Table of Contents
AT&T, BT, Orange and Verizon joined the 6G Forum to describe a disciplined approach to network evolution that protects existing investments, defines commercial value and reserves architectural change for capabilities that genuinely require it
The mobile industry’s technical roadmap to 6G is becoming clearer. 3GPP’s Release 21 schedule calls for the first normative 6G specifications to reach a functional freeze in December 2028, with initial commercial systems generally anticipated around the turn of the decade. But a standardization deadline is not an investment thesis. Operators still have to determine what 6G will do that 5G-Advanced cannot, how much of their existing infrastructure they can retain and, critically, who will pay for the difference.
Across discussions at RCRTech‘s recent 6G Forum (available on demand here), representatives of AT&T, BT, Orange and Verizon articulated a recognizable, if not unanimous, position. 6G should evolve from 5G without obligating operators to rebuild networks on a generational timetable, while making room for capabilities that deliver meaningful new value. The central tension is that emerging artificial intelligence (AI) applications and integrated sensing and communications (ISAC) may demand precisely the architectural changes that operators would otherwise prefer to avoid.
BT Senior Network Architecture and Standards Manager Rana Kamill put the commercial lesson plainly. “Technology adoption doesn’t guarantee business success by default,” she said. Her point was not that 5G delivered no value. Capacity expansion, modernization and the success of fixed wireless access in some markets is clearly significant. Rather, deploying new capabilities has not consistently translated into the differentiated services and enterprise revenue once expected of 5G. With 6G, commercial relevance has to inform the design process itself.
Evolution without mandatory replacement
AT&T Labs’ Milap Majmundar described a migration strategy built around flexibility. Multi-RAT spectrum sharing (MRSS), under discussion for the 5G-to-6G transition, is intended to avoid some of the compromises associated with dynamic spectrum sharing (DSS) between LTE and 5G. Majmundar pointed to the absence of LTE’s fixed common reference signals in 5G NR and cited work to limit shared downlink control-channel overhead to below 3.5%. Taken together, he argued, those characteristics should permit more efficient coexistence than the earlier DSS approach. The overhead figure reflects his account of ongoing standards work, not a demonstrated commercial-network result.
The more important implication is economic. Where deployed 5G radios have sufficient processing headroom and appropriate hardware, Majmundar said, a software upgrade to support the baseline 6G air interface may be possible. Operators could then purchase new radios for capabilities or spectrum bands that warrant them rather than because a new generation has arrived. From his point of view, the objective is “giving operators the choice rather than restricting any capabilities of 6G.”
Orange’s Alok Srivastava expressed the same principle as an sequenced operational movement from 5G to 5G Standalone to 5G-Advanced to a 6G-ready network and then 6G. “We will go step by step.” That does not eliminate hardware investment. New requirements associated with ISAC, non-terrestrial networks (NTNs) and AI will create equipment and ecosystem dependencies. It does, however, make replacement selective rather than automatic, with 5G and 6G coexisting according to service requirements.
While this conclusion was drawn from speakers at the 6G Forum, it’s aligned with broader industry discourse. In its September 2026 migration guidance, the operator-led NGMN Alliance treated MRSS as a preferred direction while stressing unresolved questions about implementation, hardware impact and total cost of ownership. The emerging operator demand is for optionality marked by modernization driven by measurable benefits, not a synchronized infrastructure refresh.
AI-native should describe an operating model, not a marketing claim
The same evolutionary logic applies to AI-native 6G. Verizon CTO Yago Tenorio described a network already moving from conventional automation toward Level 4 autonomy. Verizon executed roughly 70 million closed-loop automations during 2025, he said, but the challenge now is to move beyond predefined responses toward systems capable of handling circumstances for which they were not specifically programmed.
That change has less to do with a new air interface than with difficult work across legacy operational support systems, network functions and management interfaces. “Autonomy will not be an afterthought. Autonomy will be embedded natively,” Tenorio said. His expectation is that 6G functions will join an autonomous operating environment already developed through 5G, rather than requiring engineers to retrofit a parallel system after launch.
Orange’s emphasis on AI-based optimization (radio-resource utilization, load balancing, user-plane decisions and service availability) makes the operational rationale tangible. Those functions must create value on the installed network well before 6G is widely deployed. As explored in our earlier note on what AI-native 6G actually means, the distinction is not about AI-enhanced 5G versus AI-enabled 6G; it is whether intelligence becomes part of how the communications system is designed and managed. NGMN’s August 2026 autonomy report likewise identifies governance, trust, interoperability and operational transformation as prerequisites for credible Level 4 networks.
