The transition from 5G to 6G will be far more than a speed upgrade. It represents a fundamental shift in how networks are designed, operated and monetized. 6G is expected to bring intelligence and sensing into the core of the radio system, redefine spectrum strategies and reset energy and cost models. In the sections that follow, I’ll highlight the breakthroughs likely to accelerate 6G technology innovation, the under-the-radar developments that could surprise the industry and the technical challenges that must be solved for 6G to achieve International Mobile Telecommunications-2030 (IMT-2030) ambitions.

BREAKTHROUGHS DRIVING 6G FORWARD

AI-Native Networks, from Optimization to Core Design Principle

Across industry discussions, one theme is clear: AI moves from a bolt-on feature to a foundational element of network design. In 5G, AI is mostly applied at the network level and through RAN intelligent controller functions; in 6G, it infuses the RAN and the physical (PHY) layer. This shift demands standardized workflows for training, model exchange and on-device inference; defining key performance indicators (KPIs) that measure accuracy, latency, energy and memory; and interoperable interfaces that allow AI models to operate across multi-vendor environments.

The immediate value of two-sided AI for the PHY lies in cooperation between the base station and user equipment (UE) to compress and enrich channel state information, steer beams faster under mobility and apply AI to joint source–channel coding. These capabilities are not science projects; they are being scoped precisely because traditional algorithms are hitting complexity limits in frequency range 3 (FR3) with thousands of antenna elements and near-field effects. AI will also be instrumental in the RAN, improving spectrum usage, reducing costs and enhancing energy efficiency through AI-for-RAN, maximizing infrastructure utilization with AI-and-RAN and enabling new services and applications in wireless networks with AI-on-RAN. Agentic AI will likely play an essential role in orchestrating optimization policies across layers as well.

The philosophical shift is as important as the technical one: We are no longer optimizing fixed stacks; we’re designing adaptive stacks that learn. But “AI-native” only works if we can test, benchmark and trust what we deploy. Testing and validating AI-native designs under real-world impairments will require advanced measurement solutions and expertise from partners who understand both wireless and AI domains.

New Spectrum, New Rules: FR3 Takes the Lead

Early 6G enthusiasm gravitated toward sub-terahertz frequencies. Reality has since redirected focus toward FR3, the centimeter-wave (cmWave) span between frequency range 1 (FR1) and frequency range 2 (FR2). Why? Because it’s where we can access more bandwidth with viable coverage from existing macro sites, if we scale antenna elements and beamforming, while improving energy efficiency.

The FR3 agenda centers on three imperatives:

Coverage parity from existing towers: Array sizes will jump into the hundreds or thousands of elements to maintain link budgets. This pushes innovation in front-end efficiency, thermal design, calibration and near-field beamforming.

Coexistence and sharing: FR3 is crowded with incumbents, including satellite communications, earth exploration and defense radars. Expect sharper filters and guard bands, interference cancellation and dynamic sharing policies across terrestrial and non-terrestrial networks (NTNs) to be as strategic as raw throughput.

Global harmonization: Fragmented allocations strand device volumes and delay coverage. With the World Radiocommunication Conference 2027 (WRC-27) on the horizon, harmonization is not a nice to have, it’s foundational to 6G market adoption.

Hardware That Changes the Game

Breakthroughs in RF photonics and heterogeneous integration are moving from papers to prototypes. Multi-frequency, wideband photonic front-ends promise lower loss, reconfigurable links spanning microwave to mmWave, while tighter complementary metal oxide semiconductor (CMOS) integration of “RF + mixed-signal + control” shrinks size, weight, power and cost at the radio head.

At the same time, reconfigurable intelligent surfaces (RIS) and new materials may make extreme MIMO practical. If RIS clears hurdles in calibration, control, reliability and cost, it can reshape the coverage/energy tradeoff, especially in FR3, by steering fields rather than only boosting power.

UNDER-THE-RADAR DEVELOPMENTS TO WATCH

Energy as a first-class design constraint. Beyond clever schedulers, watch for AI-driven hardware adaptation: deep sleeping entire RF chains and antenna subarrays, envelope tracking, fine-grained voltage/frequency scaling and traffic-aware wake/sleep orchestration. Energy is the budget that must go down even as antenna counts go up.

Digital twins and high-fidelity simulation. As antenna counts, NTN links, RIS panels and sensing overlays multiply, physical trial-and-error becomes less viable. Digital twins grounded in impairment-rich models are becoming necessary to de-risk PHY choices, coexistence policies and city-scale deployments before field trials.

Standardization of AI practice. Less glamorous than the hype around AI, but critical: Shared datasets with realistic impairments, transparent model documentation, repeatability requirements and interoperable model exchange formats so a base station’s encoder can seamlessly interact with a UE’s decoder.

THE HARD PROBLEMS WE MUST SOLVE

6G represents a new frontier in wireless network possibilities. Before it can be explored, we must first overcome numerous challenges, including:

Spectrum: Coexistence, Sharing and Harmonization

Engineering is inseparable from policy. Coexistence with satellite and radar incumbents will require smarter sharing and tighter RF hygiene via advanced filtering and guard strategies. Multi-radio spectrum sharing (MRSS) will be critical to dynamically balance terrestrial and non-terrestrial demands across FR3 and beyond. NTN–TN integration introduces cross-layer interference challenges, demanding rigorous guard-band planning, power coordination and interference cancellation. Global alignment, driven by WRC-27 decisions and regional policy frameworks, will ultimately dictate FR3 availability and the pace of ecosystem maturity. While FR3 leads the pragmatic curve, FR2 will persist as the go-to band for ultra-high-capacity hotspots and specialized deployments, complementing FR3’s broader coverage ambitions.

