How AI, Uplink Traffic and Spectrum Are Shaping the Case for 6G

With 5G still being deployed and 5G-Advanced only beginning to arrive, one of the most reasonable questions about 6G is simply: why do we need it? There is no shortage of 6G vision documents and white papers describing AI-native networks, Integrated Sensing and Communication (ISAC), Non-Terrestrial Networks (NTN), immersive communications and new spectrum. What is often less clear is why these capabilities require a new generation of mobile technology rather than continued evolution of 5G.

The XG Mobile Promotion Forum (XGMF) in Japan has attempted to answer exactly this question in a recently published document appropriately titled "Why 6G?". There is also an extended version with additional notes. 

One of the more interesting aspects of the paper is that XGMF does not define 6G simply as a faster version of 5G. It sees 6G as part of a wider digital infrastructure connecting people, things and AI. Distributed computing resources across the cloud, edge and devices would be connected through the network, with AI increasingly involved both in applications and in operation of the network itself. XGMF describes this as a kind of "nervous system" connecting computing resources across clouds and devices. 

This leads to an important distinction between 5G and 6G. Improvements in spectral efficiency, user experience, uplink and downlink performance, reliability, resilience and energy efficiency will continue, but XGMF argues that 6G introduces capabilities that go beyond conventional connectivity. These include AI-native operation from the outset, positioning and sensing, and ubiquitous connectivity using technologies such as NTN and HAPS.

Perhaps the most interesting part of the document, however, is the discussion around uplink traffic. Mobile networks have traditionally been heavily downlink dominated. Users consume video, social media, webpages, software updates and other content from the network, while comparatively less data travels in the opposite direction. The arrival of AI could start changing this balance.

XGMF identifies several drivers for increasing uplink demand. Industrial applications already generate uplink traffic through video streams and remote-control systems. AI-enabled phones, glasses and vehicles could send increasing amounts of data towards network and cloud-based AI systems. Autonomous AI agents may continuously generate sensing and contextual information, while Physical AI applications involving robots, vehicles and other machines could increasingly depend on network-side computing.

There is already evidence of this shift. In data included in the XGMF appendix, based on measurements from 55 operators worldwide during 2025, 43 of the 55 operators experienced faster uplink traffic growth than downlink traffic growth. For 17 operators, uplink traffic growth was more than 1.5 times the downlink growth rate. XGMF is careful to point out that downlink still represents the majority of traffic in absolute terms, but the direction of travel is significant.

The document also highlights a scenario where uplink traffic in 2031 could reach around three times the 2025 level. AI agents and sensing applications are particularly important because, unlike a human occasionally uploading a photograph or video, machines could generate data continuously and autonomously.

This could have important implications for mobile network design. Networks have historically been optimised around asymmetric traffic. If AI devices, cameras, vehicles, robots and sensors begin producing much larger amounts of uplink data, operators may need to reconsider radio capacity, spectrum configurations, network dimensioning and where computing resources are located. This does not mean that mobile traffic suddenly becomes uplink dominated, but it does suggest that the old assumption of overwhelmingly downlink-centric demand may become less useful during the 6G era.

Another reason XGMF gives for 6G is the familiar problem of continued overall traffic growth. The organisation expects mobile traffic to keep increasing and argues that meeting this demand will require a combination of additional spectrum, network expansion and improvements in technology efficiency. 6G therefore becomes part of a wider capacity solution rather than simply introducing new applications.

This is where the spectrum discussion becomes particularly interesting. XGMF specifically highlights the 6425–7125 MHz and 7125–8400 MHz ranges as important candidates for new 6G mid-band spectrum. It argues that contiguous bandwidth of approximately 200–400 MHz, potentially involving multiple blocks, will be needed to achieve a useful balance between capacity and coverage while supporting capabilities such as high-resolution sensing. 

This reinforces something that is becoming increasingly clear from global 6G discussions. While early research spent considerable time looking at extremely high frequencies and sub-THz spectrum, the practical foundation for wide-area 6G is increasingly expected to include additional spectrum in the upper mid-band range. These frequencies offer considerably better coverage than very high-frequency bands while potentially providing wider channels than today's traditional mobile bands.

XGMF also makes an interesting point about the transition from existing networks. It says that by the expected introduction of 6G around 2030, 5G should be mostly Standalone (SA). 4G would continue for some time, but the objective should be to avoid operating too many generations and network architectures simultaneously.

This is significant because it connects today's 5G deployment decisions with the future 6G transition. In many countries, much of the 5G deployed so far still relies on Non-Standalone (NSA) architecture using a 4G core and 4G signalling anchor. Moving towards 5G SA is therefore not just about enabling the full capabilities of 5G; it can also be seen as an important step towards simplifying eventual migration to 6G.

Another refreshing aspect of the XGMF paper is that it does not ignore the difficult question of who pays for all of this. XGMF describes 6G investment as early investment in "AI-era industrial infrastructure". Potential value comes not only from conventional connectivity but from connecting Physical AI, accessing distributed AI compute, providing resilient wide-area connectivity through NTN and HAPS, extracting value from sensing, and using AI-native automation to reduce operating costs. 

The proposed monetisation evolution is particularly noteworthy. XGMF sees operators progressing from monetising basic connectivity and network quality towards value-added capabilities such as authentication, location, sensing and network APIs. A further stage could involve monetising distributed computing resources and network architecture for AI workloads. The document even references the emerging concept of an "AI token economy", although this should probably be regarded as a possible future business model rather than an established part of 6G.

Taken together, this provides a slightly different answer to "Why 6G?" than the usual promise of higher data rates. The argument is that the environment in which mobile networks operate is changing. AI is becoming distributed between devices, edge infrastructure and large data centres. Machines could generate significantly more uplink traffic. Networks may increasingly need to sense their surroundings as well as communicate. NTN and HAPS may become integrated parts of the connectivity system. Mobile infrastructure will need to interact much more closely with computing infrastructure, while operators need new ways of creating revenue from networks that are becoming increasingly expensive and complex.

At the same time, conventional requirements have not disappeared. Traffic continues to grow, more spectrum will be required, spectral efficiency must improve, networks must become more energy efficient, and migration from existing generations needs to remain economically sustainable.

This perhaps provides a better way of thinking about 6G. The justification for 6G is unlikely to be one killer application or one spectacular new radio capability. It is the combination of changing traffic patterns, AI, distributed computing, sensing, ubiquitous connectivity and continued capacity growth that could make the mobile network of the 2030s fundamentally different from the one designed at the beginning of the 5G era.

The XGMF paper is worth reading precisely because it tries to connect these technology discussions with the much more fundamental question that the industry will increasingly have to answer over the next few years: why 6G at all?

Posted by on

Comments

Post a Comment