6G news and analysis for the week ending Friday 11th September, 2026.
Before we begin, a quick note to say that I am just back from TelecomTV’s ‘AI-Native Telco Forum’ event in Düsseldorf. It was great to see 6G getting more public mentions, with telcos now less reticent to mention next generation wireless, especially as the foundations are going to be AI-native. And it was a pleasant surprise to see so much 6G news this week, after a few weeks where the pace slackened.
Next week the 3GPP plenaries convene in Madrid, Spain. The migration decision that RAN deferred in June is now due, which is why the NGMN Alliance has chosen its moment to update its operator guidance. It keeps Multi-RAT Spectrum Sharing (MRSS) as the baseline for 5G-to-6G migration, but with a pointed condition: the evidence on overhead, hardware reuse and feasibility is not yet good enough. It also asks a question 3GPP has been circling for months, namely whether the 6G core is an evolution of 5GC or something more distinct.
Spectrum is our second theme for this week and shows no sign of being settled. The GSMA issued a 60-member declaration that the ecosystem is ready to deploy in upper 6 GHz and urged governments to plan for full-power macro use. Across the Atlantic, the Next G Alliance published spectrum recommendations for US 6G launch by 2030. Both documents are advocacy and both should be read alongside the Plum study we covered a fortnight ago. Taiwan’s ministry hosted the same regulators in Taipei on Wednesday.
I love it when an unexpected source pops up on the 6G Futures radar. This week, a Europol and JRC foresight report devotes a full chapter to 6G and it comes across as a law enforcement wish list for network sensing, much to the delight of the ISAC advocates. It also flags that AI-run networks may be hard to interrogate in court.
Elsewhere, KDDI and Samsung push the 7 GHz record to 3.6 Gbps, India joins the US-led coalition and we have four thought-provoking research papers covering agents, tokens and HAPS. Settle down, it’s going to be a long read.
NGMN sets conditions on MRSS before 3GPP decides
The NGMN Alliance has issued updated operator guidance on 6G architecture and migration, timed for the 3GPP plenaries in Spain from 14 to 18 September. It follows the operator-view publications released in June, and NGMN had committed to this update when RAN #112 failed to reach a migration decision. The headline position is unchanged: MRSS (where 5G and 6G share the same carriers) should remain the baseline. However, the tone has hardened. NGMN says more evidence is still required on MRSS efficiency, implementation feasibility, performance and hardware requirements before it can be confirmed as the preferred solution. Laurent Leboucher, NGMN Board Chairman and Orange Group CTO, said:
“The 5G to 6G migration path is an essential component of the 6G design, because it will determine the cost of the 6G deployment and hence the economic viability of the associated new features and capabilities.”
Guangyi Liu of China Mobile was more blunt, saying “5G taught us a hard lesson: too many migration options led to fragmentation, delayed rollouts, and costly retrofits.”
What’s new: Two things. First, NGMN sets explicit tests for any migration option beyond MRSS. These include deployment need, hardware impact, implementation feasibility, total cost of ownership and commercial viability. Any alternative must justify the extra standardisation and operational complexity it introduces. Second, the guidance asks the industry to decide whether the 6G Core is an evolution of 5GC or a distinct architecture, and to trace the consequences for devices, RAN and the relative value of each migration option. That question was absent from the June outputs.
Why this matters: This lands directly on the decision 3GPP has been unable to make. At RAN #111 in March, the options narrowed to a 6G-anchored approach and a dual-stack approach, with a September deadline. At RAN #112 in June, the MRSS overhead numbers operators had demanded were delivered and noted, but no consensus emerged, and the RAN Chair could offer the joint session no input on migration until TSG #113. NGMN’s insistence on “more evidence” is therefore a message to the working groups as much as to the plenary. Luke Ibbetson of Vodafone framed the commercial position, saying 6G must arrive as “a smooth, software driven upgrade in existing frequency bands” without triggering replacement of deployed infrastructure. The core question is equally pointed, since Apple observed in June that the shape of the 6G core would not be known by September anyway.
