The hidden barrier to AI transformation: connectivity (www.techradar.com)

Despite widespread AI experimentation, adoption is being held back not by models but by connectivity. Ericsson’s State of Enterprise Connectivity research and industry reporting warn that legacy networks — slow, costly-to-upgrade fiber and unstable links — are constraining AI’s data-hungry workloads. Key findings: only 25% of AI projects hit expected ROI (IBM), 88% of European business leaders say new connectivity is necessary for AI benefits, 28% tie unreliable networks to revenue loss and 46% to rising operating costs. Organizations increasingly view 5G and Wireless WAN (WWAN) as the practical fix: low-latency, high-bandwidth cellular links support many IoT endpoints, enable real-time inference and cloud-offload, deploy faster than fiber in remote or urban sites, and help capture emissions and other distributed data critical for AI use cases. The relationship is reciprocal: 85% of firms report AI is already improving network performance through traffic prediction, automated resource allocation, anomaly detection and root-cause remediation. For the AI/ML community this means two imperatives—treat connectivity as a strategic axis for model deployment (edge vs. cloud, latency budgets, telemetry) and leverage AI to optimize network capacity and resilience. Faster, managed 5G/WWAN infrastructure will unlock time-sensitive applications (emergency response, live video analytics) and sustainable monitoring, while network-aware ML will maximize ROI across enterprise deployments.