After a week of reporting by RCRTech and RCRWireless on Mobile World Congress in Barcelona, the biggest takeaway in terms of AI infrastructure is that AI is moving from the data center to the network edge, with a lot of news about the shift to “AI-native,” software-defined infrastructure designed for real-time applications like physical AI and autonomous agents. Throughout the week, it became clear that the industry is pivoting from simply using AI to optimize existing networks to rebuilding networks to embed GPU/NPU compute directly into the radio access network (RAN). For example, AT&T’s connectivity offering with AWS will prepare networks for the AI era – embedding last-mile fiber and 5G directly into the AWS environment and pushing AI into the RAN and expanding its edge ecosystem with Microsoft Azure. AT&T also announced it would work with Nvidia, Microsoft and MicroAI through its Connected AI platform for “smart manufacturing.” Korean operator SK Telecom unveiled a full-stack AI strategy, which it said would include expanding data center capacity to GW scale, collaborating with Schneider Electric and Supermicro to build modular, pre-fabricated data centers. SKT also announced a “sovereign AI package,” which integrates AIDC infrastructure and AI services tailored for industrial and enterprise use. Also important were the fiber-heavy data center interconnects, with companies like Cisco capitalizing on the east-west traffic surge for training workloads in hyperscale data centers, as covered below in “Top Stories.” Check out this week’s reporting by RCRTech and RCRWireless for news about how telecom operators are upping their AI infrastructure game – no longer sinking into the “dumb pipe” narrative that permeated the industry before the AI boom.