Implications for 6G standards: MRSS as baseline aligns with where RAN1 and RAN2 have been heading but NGMN is now conditioning that support on quantified performance. If TSG #113 endorses a 6G-anchored approach with MRSS, operators will expect signalling overhead and hardware reuse to be evaluated as Release 21 requirements, not deferred to implementation. The 6GC question feeds straight into the SA WG2 architecture study, which carries a March 2027 deadline. We will report on the plenary outcomes in the coming weeks.
View source material to read more
GSMA declares upper 6 GHz ready for deployment
The GSMA has issued a statement from Kuala Lumpur that the mobile ecosystem is ready to deploy in the upper 6 GHz band (6425 to 7125 MHz), and is calling on policymakers to plan the spectrum for full-power macro-cell mobile use. More than 60 operators, device makers, chipset suppliers and equipment vendors have signed the message. The GSMA says countries representing over 80 per cent of the world’s population now support mobile use of the band following its harmonisation at WRC-23, and that the first assignments have been completed. It argues the band can deliver the 200 to 400 MHz channel bandwidths it expects 6G to require. Vivek Badrinath, GSMA Director General, said:
“Governments now have an important opportunity to turn international harmonisation into real-world connectivity by prioritising this spectrum for full-power mobile services.”
Meanwhile on Wednesday, Taiwan’s Ministry of Digital Affairs hosted an international spectrum symposium in Taipei with speakers from the GSMA, NTIA, Ofcom and the ITU-R Study Group 5 chair. Deputy Minister Yeh Ning said Taiwan’s 6G preparation “cannot wait until the technology and applications mature.”
What’s new: The 60-signatory ecosystem statement is the substance, since it is intended to answer the regulator’s standard objection that equipment does not yet exist. The 200 to 400 MHz channel figure is also notable. The GSMA has previously argued for a 200 MHz minimum per operator, framing 400 MHz as the upper end of what 6G expects raises the ask. The Taiwan event adds little to the spectrum review the ministry opened in May, with an allocation plan still promised for the end of 2026, but the guest list shows where Taipei is taking its cues.
Why this matters: This is industry advocacy, and it should be read against the Plum Consulting study we covered a fortnight ago, which found no meaningful consumer loss from capping IMT at 600 MHz and questioned whether every operator needs 200 MHz. The GSMA does not engage with that argument here. Where the statement is on firmer ground is the field evidence. Optus and Nokia reported 3.5 Gbps over 200 MHz in Sydney, and du, Telefónica and SoftBank have each reported coverage comparable to existing mid-band from existing sites. The pressure point is Europe, where Ofcom settled on 540 MHz of mobile priority spectrum in July and the GSMA called even that insufficient.
Implications for 6G standards: The band is already 3GPP n104, so the specification question is channel bandwidth rather than band definition. If the GSMA’s 400 MHz expectation gains traction in Release 21 evaluation assumptions while European regulators hold near 600 MHz total, the gap between regional band plans widens further. That is the same fragmentation risk we flagged when Asia-Pacific tabled a 6425 to 8400 MHz tuning range.
View source material to read more
View source material to read more — Taiwan MODA
Next G Alliance maps a US spectrum route to 2030
ATIS has published a Next G Alliance white paper, 6G Spectrum Recommendations, described as a strategic blueprint for securing the spectrum needed to launch commercial 6G networks in the United States by 2030. Produced by the NGA Spectrum Working Group, it evaluates five frequency ranges in the emerging US pipeline: 1.675 to 1.695 GHz, 2.69 to 2.9 GHz, 3.98 to 4.2 GHz, 4.4 to 4.94 GHz and 7.125 to 7.4 GHz. It also examines the ‘One Big Beautiful Bill Act’ mandate to repurpose 800 MHz for commercial mobile use by 2033 and the complementary role of mobile satellite services and supplementary coverage from space. Jaydee Griffith, NGA Managing Director, said:
“This white paper provides a practical roadmap for policymakers, regulators, and industry leaders to align spectrum decisions with the demands of next-generation applications.”
