Industry Outlook: Telecoms & Connectivity — Week of May 25, 2026

Market Outlook

• AI-native RAN trials hint at capex-light efficiency. AT&T and T-Mobile’s tests of Ericsson’s AI-native link adaptation in the RAN—delivering spectrum and performance gains without GPUs—signal a pragmatic AI path that avoids today’s accelerator cost and power constraints. For CTOs, this validates near-term focus on embedded AI features in vendor roadmaps over large bespoke GPU-heavy RIC/SON builds.
• Copper retirement accelerates fiber and 5G densification. AT&T’s legal push in California to retire copper, paired with a $19B commitment to fiber and cell site buildouts, underlines how regulatory wins can unlock rapid fixed and mobile upgrades. Similar copper sunset programs already underway in 20+ states show a structural shift toward all-IP, fiber-first access footprints that will shape 5G/6G coverage economics.
• Rural broadband race tightens with FWA and fiber. Ethos Broadband’s fixed wireless build completion in rural New Mexico, plus fresh builds by TWN, Shentel, WOW and Comcast, underscore how BEAD-style funding and competitive overbuilds are compressing the window for Tier 3 operators to lock in share. Satellite broadband’s rise as a credible alternative further limits the margin for operational inefficiency in low-density markets.

Discussion: This week favors operators that can convert regulatory and funding tailwinds into rapid copper-to-fiber migration and AI-enabled RAN efficiency. Watch how quickly AI-native features move from PoC to standard software loads and reassess your rural playbook in light of intensifying FWA, fiber and LEO competition.

Headwinds

• Spectrum renewal sticker shock constrains 5G capex. Australia’s ACMA is demanding AU$5.22B in spectrum renewal fees, triggering operator warnings about higher tariffs and reduced network investment. This is a stark reminder that spectrum policy can abruptly rebalance cash flows away from 5G densification, Open RAN trials and edge builds, especially in mid-band-heavy markets.
• Labor cuts and AI backlash complicate automation. BT’s plan to cut at least another 27,500 jobs by 2030, after 8,500 last year, exemplifies the pressure to automate in the absence of topline growth. At the same time, emerging societal backlash against AI “slop” and job-loss fears—highlighted in sector commentary—raises reputational and regulatory risk for aggressive AI-driven workforce reductions.
• Smartphone softness clouds 5G monetization runway. Omdia’s forecast of a 22% smartphone market decline in the Middle East after a weak start to 2026 underlines slowing device replacement in key emerging markets. With Fierce noting the industry still lacks a clear 5G monetization story beyond basic connectivity, a weaker upgrade cycle makes it harder to justify premium 5G ARPU assumptions.

Discussion: CTOs should stress-test 5G and 6G roadmaps against higher spectrum and energy costs, slower device cycles, and constrained headcount. Prioritize automation that clearly improves service quality and opex while building a defensible narrative for regulators, employees and customers around “responsible AI operations.”

Tailwinds

• AI-native networks open new efficiency frontier. The AT&T/T-Mobile–Ericsson RAN trial, combined with growing interest in agentic AI for WAN optimization and AI-verified operations architectures, highlights a maturing toolkit for closed-loop optimization. These capabilities promise measurable spectrum gains, energy savings and faster fault resolution without wholesale hardware refreshes.
• Sovereign cloud and AI demand favor telco-cloud. Deutsche Telekom and SAP’s win on Germany’s sovereign AI project and the Microsoft–Proximus NXT sovereign cloud collaboration show governments leaning on telcos as trusted cloud and AI infrastructure partners. This aligns with rising demand for data residency, regulated-industry compliance and low-latency edge compute—natural adjacencies for telco-cloud and network-as-a-service offerings.
• LEO-to-mobile momentum expands coverage options. SpaceX’s Starship successfully deploying mock satellites and AST SpaceMobile’s BlueBird launch preparations—alongside Starlink Mobile’s broader ambitions flagged in its IPO S‑1—signal accelerating direct-to-device and LEO backhaul capabilities. For MNOs, this opens hybrid terrestrial–satellite propositions for maritime, aviation, remote enterprise and disaster recovery segments.

Discussion: To capitalize, align AI investments with vendor roadmaps and focus on concrete KPIs like spectral efficiency and MTTR. At the same time, deepen partnerships around sovereign cloud and LEO to position your network as the default fabric for compliant AI workloads and ubiquitous connectivity.

Tech Implications

• AI in RAN and WAN without GPU dependency. Demonstrations of AI-native link adaptation running without GPUs point toward lighter-weight inference models embedded in baseband and radio software. Combined with emerging multi-agent control-plane (MCP) architectures that verify agent actions before execution, this suggests a near-term focus on integrating vendor AI modules, telemetry pipelines and policy engines rather than building bespoke GPU clusters at every edge site.
• Global memory shortage reshapes telco-cloud design. Reports of a severe global memory shortage blowing up IT modernization budgets will directly impact large NFV, 5G SA core and AI buildouts. Architectures that assume generous DRAM/HBM footprints per vNF or AI workload will need to be revisited, pushing designs toward memory-efficient network functions, shared inference services and more granular workload placement across core and edge.
• Copper sunset drives all-IP, Wi‑Fi-centric access. AT&T’s copper retirement push and the Fiber Connect commentary that fiber’s value depends heavily on in-home Wi‑Fi performance both reinforce the need to treat access as an end-to-end IP service. This elevates customer-premises Wi‑Fi optimization, cloud-based CPE management and standardized service differentiation (e.g., slices, QoS tiers) as first-class engineering concerns alongside fiber and 5G radio planning.

Discussion: Engineering teams should re-baseline network designs for constrained accelerators and memory, adopting more efficient models, smarter telemetry and policy-driven automation. At the access edge, prioritize CPE/Wi‑Fi management platforms, IP-only service models and tight integration between RAN, transport and home networks to unlock the full value of fiber and 5G.

CTO Action Items

This week, prioritize a review of your AI-in-network roadmap: benchmark vendor-native RAN and WAN optimization features against your own AI initiatives, and reallocate spend toward those that deliver measurable spectrum and energy gains without large GPU estates. Revisit spectrum and capex plans under more punitive fee and device-replacement scenarios, ensuring optionality via Open RAN, shared infrastructure and LEO partnerships. On the telco-cloud side, accelerate work on sovereign, compliant cloud offerings and edge platforms that can host AI workloads, while adjusting designs for the current memory crunch. Finally, treat copper retirement and rural buildouts as catalysts to modernize end-to-end IP access, including managed Wi‑Fi and differentiated QoS, so that new fiber and 5G capacity translates into premium, monetizable services rather than undifferentiated bandwidth.