Why the first AI-native 6G cores will run on 5G foundations

wpnews.pro

We already know two big things about the 6G system architecture: it will be standalone from Day 1, and it will be "AI native."

Actually, make that three things: for some, 6G should also be "service aware."

In September 2026, 3GPP will make critical decisions on 6G RAN, core and end-to-end system architecture at its plenary meetings. Ahead of that meeting, this article discusses how the three high-level themes listed above might combine to shape the emerging 6G architecture.

Keep in mind there is still lots of uncertainty and detail to be decided. Nevertheless, to set the context behind the three themes:

Standalone Day 1 requires a discrete 6G core network (CN). What that means in practice is an open question. Is it a lightly upgraded version of a 5G core, a heavily evolved 5G core with lots of new 6G functions or an entirely new core with new infrastructure, functions, protocols and interfaces?

6G will be an AI native system. A new AI domain will be integrated into the CN in all the above scenarios. This is sometimes also described as an "AI plane" that will run across the system architecture to augment the classic control and user planes. The details of this AI domain/AI plane are a work in progress.

Service-aware networks. This is more nuanced because so-called intelligent networks have historically been outpaced by innovation in devices and applications. Nevertheless, new agentic and token-driven AI traffic profiles are likely to make service awareness important in 6G. (For background, see this 3GPP article on "Service characteristics awareness in the RAN," July 2026).

Related:Protecting spectrum for the 2028 Wi-Fi Olympics The concept of service intent draws these three themes together. The idea behind service intent is that the device signals its connectivity requirements to the network (for a given application), and the AI plane acts as an orchestration engine to configure the network accordingly. Crucially, service intent is driven by the end-user application – that is, what the customer wants.

In this architecture, the AI orchestrator controls fleets of agents both operating in the new AI domain and embedded within network functions (NFs; see the figure below). Service intent is thus central to the 6G core and AI plane/AI domain discussions.

For an extended version of this article, with more technical detail and context, download this free-to-view Omdia white paper: AI Native 6G System Architecture.

5G interworking for 6G Day 1 architecture

It appears likely that for most early deployments (say, around 2030), the 6G core will be based on an evolved 5G core infrastructure enhanced with new 6G core control plane functions and a new AI domain.

Related:[Ericsson and Nokia shed thousands of China jobs at likely 6G cost](/6g/ericsson-and-nokia-shed-thousands-of-china-jobs-at-likely-6g-cost)

For example, a new "6G AMF" may be introduced alongside combo 5G/6G nodes such as SMF and UPF (or some variation of new and combo nodes). And because voice and emergency calling are Day 1 requirements, this means reusing the 4G/5G IMS core. This initial 6G SA core will interwork with the existing 5G core, then evolve and refresh independently over time.

Does this risk a two-phase 6G akin to the NSA to SA transition we saw in 5G?

Sort of, maybe. Yes. The difference is that the architecture is SA from Day 1, and therefore the core can evolve quasi-independently of the RAN.

Introducing the AI domain/AI plane

The 6G architecture is likely to formalize a new AI domain in the CN. Based on Omdia's reading of 3GPP submissions and other market commentary, a draft architecture appears to be emerging, as shown in the figure below. In this architecture, NFs in the control and user planes are built with embedded agents that are orchestrated by the new agentic AI plane.

The elements and interfaces in the emerging 6G AI plane look a lot like existing hyperscaler and telecom vendor agent fabrics, and the emerging TM Forum Open Digital Architecture (ODA) AI-native canvas. See the full report for more details.

Emerging AI architecture for 6G core

The case for an all-new AI native 6G core

There is a strong technical case for an entirely new 6G core built as AI native from the ground up for Day 1 deployment. For example, the disaggregated architecture could align more closely to state-of-the-art AI compute, and NFs could be defined as larger functional blocks and connected over an agent fabric. Automation, monitoring, assurance and so on would follow suit. Potentially, an all-new AI native 6G core would allow the network to match service intent more efficiently to network resources and deliver optimal customer outcomes.

However, in practical terms, the supplier ecosystem will be challenged to deliver such a core in 2030 (although some vendors might have prototype and pilot products). Similarly, the early 2030s is too soon operationally and financially for operators to make such a large and radical shift in core infrastructure and operations.

Ironically, an issue with the evolution argument is that 5G core (and the service-based architecture) is now often viewed as overly complex, too brittle/fragile and overly disaggregated. Nonetheless, it is the baseline the industry must work from.

An all-new 6G core is thus appealing in theory, but is unlikely to be widely implementable in practice in the 2030 timeframe in a manner that is scalable and deployable for the global mainstream.

An EV analogy

Some argue that retrofitting AI to an evolved 5G architecture and calling it 6G SA is a losing proposition from the start. To use an automotive analogy, evolving an existing 5G core for AI native 6G is like building an electric vehicle (EV) on a combustion engine chassis. It works, but the EV native design will ultimately win. To extend the analogy further, one can only build a successful EV platform if one also builds the production line to produce it.

This automotive analogy does not map perfectly to 6G because nearly all mobile operators run brownfield infrastructure. Yet, it shows the need for vendors and operators to balance ambitious innovation with practical implementation constraints when developing AI native 6G core technology.

End users are the North Star

The potential for AI services is immense. This technology will expand the addressable market for mobile wireless technology and network operators, creating value for end customers. Wireless networks are fundamental to delivering services to AI-enabled endpoints and to realizing the full utility of AI in the wider economy and society.

The critical questions are about when and how this happens. Service intent (what users/customers want) is a useful guiding principle to determine what the architecture should look like, how to build it and how to operate 6G networks.