What’s new: The band-by-band reasoning is the useful part. Four of the five ranges are the federal bands the NTIA cleared for study in August. The fifth, upper C-band at 3.98 to 4.2 GHz, is non-federal satellite spectrum under FCC evaluation. NGA argues that clearing it would extend existing C-band holdings into a contiguous 400 MHz allocation from 3.7 GHz, and it explicitly names Multi-RAT Spectrum Sharing between 5G and 6G as the route to introducing 6G there on existing site grids. The 7.125 to 7.4 GHz range is described as suited to “pure 6G waveforms”, with coverage parity achievable only through improved antennas and higher radiated power. The recommendations follow from that analysis: prioritise large contiguous blocks, accelerate incumbent studies, harmonise internationally where feasible, and set transition plans realistic enough to permit early deployment.
Why this matters: The paper arrives as the NTIA pipeline moves from plans to studies, with the 7.125 to 7.4 GHz report due in December. Its most telling judgement is that the bands most compatible with existing infrastructure are the C-band extension and 4.4 GHz, not 7 GHz. That is an acknowledgement that the “6G band” of global regulatory focus is the hardest of the five to deploy economically. The MRSS reference is also notable, since it ties a US industry spectrum paper to the migration debate NGMN has just reopened at 3GPP. The 2030 launch framing deserves a caveat: Release 21 specifications are not expected before late 2029 and NGMN’s operator members talk of commercial introduction in the early 2030s. A US 2030 launch would be first-wave at best. The demand case built on robotics, automation and immersive applications also sits uneasily with the Plum finding that traffic growth alone does not justify larger allocations.
Implications for 6G standards: The US pipeline is broader than the upper mid-band consensus forming in Asia-Pacific and Europe, spanning L-band to 7 GHz. Contiguous block size feeds directly into 3GPP channel bandwidth assumptions, and pairing 2.69 to 2.9 GHz with Band n41 or extending C-band to 400 MHz would both require new band definitions in Release 21. The MSS and SCS angle connects to the NTN work already scoped for 6G. Harmonisation “where feasible” is the phrase to watch at WRC-27.
View source material to read more
Europol sees 6G sensing as a policing tool
Europol and the European Commission’s Joint Research Centre have published a joint horizon-scanning report on privacy-enhancing technologies, and it contains a full chapter on 6G. The exercise was designed to identify technologies that could help or hinder European law enforcement and 6G crops up on both sides. The report describes network-integrated sensing as a way for authorities to track movement patterns, detect anomalies and improve crowd management. It also warns that AI embedded across RAN, core and edge will make digital evidence harder to trace and AI-driven decisions harder to challenge in legal proceedings. One interviewee is quoted: “as standards are being discussed by industry, now is the time to act.”
What’s new: A law enforcement agency has articulated, in a formal EU publication, an operational interest in 6G sensing. The policy requests are equally direct: mandated EU-based cloud for European telecom providers, multi-cloud redundancy requirements, and early regulatory engagement before standards close. The report also flags data mobility between cloud, network nodes and devices as a jurisdictional problem for lawful interception.
Why this matters: Integrated sensing and communication (ISAC) is usually discussed here as a capability, with drone detection and factory safety as the example use cases. This report reframes it as surveillance infrastructure and it does so from within the EU’s own institutions. That will no doubt sharpen the privacy debate that has run through the standards work since the 3GPP 6G Workshop. The AI opacity point is well made, as an autonomous network that cannot explain a routing or security decision is a problem for accountability, not just operations.
