MWC 2026 made one thing clear: the industry is done with flashy demos and is building the plumbing to run AI across real networks.
At the show, chip designers, infrastructure vendors and software houses spent less time on proofs of concept and more on practical steps—shared road maps, go‑to‑market partnerships and co‑funded pilots—designed to move carriers from lab experiments to paid deployments. The conversation centered on engineering whole systems, not just individual components: silicon that understands radio workloads, software stacks that span cloud and edge, and business models that lower the financial and operational risk for operators.
Three themes dominated the announcements
– Tight silicon–radio integration. Chipmakers and radio vendors are aligning road maps so processors and RF systems are optimised for AI tasks like inline packet processing and low‑latency inference. That promises more efficient on‑site inferencing and less friction when integrating new hardware into live networks.
– Cloud-to-edge AI stacks. Vendors showed software designed to push models out to the edge—closer to users and sensors—while keeping orchestration, telemetry and security consistent across sites.
– Financing and service playbooks. Suppliers are offering managed services, reference deployments and joint investments so operators—particularly those without vast CAPEX—can adopt edge AI without shouldering all the risk.
Who’s best positioned
Tier‑1 operators with extensive edge footprints and big CAPEX programs will be first movers. Smaller carriers will likely adopt managed offerings or shared infrastructure. In every case, early success will hinge on end‑to‑end readiness: validated stacks, production verification and measurable SLAs for latency, resilience and automation.
Chips and telco compute: AMD, Intel, Nvidia and friends
Leading silicon suppliers framed 2026 as the year they translate telco requirements into product lines. AMD expanded its Ryzen AI family with new 400‑series variants aimed at mobile workstations and distributed edge use. Intel doubled down on secure, cloud‑native building blocks with Ericsson to push research into commercially viable 6G components. Nvidia released telco design guides and orchestration patterns to make large telco models and automation more operationally tractable.
Across vendors the message was the same: interoperability matters. Coalitions, open‑source reference architectures and cross‑vendor testbeds are meant to shorten integration cycles and provide repeatable blueprints operators can trust.
What operators need to do now
– Run pilots tied to clear KPIs (revenue uplift, cost reduction, SLA improvements) and prefer coalition‑backed references.
– Adopt orchestration frameworks that support containerised, multi‑agent workflows and multi‑vendor interoperability.
– Harden supply chains and edge platforms: secure boot, telemetry encryption and platform attestations should be baseline requirements.
– Reskill teams for model lifecycle management and hybrid cloud operations.
– Negotiate procurement terms that mandate third‑party interoperability testing and measurable SLAs.
Connectivity and integrated telco stacks
A noticeable shift at MWC was the consolidation of formerly separate functions—connectivity, routing, orchestration—into unified, composable stacks. Vendors like Qualcomm pushed new connectivity silicon (Wi‑Fi 8, FastConnect 8800, Dragonwing platforms and the X105 5G Modem‑RF) engineered for higher throughput and lower latency for on‑device and access‑point AI workloads. HPE and Juniper demonstrated converged portfolios that combine modular routers with agentic AI routing controls and dense compute chassis to reduce footprint and latency.
Adoption speed will depend on three things: published interoperable interfaces and validated test plans, procurement models that reward modular upgrades, and workforce reskilling to manage distributed AI workloads. Standardised stacks can lower TCO by sharing silicon across functions—but they also spread risk if reference designs carry vulnerabilities. Early interoperability testing and third‑party validation will therefore be strategic priorities.
Software, security and ecosystem commitments
Software vendors and cloud providers emphasised integrated stacks that bundle multi‑function servers, cloud‑native routing and orchestration—deployed close to traffic at carrier and edge data centers. Security was a headline: suppliers are standardising endpoint hardening, encrypted telemetry, secure boot and joint vulnerability processes to reduce cascading risks.
At the show, chip designers, infrastructure vendors and software houses spent less time on proofs of concept and more on practical steps—shared road maps, go‑to‑market partnerships and co‑funded pilots—designed to move carriers from lab experiments to paid deployments. The conversation centered on engineering whole systems, not just individual components: silicon that understands radio workloads, software stacks that span cloud and edge, and business models that lower the financial and operational risk for operators.0
At the show, chip designers, infrastructure vendors and software houses spent less time on proofs of concept and more on practical steps—shared road maps, go‑to‑market partnerships and co‑funded pilots—designed to move carriers from lab experiments to paid deployments. The conversation centered on engineering whole systems, not just individual components: silicon that understands radio workloads, software stacks that span cloud and edge, and business models that lower the financial and operational risk for operators.1
At the show, chip designers, infrastructure vendors and software houses spent less time on proofs of concept and more on practical steps—shared road maps, go‑to‑market partnerships and co‑funded pilots—designed to move carriers from lab experiments to paid deployments. The conversation centered on engineering whole systems, not just individual components: silicon that understands radio workloads, software stacks that span cloud and edge, and business models that lower the financial and operational risk for operators.2
At the show, chip designers, infrastructure vendors and software houses spent less time on proofs of concept and more on practical steps—shared road maps, go‑to‑market partnerships and co‑funded pilots—designed to move carriers from lab experiments to paid deployments. The conversation centered on engineering whole systems, not just individual components: silicon that understands radio workloads, software stacks that span cloud and edge, and business models that lower the financial and operational risk for operators.3
At the show, chip designers, infrastructure vendors and software houses spent less time on proofs of concept and more on practical steps—shared road maps, go‑to‑market partnerships and co‑funded pilots—designed to move carriers from lab experiments to paid deployments. The conversation centered on engineering whole systems, not just individual components: silicon that understands radio workloads, software stacks that span cloud and edge, and business models that lower the financial and operational risk for operators.4