6G Standardization Is Advancing, but Operators Are Planning for Evolution Rather Than Replacement

The 3rd Generation Partnership Project’s (3GPP) September RAN#113 meeting reported progress on several parts of the 6G Radio (6GR) physical layer: uplink channel coding, constellation shaping, higher-order modulation, and Synchronization Signal Block (SSB) design. This discussion took place after the detailed RAN1 discussion held during Release 20, with vendors still in the process of making their technical decisions regarding the commercial air interface.

These choices are important for the infrastructure because they impact different segments in the chain. The modulation and constellation shaping choices put pressure on radio linearity and baseband processing, and determine whether the implementation could be done in software or would require the development of new silicon. The uplink channel coding impacts the baseband compute power consumption. The SSB design impacts how 6G can coexist with 5G in shared spectrum and whether the migration could be done via software update only. In all of these cases, the decision determines whether the current hardware should be reused or should be decommissioned.

Release 20 is a study phase, with normative specifications expected in Release 21 and commercial launches around 2030, so operators' next investment decisions will be taken before the standard is final. The Next Generation Mobile Networks Alliance (NGMN) names Multi-Radio Access Technology (RAT) Spectrum Sharing as its preferred migration approach, which assumes 5G and 6G will coexist for a long transition. Ericsson positions 6G as an evolution of 5G Standalone (SA) and 5G-Advanced, and Nokia's AI-RAN work is already bringing Artificial Intelligence (AI)-native capabilities into existing infrastructure. Standardization will therefore decide how much of today's network carries forward and where genuinely new infrastructure is needed.

Considering this, the implementation of 6G standards is not expected to give rise to one single replacement cycle like the first rollout of 5G did. Significant investments have been made by operators into 5G SA core, cloud-native, and Software-Defined Radio Access Networks (SD-RANs), and decommissioning these for the sake of introducing new radio standards wouldn't be economically viable. The greater risk is that equipment bought today proves too inflexible to upgrade, forcing early replacement or leaving operators running parallel networks. Introduction is more likely to proceed through software upgrades and spectrum sharing, with hardware replaced only where needed. A radio that supports multiple RATs and additional spectrum, and can take on software functionality added later, will have a longer commercial life than one designed around a fixed 5G configuration. Operators with more advanced 5G cores and orchestration will also have less to replace when 6G scales.

Where physical replacement is necessary, spectrum is the main determinant. Low-band and mid-band spectrums would continue to work in terms of coverage and capacity, while upper mid-band and higher frequencies would require additional radio and antenna equipment. As a result, operators will implement the new layers in cases where capacity is needed, while leaving the rest of the RAN working as before. Site infrastructure (towers, power, fiber, and backhaul), 5G SA cores, orchestration layers, and software-defined low- and mid-band multi-RAT radios with extra processing capacity become the most likely to carry forward. The most vulnerable elements include fixed-function radios and basebands for one 5G configuration, upper mid-band and higher frequency radios and antennas, transports unable to transmit wide band, and AI accelerators and servers.

A second source of change lies in AI-native networking, with a tighter integration of the RAN with compute as network optimization and ultimately Integrated Sensing and Communication (ISAC) processing moves toward the edge. Nokia’s AI-RAN projects demonstrate that operators are already exploring such an approach, and compute is likely to become one of the fastest-growing elements of the infrastructure stack. The challenge lies in the mismatch in the lifetimes: radios will stay in place for 10 years or more, whereas accelerators and servers need to be replaced more often. Operators need to plan for the compute of the next-next generation, because retrofits are more disruptive than hardware refreshes, and separate their RAN software from the vendor’s hardware so as not to inherit its replacement cycle and pricing policies. Compute will have its own lifecycle and its own funding, more akin to Information Technology (IT) rather than RAN Capital Expenditure (CAPEX), and where inference will happen (cell site, aggregation site, or regional edge) is a design decision with its own trade-offs in latency, power, and compute utilization.

Constraints on energy are increasingly becoming a problem. AI will reduce energy requirements for the RAN, but energy and cooling required by its compute requirements become equally important. All this results in the market becoming one of constant upgrades and spending split between the RAN and the layers surrounding it rather than a singular hardware refresh.

6G will evolve gradually, so the key purchasing decisions must already be happening with regard to 5G-Advanced rather than at the outset of 6G. Flexible purchasers will have the opportunity to roll out upgrades in phases, whereas locked configuration purchases can double costs. The most useful step now is to write 6G requirements into current 5G-Advanced purchases without waiting for the full specifications. Priority should go to RAN equipment that combines multi-RAT capability, flexible spectrum support, and software upgradeability, along with 5G SA cores and orchestration platforms. This is a near-term priority, running to roughly 2028. It costs more up front, and some capability may go unused if standards shift, but the larger risk is early replacement. Operators can limit the downside by paying for headroom only on the sites with the strongest capacity case.

Spectrum planning needs to be integrated into equipment refresh cycles, as spectrum defines what kind of equipment is necessary. Flexible infrastructure may be deployed where there is enough certainty about future spectrum use, whereas more specific investments in radio and antennas may wait until milestones related to regulation and standardization, especially leading up to WRC-27. The latter need to happen in the medium term, around 2028 to 2030. Investing too early risks stranded assets, but investing too late risks falling behind competing solutions; milestone-based phased spending addresses both risks.

Compute needs to be considered its own infrastructure layer, as AI-RAN, edge inference, and ISAC will increase the demand for processing on a much shorter cycle than RAN itself. Site energy, cooling, connectivity, and capacity must be considered in terms of evolving compute platforms rather than just one AI platform. From 2030 onward, compute will be refreshed on its own cycle while 6G layers are added selectively. The main exposure then is energy, because compute load outgrowing efficiency gains could erode the business case, making total site efficiency a core metric. Vendors face the same logic from the supply side: a credible account of how an operator's existing infrastructure evolves into the 6G network will matter more than a newer radio.