Implications for 6G standards: 3GPP has already scoped the relevant work. SA #110 approved a 6G lawful interception study for SA WG3-LI, held as an internal technical report with a December 2027 target, which cannot start until the other 6G groups complete their studies. That sequencing means law enforcement requirements will arrive late in the Release 21 cycle. Europol’s report is an early signal of what those requirements may contain and of how much political weight the ISAC privacy question will carry in Europe.
View source material to read more
KDDI and Samsung push 7 GHz downlink to 3.6 Gbps
KDDI, KDDI Research and Samsung Research have reported an outdoor demonstration in the 7 GHz band at Samsung’s R&D campus in Seoul, conducted from 18 to 20 August. Using a base station with Extreme Massive MIMO, more than a thousand antenna elements and 100 MHz of bandwidth, the trial achieved 3.6 Gbps downlink to a single terminal across eight spatial layers. The gain came from applying 1024QAM modulation, which delivered a 20 per cent uplift over the 3.0 Gbps achieved with 256QAM in the same configuration. The partners also report that 95 per cent of measurement locations met the received-power threshold for stable service, matching a sub-6 GHz base station at the same site. KDDI links the work to its “Digital Belt Initiative” for AI-era capacity.
What’s new: The modulation step is the new element. Reaching 1024QAM outdoors at 7 GHz requires signal quality that the beam concentration from a very large array is claimed to provide. The coverage parity figure is also a useful data point, though the measurement area is not described.
Why this matters: The 7 GHz evidence base keeps growing ahead of WRC-27, where the 7125 to 8400 MHz range sits under Agenda Item 1.7. This trial is a direct descendant of the Samsung and KT result from February, which hit 3 Gbps on the same campus with the same 256-port class of prototype, and of the arXiv field paper we covered in March. The change of operator partner from KT to KDDI suggests Samsung is using the Seoul testbed as a showcase for multiple customers. The propagation challenge at 7 GHz is a real concern and coverage parity claims from Seoul, Sydney, Tokyo, Valencia and Dubai are now consistent enough to carry weight with regulators.
Implications for 6G standards: Eight-layer single-user MIMO over 100 MHz with 1024QAM is a reference point for Release 21 evaluation assumptions on extreme MIMO, which Samsung has championed in 3GPP under its NeX-MIMO branding. The result does not change the spectrum arithmetic. Operators are asking for 200 to 400 MHz channels precisely because 100 MHz is the ceiling here.
View source material to read more
India joins the US-led 6G coalition
India has become the 26th country to affirm the US “Call to Action for 6G Leadership and Security”, the US Department of Commerce announced via X (Twitter). Union minister of state for commerce Jitin Prasada met Commerce Secretary Howard Lutnick on the sidelines of the G20 Innovation Ministerial, where the two discussed 6G. The department said India’s accession would help ensure next-generation networks “reflect our shared security interests, strengthen our competitiveness, and drive innovation in 6G.” Prasada said the joint effort “will give a new direction to the global digital landscape.” The story was reported by (amongst many others) the Times of India and India Weekly.
What’s new: Only the accession itself. The coalition launched on 27 July with 25 governments and India was a notable absentee given its 2023 Bharat 6G Vision and its stated aim of frontline contribution by 2030.
Why this matters: India’s participation changes the coalition’s character more than its numbers. It is the only signatory with a 3GPP Organisational Partner (TSDSI) whose contribution count has risen fifteenfold in five years, and it is a BRICS member that only last month was fielding a Chinese proposal for BRICS 6G cooperation. Joining a Washington-led security framework while keeping that door open is characteristic of Indian technology diplomacy. The commitments India now inherits are the coalition’s dated deliverables: a designated 6G agency within a month and industry engagement within three.
Implications for 6G standards: The coalition disclaims any intent to bypass 3GPP or ITU, but it does commit members to coordinate positions inside them. India’s WRC-27 stance on the 6425 to 8400 MHz tuning range, which the ITU-APT Foundation of India championed at APG27-3, is now nominally aligned with a US bloc pursuing a different pipeline. Whether those positions converge or merely coexist will be visible in Shanghai next year.
