes this railway much more than an impressive 5G coverage project. It is an experiment in monetizing control.
The network knows you are on the train
Providing reliable mobile connectivity on a train moving at 350 km/h is already difficult. A handset moves rapidly through successive cells, creating repeated handovers while hundreds of passengers simultaneously consume bandwidth inside a metal tube.
China Mobile has built dedicated capacity along the Beijing-Shanghai route using 2.6-GHz 8T8R and 700-MHz spectrum in a 5G-A three-carrier network. Those frequency ranges are not unique to China: U.S. operators also use 700-MHz low-band spectrum and extensive mid-band holdings around 2.5 GHz. What is unusual is the way China Mobile has combined the two into a dedicated high-speed rail network and then layered AI-driven user and application prioritization on top. Clever.
The company says that combination, together with technologies including multiband load balancing, carrier aggregation and Huawei's train-specific power adaptation, has improved perceived uplink rates by 38.5% and downlink rates by 9.5%.
But adding radio capacity is just the foundation. The more interesting development is in the control system.
China Mobile says it has integrated AI inference into the Network Data Analytics Function, or NWDAF, in its core network. The system studies three minutes of user movement and uses characteristics including speed, base-station transitions and network topology to determine whether someone is traveling on the railway or merely using the public mobile network nearby.
China Mobile claims the system identifies high-speed rail passengers with more than 95% accuracy. Once it has made that determination, the network can adjust frequency-selection priorities to keep railway passengers on dedicated capacity while steering ordinary users elsewhere. At typical base stations, it says as many as 60% of public-network users can be migrated away from the railway network.
In this implementation, the network is no longer merely transporting packets. It is interpreting context and making decisions about how its resources should be used.
From bandwidth to experience
But wait, there’s more.
China Mobile says its NWDAF can autonomously detect deteriorating application performance and dynamically establish dedicated guaranteed-bit-rate bearers for eligible subscribers and services.
In 2025, the system could accelerate 27 applications across five categories. China Mobile says that has now expanded to 55 applications across six categories, including multistream acceleration for short-form video and livestreaming services.
Intelligent processing inside the radio network adds another layer, performing millisecond-level management of rate, latency, power and carrier selection for particular users and services.
This begins to look less like conventional mobile broadband and more like an intelligent service-delivery platform.
Which is a bigly deal, because 5G operators have spent years searching for precisely this distinction.
Selling another 50 Mbit/s to a consumer who can already stream Netflix perfectly well is a stretch. Selling reliable performance for a high-pressure corporate Teams meeting while traveling between Beijing and Shanghai is a no-brainer.
One is just bandwidth. The other is an experience.
Huffy-puffy important business types travel in huffy-puffy important business class. Now they can ensure their data does the same thing. They will pay for that assurance with about as much hesitation as they order the £9 airport coffee when the company is paying.
China Mobile's branding doesn't hurt, either. Premium subscribers get a dynamic China Mobile HSR VIP logo displayed on their phones. And who doesn't want that?
5G finally finds something to sell
Railways make an unusually good environment for proving this model because the value of network performance is obvious. A passenger trying to conduct a video call does not care whether the network is technically 5G, 5G-Advanced or eventually 6G. They care whether the call works.
That is why China Mobile's approach may matter beyond China. And the contrast with some other major markets is fairly savage.
In Britain, Ofcom reported in June that mobile performance was poor on between 58% and 83% of the train tests it conducted, depending on the network. Its definition of a good connection was hardly extravagant: 5 Mbit/s down, 1.5 Mbit/s up and latency no more than 50 milliseconds — enough to make an entirely OK video call.
China Mobile is using AI to decide which passenger's video call should get priority. Britain is still working on making sure the bleedin' thing connects.
London provides a useful footnote to just how quickly this can change. The Elizabeth line originally opened without mobile coverage. The consortium building it "forgot" to install it. But in true Spirit of the Blitz style, TfL rallied and now says all Elizabeth line stations have 4G and the tunnel sections have 4G and 5G coverage — evidence that basic connectivity problems can be solved when infrastructure investment, and common sense, finally catch up.