AI in the PHY and RAN: From Promising to Proven

AI will only earn its keep if it survives the brutal constraints of radio:

Data fidelity: Training on impairment-rich datasets that include phase noise, in-phase and quadrature imbalance, power amplifier distortion and realistic interference. Synthetic-only training is insufficient; hybrid datasets are table stakes.

Explainability and repeatability: Engineers must be able to trace decisions and expect deterministic behavior under identical conditions. That means model introspection and robust evaluation protocols.

Latency, complexity and energy: Models must meet sub-millisecond deadlines and handset power budgets. Expect aggressive compression, quantization, pruning and sparsity, co-designed with hardware accelerators.

Standardization and interoperability: Without shared model interfaces and metadata, two-sided AI cannot scale across vendors. KPIs must measure not only link gains but also compute/energy overhead.

Testing Ultra-Massive Arrays

With 1000–2000 element arrays on the table, testing becomes a first-order challenge. OTA strategies must handle near-field conditions, fast phase-coherent calibration across subarrays and production-rate test cycles without breaking the economics. Expect multi-probe chambers, near-field to far-field transformations and new system identifiers for array health monitoring in the field.

Built-In Security

Every generation has treated security as a layer; 6G must treat it as fabric:

• Zero-trust architectures with hardware roots of trust, secure boot and continuous attestation from chip to cloud.
• Quantum-safe cryptography and crypto-agility, because lifecycle horizons exceed the time to a practical quantum threat.
• AI-era threats such as data poisoning, model theft, adversarial perturbations and cross-domain exploits necessitate AI for defense (anomaly detection, behavior modeling) as much as AI for performance.
• PHY-layer resilience: Anti-jamming, spoofing resistance, channel-based keying and beam-level privacy, especially as FR3 and FR2 enable highly directive links that can be both a shield and an attack vector.

Selecting and Proving First-Wave ISAC

The most compelling ISAC early wins cluster where macro-footprints become macro-sensors:

Critical infrastructure and smart cities: Continuous structural health monitoring of bridges and tunnels, road surface change detection, crowd density and traffic compliance. Centimeter-level positioning transforms municipal operations and emergency response.

Autonomy and advanced mobility: Cooperative perception across vehicles and infrastructure, network-assisted detection for drones and micro-mobility and higher confidence in pedestrian and cyclist protection through fused sensing.

Industrial automation and healthcare: Millimeter-precision tracking for collaborative robots and logistics, non-contact fall detection and vital-sign proxies for elder care. Crucially, these use cases monetize sensing data, creating revenue paths beyond connectivity.

Aerial and NTN Scenarios: Unmanned aerial vehicle traffic management and NTN–TN handover sensing, where persistent awareness of airspace users becomes a prerequisite for safety and spectrum hygiene.

Pragmatic Steps for Industry Readiness

1) Align early on AI interfaces and KPIs

If two-sided AI is to scale, the industry must converge on model exchange formats, metadata schemas and evaluation KPIs that include latency, energy and compute overhead alongside link-level gains. Interoperability beats isolated excellence.

2) Invest in FR3 field data

City-scale propagation and interference datasets for FR3 spanning varied climates, morphologies and incumbents are essential. Real-world data will shape guard bands, filter specifications and interference cancellation strategies far better than simulations alone.

3) Treat energy as a design input, not an outcome

From baseband to RF to antenna controls, plan for AI-driven sleep orchestration and adaptive duty cycling. Publish energy KPIs with the same rigor as throughput and latency and make them pass/fail gates for features.

4) Operationalize digital twins

Use them to iterate on PHY choices, ISAC tradeoffs, RIS placement, NTN/TN handovers and city-scale deployments before field trials. Close the loop by feeding models with real impairment data captured from live deployments.

5) Make security foundational and adversarial

Adopt zero-trust by default. Run red-team exercises that include AI adversaries and NTN vectors, and practice crypto-agility to rotate algorithms without ripping infrastructure. Integrate model provenance and attestation into AI pipelines to know what has been deployed and whether it is intact.

6) Embrace multi-party demonstrations

Some of the most valuable 2026 milestones will be interoperable demonstrations. Operators, infrastructure vendors, chip suppliers and test and measurement partners must collaborate and leverage shared validation frameworks to accelerate multi-party proofs. Confidence, not perfection, is the immediate goal.

CONCLUSION: BEYOND SPEED

6G is not a race to the highest data rate; it’s an opportunity to rethink wireless network capabilities. The near-term reality is FR3 pragmatism, AI-native workflows and sensing that turns ubiquitous coverage into ubiquitous awareness. If we can harmonize spectrum, tame the energy curve, standardize AI practice, scale testing for ultra-massive arrays and build security in from the start, 6G will earn its next-generation status not by headline speeds, but by system-level intelligence and reliability.