View source material to read more — Times of India
View source material to read more — India Weekly
Selected research papers published this week
When RAN agents need a theory of mind
Hatim Chergui, Mehdi Bennis, Mérouane Debbah and colleagues argue that LLM agents managing the 6G RAN cannot treat messages from peer agents as objective facts. Each message is a trace of the sender’s reasoning, so a syntactically valid report can carry a hallucination and trigger a cascading outage that protocol validation would never catch. The paper models agent interactions as cognitive channels on a cellular sheaf and derives five design principles, including a continuous “cognitive SNR” measure of trust and a rule that peer modelling should halt at exactly two levels.
Why this Matters: A signalling-storm study on locally deployed 1B-parameter telecom language models shows the cognitive SNR isolating a hallucinating peer that three of its four neighbours agreed with. That is a concrete failure mode for the multi-agent architectures now being proposed for autonomous networks, and a reminder that resilience must be designed into agent-to-agent trust rather than assumed from message validity.
View source material to read more
Agentic AI layered onto the GSMA Operator Platform
Claudia Carballo González, Hatim Chergui, Christos Verikoukis and colleagues propose an agentic intelligence extension to the GSMA Operator Platform, using ETSI’s open-source OpenOP as the reference implementation. Autonomous agents hold persistent service context and drive closed-loop control through CAMARA APIs, with a declarative monitoring and alerting system added to the stack and a decentralised negotiation protocol for cross-domain conflicts. A URLLC and eMBB coexistence case is used to show agents reconciling SLAs across federated domains while improving energy efficiency.
Why this Matters: Operator platform federation is where multi-operator 6G services will be exposed, and today’s frameworks are stateless. This paper is one of the first to take the OpenOP release we covered in March and build agentic autonomy on top of it, giving the Network-as-a-Service model a reference architecture aligned with existing standards.
View source material to read more
Tokens as the unit of 6G communication
A large author group including our good friend Rahim Tafazolli, Tony Quek and Ping Zhang surveys the shift from bit delivery to meaning-aware connectivity and argues that the token, the native processing unit of large models, is the missing abstraction. Unified multimodal models already encode text, images, audio, video and robot actions in one token space, and distributed inference already generates token-level traffic. The paper traces the path from large-model-driven semantic communication to what it calls TokenCom.
Why this Matters: Semantic communication has lacked a common unit, which has limited interoperability and theoretical unification. Proposing the token as that unit lets importance weighting, error handling and resource allocation operate at token granularity, which is a plausible route to AI-native design. It remains a survey and a proposal, and the physical-layer implications are far from settled.
View source material to read more
HAPS-mounted RIS beats relays under real-world impairments
Bilal Karaman, Halim Yanikomeroglu and colleagues compare high-altitude platform stations equipped with a reconfigurable intelligent surface against those using a conventional relay, under NOMA multiple access with hardware impairments and imperfect channel knowledge. The RIS configuration delivers better sum-rate and energy efficiency because its passive reflection avoids noise amplification. RIS is more sensitive to channel estimation errors, but large element counts compensate.
Why this Matters: HAPS are a live topic in 6G NTN and this is a rare study that models impairments rather than ideal conditions. The finding that passive RIS retains its advantage under those conditions strengthens the design case for RIS-assisted aerial platforms, though field validation of RIS at altitude remains absent.
Guy, I’d make the existing 5G user part of the MRSS acceptance test. Same spectrum, same offered load, then progressively increase the 6G traffic share and measure what happens to the 5G users: goodput, latency under load and session continuity. An aggregate capacity gain could still hide a worse experience for part of the installed base. I’d also want the exact radio hardware and software baseline alongside the results. Otherwise, “hardware reuse” is difficult to turn into a deployment decision.