The uplink is where evolution may run out of room
Tenorio’s most provocative argument concerned a different role for AI that goes beyond a network-management tool, to a source of new connectivity demand. Today’s mobile networks were optimized largely for people consuming content. Context-aware agents, wearables and physical AI systems could instead transmit sustained streams of video, audio and sensor information upstream for inference. In that scenario, the traffic relationship between downlink and uplink changes materially.
Tenorio suggested that traffic patterns commonly described as 9:1 or 8:2 downlink to uplink could eventually move toward something resembling 1:1. That is a forward-looking scenario, not an established forecast. But the underlying constraint is credible. A wearable may need to send considerably more information while having less space for antennas, batteries and thermal management than a smartphone.
Some radio techniques, including subband full duplex, could help. Tenorio nevertheless argued, “We’re missing a breakthrough that will probably have to go all the way down to the architecture.” He raised the possibility of separating uplink and downlink deployment more fully, with additional, geographically pervasive uplink receivers positioned close to users. He explicitly offered that as an idea to investigate, not a settled Verizon design.
The concern builds on the traffic analysis in our September note, “AI flips the mobile network — the uplink wave is coming.” The observed increase in uplink demand is not yet proof that AI agents will drive a wholesale reversal in traffic symmetry. Indeed, current AI traffic remains a small portion of overall mobile volume. Yet network architecture must be planned before demand fully materializes. The question is whether AI-driven uplink requirements become the point at which incremental 5G evolution is no longer economically or technically sufficient. NGMN has put the broader issue on the standards agenda in AI Surge and its Implications for 6G.
ISAC: The distinction between demonstration and scale
ISAC exposes a related problem. BT’s Kamill identified sensing, digital twins and physical AI as opportunities to create outcomes beyond connectivity, but conceded that extensive use-case work has not yet identified which applications will reliably generate substantial operator revenue. 6G is being designed around prospective services before their economics are proven.
Verizon’s experiments demonstrate both the opportunity and the ambiguity. In September 2026, the company described 5G-Advanced-based trials involving crowd-density estimation and drone or vehicle sensing. If existing radio technology can already perform sensing, why should it be positioned as a defining 6G capability?
Tenorio’s answer was not that 6G invents radar-like sensing. Radio waves can already reveal features of their environment. The difference, he explained, would be the ability to industrialize sensing “without any compromise in the primary function of a network, which is providing connectivity.” In other words, 6G should allow operators to support sensing broadly while protecting the capacity and performance of the communications network. That is a design aspiration, not yet proof that all practical tradeoffs can be eliminated.
The distinction between a successful proof of concept and an economically scalable service could be key to ISAC as a breakout 6G use case. Sensing only becomes an operator business if it can be offered reliably, at defensible cost, without degrading existing services, and if customers see enough value to pay for it. That is a harder test than demonstrating that a base station can detect an object or characterize a crowd.
A different bargain with the 6G ecosystem
The operators’ positions converged more clearly on investment conditions than on a single winning use case. BT wants a defined problem and paying customer. Orange wants new value streams that justify capex. AT&T wants deployment flexibility and spectrum rules that permit operators to realize the value of their holdings. Verizon wants use cases, spectrum availability and globally consistent standards aligned before commercialization, while avoiding the architectural fragmentation that complicated parts of the 5G transition.
Kamill summarized the desired inversion of the old development sequence. “Let’s work on the business case and then build the technology around it as well.” For his part, Tenorio said that the industry has a narrow window to carry learnings from trials into the specifications, rather than standardizing around performance metrics and hoping attractive applications subsequently emerge. The ITU’s IMT-2030 framework includes new scenarios for AI and communication, ubiquitous connectivity and ISAC; those categories still need to become deployable services with credible economics.
There is a consequential distinction worth keeping in mind. Operators are not saying that 6G must be minimally different from 5G. They are saying the cost of difference needs to be justified. Continuity is valuable where existing infrastructure works; new architectures become necessary where they enable capabilities that cannot be delivered efficiently at scale. The hard work between now and 2030 is determining which side of that line each proposed 6G feature occupies.
