What's Happening?
AT&T's Chief Technology Officer, Yigal Elbaz, has expressed skepticism about the viability of AI compute at the 'far edge' of networks, a concept that involves deploying AI capabilities at remote network locations
such as radio access network (RAN) cell sites. This stance was shared during the New Street Research and BCG Global Connectivity Leaders Conference. Elbaz highlighted that while there is significant investment in AI infrastructure across the U.S., with $650 billion planned for data centers, the necessity of extending AI compute to the far edge to save milliseconds of latency is questionable. Instead, AT&T is focusing on leveraging its fiber and wireless networks to provide a 'deterministic experience' for new use cases, connecting them intelligently to the necessary models and infrastructure. This approach contrasts with other telcos like T-Mobile, which supports the AI-RAN concept, aiming to deploy AI capabilities at cell sites.
Why It's Important?
The skepticism from AT&T's CTO regarding AI compute at the far edge could influence the strategic direction of telecommunications companies in the U.S. The decision to focus on centralized AI infrastructure rather than distributed edge computing could impact how telcos allocate resources and develop new technologies. This stance may affect the adoption of AI-RAN, a concept that has seen limited support from major carriers. The implications are significant for the telecom industry, as it navigates the balance between investing in cutting-edge AI technologies and managing costs and complexity. Companies that choose to invest heavily in edge computing may gain a competitive advantage in offering low-latency services, while those like AT&T may benefit from a more centralized approach.
What's Next?
AT&T's current strategy involves preparing its network to be 'AI-ready' and 'AI-native,' focusing on integrating AI capabilities within its existing infrastructure. The company is collaborating with Cisco and Nvidia on an AI grid for enterprise IoT, which aims to bring AI inference closer to data generation points. This approach may lead to further developments in public safety and enterprise site security applications. As the industry continues to evolve, other telcos may reassess their positions on AI compute at the edge, potentially leading to a divergence in strategies. The ongoing debate over the best approach to AI deployment in telecom networks will likely continue, with companies weighing the benefits of edge computing against the challenges of implementation.