China Mobile has already moved beyond such engineering questions. It is asking what you can sell once the connectivity actually works.
And, after searching for comparable deployments, I could find no other commercial rail service that combines 5G-A, AI-based passenger identification and application-level network prioritization in quite this way.
That doesn't mean other rail networks are standing still. Germany, for example, is testing 5G standalone for railway communications. But Deutsche Bahn's work is still focused on establishing the next-generation communications architecture. In 2025 it began testing what Fierce Network described as the world's first 5G-based Future Railway Mobile Communication System (FRMCS) on live outdoor tracks.
The contrast is clear. Germany is testing how 5G should run the railway. China Mobile is testing how 5G can make money from the people sitting on it.
Mobile operators have traditionally sold fairly blunt instruments: minutes, messages, gigabytes and speed tiers. 5G promised network slicing, deterministic performance and differentiated quality, but translating those capabilities into products consumers would willingly buy has proved far harder. Less so in vertical industry, admittedly, but that's a whole other market.
I saw one version of that industrial future myself at the East-West Gate intermodal terminal in Hungary, where Huawei private 5G provides the high uplink bandwidth, reliability and guaranteed latency needed to automate railway logistics. The Beijing-Shanghai project takes that basic proposition and applies it to a very different customer: the passenger.
The deployment starts to make 5G's consumer proposition tangible.
Huawei's own answers to FNTV's questions about the case study are revealing. It says additional spectrum and capacity provide the basic experience. AI profiling and intelligent radio processing then provide the fine-grained resource management that improves it. In other words, capacity creates the network. Control creates the product.
That distinction will become increasingly important to operators.
Hyperscale cloud companies built extraordinarily profitable businesses partly by placing proprietary software and control layers above commodity infrastructure. Telecom operators have spent much of the same period worrying that connectivity itself was becoming commoditized.
An intelligent mobile network gives them another option.
T-Mobile provided another glimpse of this direction when its executives described to me how AI is moving from the cloud directly into its 5G SA; abbreviated on later reference== core, with orchestration extending across the core, radio and device layers.
The carrier owns the spectrum. It operates the radio network. It controls the core. Increasingly, it can use software and AI to determine how those assets behave for an individual customer or application. And that control can have commercial value.
The real test starts now
There are still questions. The performance figures supplied to FNTV are China Mobile and Huawei's own measurements and have not been independently verified. More importantly, the long-term commercial test will be whether customers consistently perceive enough difference to keep paying for premium treatment.
Having said that, this passes the Saunders Good Housekeeping Seal of Approval test. I'd pay for it.
And the technological direction is becoming clearer.
I have covered railway communications from Huawei's private 5G deployment at Hungary's East-West Gate terminal to the rather less futuristic reality of a German railway brought to a halt by 2G in 2026. China is pushing the question one stage further. What happens after connectivity becomes reliable?
The network starts becoming programmable. Then intelligent. And finally commercial.
At 350 km/h between Beijing and Shanghai, China Mobile is beginning to demonstrate what that progression looks like. 5G's route to revenue may turn out not to be selling passengers a faster network. It may be selling them a better journey.
Learn more
Opinion: Telecom's great AI dilemma — everybody wants the future, nobody knows how to monetize it
T-Mobile's Ankur Kapoor: AI is moving from the cloud to the core network
Nokia scores first with German 5G train trial
Opinion: Ghost in the machines — why the world can't quit 2G
Stephen M. Saunders MBE is a communications analyst and USPTO-registered inventor examining how digital infrastructure — 5G, cloud and AI — is reshaping industry, power and society, as well as underpinning the emerging, ubiquitous global digital economy. As anchor of FNTV and a longtime industry insider, he focuses less on growth narratives and more on execution, risk and how hyperscale technology is distorting markets, governance and society at scale.
Opinion pieces from industry experts, analysts or our editorial staff do not represent the opinions of Fierce Network.