Cisco’s Chief Architect for Provider Mobility, Bob Everson, recently testified before a U.S. Senate subcommittee regarding the impact of artificial intelligence (AI) on network infrastructure. During the hearing titled "Intelligent Networks: Powering Artificial Intelligence and Transforming Communications," he emphasized the profound transformation AI is causing within enterprise and service provider networks, necessitating an evolution to handle AI workloads effectively.
The subcommittee hearing was aimed at examining the complexities introduced by the rapid adoption of AI, alongside the requirements for upgrading network infrastructure to accommodate increased demands. Senator Deb Fischer, chairing the subcommittee, noted the necessity for networks to expand capacity and complexity to efficiently support AI applications.
Everson reported significant changes in network traffic behavior due to AI, highlighting a fourfold increase in AI inference traffic over a mere eight months. Traditional networking strategies, which often prioritize downstream content flow, face challenges as AI applications generate more two-way and uplink-intensive traffic. In real-world terms, a typical AI agent could produce 450% more traffic than a human performing the same tasks.
At present, campuses and branch networks, including those supporting the Senate office building, experienced a reported 34% uptick in AI workload traffic over the last year, with expectations for a staggering 96% increase in the upcoming year. Furthermore, about half of enterprise customers cite AI traffic demand concentrated on their Wi-Fi networks, with a majority anticipating capacity issues within 24 months.
Everson outlined the critical factors influencing AI’s impact on networks, which include:
- Infrastructure: The rise of edge computing driven by AI traffic profiles necessitates new considerations around costs and data locality.
- Technical challenges: Applications such as robotics and autonomous vehicles may require sub-millisecond responses, making traditional round-trip latency unviable.
- Data management: AI generates massive datasets, making centralized processing costly and congestive.
- Security concerns: Many enterprises are apprehensive about transferring sensitive information over public networks due to regulations and vulnerabilities.
Despite these hurdles, the potential benefits of AI-driven networks are significant. Cisco maintains that AI can enhance network performance, resilience, and security. Everson pointed out that AI-driven networks can provide self-healing capabilities, allowing systems to auto-adjust in response to performance issues, thus improving uptime for mission-critical services.
However, a pressing challenge lies in the talent deficit in managing increasingly complex software-defined networks. Everson advocated for the automation of routine tasks through AI, enabling junior operators to take on more responsibilities and allowing experienced professionals to concentrate on strategic innovations.
As networks transition toward AI-native platforms, they are expected to evolve from mere conduits to essential components of intelligent connectivity. Cisco anticipates that integrating AI with network functions, like the Integrated Sensing and Communication (ISAC) system, will advance autonomous and safety-driven applications significantly.
In concluding remarks, Everson proposed three actionable steps for the committee:
- Accelerate the U.S. AI-native stack by investing in and developing infrastructure that supports both immediate 5G advancements and the future introduction of 6G capabilities.
- Modernize regulatory frameworks for permitting and infrastructure to facilitate responsible expansion of AI-enabled networks.
- Ensure a balanced spectrum policy to support high-capacity connectivity required by growing enterprise needs.
This testimony reflects a pivotal moment as policymakers grapple with the exponential changes AI is heralding across the telecommunications landscape.
