AT&T and AWS pivot telco AI network operations toward revenue growth

The shift from cost-saving chatbots to agentic AI for network management signals a maturation in how carriers view automation. By integrating AWS Context, telcos like BT and NTT Docomo aim to solve the unique complexity of wireless data, turning technical efficiency into a competitive pricing advantage. For the broader streaming ecosystem, smarter networks mean better QoS management and the potential for dynamic bandwidth tiering that could impact how high-bitrate content is delivered and billed. As carriers move legacy systems to the cloud using AWS Agentic Services, the industry should watch for a rise in personalized connectivity offers tied to real-time network performance metrics.

The push toward AI-native network operations is intensifying across the vendor landscape, with Ericsson positioning its own silicon and software stack as a direct alternative to cloud-hyperscaler approaches. At MWC Barcelona in March 2026, Ericsson demonstrated AI-native Link Adaptation with AT&T on an Intel Xeon 6 SoC-based Cloud RAN stack, achieving up to 20 percent higher downlink throughput compared to legacy rule-based link adaptation. The demo marked the first call with portable Ericsson AI software on AT&T's target cloud RAN configuration, underscoring how carriers are evaluating multiple paths to the same goal of autonomous network optimization that AWS Context targets from the semantic-layer side. Ericsson's strategy diverges from the AWS model by embedding AI inference directly into purpose-built radio hardware rather than relying on cloud-hosted semantic layers. The company launched new radio units containing neural network accelerators within its Ericsson Silicon beamforming chips, enabling real-time matrix computations for RAN processing without discrete accelerators. Erik Ekudden, Ericsson's head of networks strategy, framed the approach as analogous to Nvidia's GPU evolution, arguing that telecom's parallel compute architecture benefits from on-site AI inference in massive MIMO radios. This hardware-first philosophy contrasts with AWS's plan to provide the data and context layer that carriers need to build agentic AI traffic scaling workflows on top of existing infrastructure. Field validation of AI-native RAN features is accelerating, giving operators concrete performance benchmarks to justify investment. Bell Canada and Ericsson conducted the world's first field test of AI-native link adaptation in April 2025, delivering up to 20 percent higher downlink throughput and 10 percent improved spectral efficiency in live network conditions. The technology, developed at Ericsson's Ottawa R&D site, integrates AI throughout its sub-components rather than as an add-on, executing on the baseband unit in real time. Light Reading reported that Ericsson is pursuing AI-RAN without Nvidia dependency, signaling a broader industry move toward self-contained AI compute that could pressure hyperscalers like AWS to prove their semantic-layer services deliver comparable gains in operational autonomy. As these deployments scale, to ensure interoperability across multi-vendor environments.