Wireless Infrastructure is AI Infrastructure – A Policy Event

View photos of the event at ai.wia.org

Welcome & Opening Remarks

• Cathy Piche (EVP & COO, Crown Castle; Board Chair, WIA) emphasized the urgency of preparing network infrastructure for the AI era, with a focus on double-digit data demand growth over the next 5 years.
  • Highlighted the imperative for the U.S. to lead both in realizing 5G and laying the groundwork for 6G.
  • Called for “forward-looking infrastructure policy,” ample spectrum, and innovation at all network layers.

The AI Revolution and Network Infrastructure

Keynote: Yago Tenorio (CTO, Verizon)

• AI is already integral: Verizon collects 70 billion data points daily to feed its AI engines, optimizing self-organizing networks, maintenance, and predictive planning.
• Societal impact: AI is reshaping sectors—healthcare, logistics, public safety, manufacturing, and consumer services.
• Infrastructure needs: AI demands networks with low latency, high resiliency, and vast capacity. Data movement for both AI inference and continual training is a core challenge.
• Silicon and power as bottlenecks: GPUs, power, space, and cooling are “new currency” in the AI race. Verizon’s footprint offers unique advantages—these resources are close to customers, reducing latency.
• Paradigm shift is coming (“iPhone moment”): The next interface leap is imminent—moving from screens (“rectangles”) to wearables/sensing devices (e.g., smart glasses).
  • Ubiquitous sensing and persistent context will transform AI’s capacity to augment daily life.
  • AI will become proactive: Not just responding to explicit queries, but supporting based on full context (“I saw you left your keys on your desk 15 min ago”).
• Role of policy: Differentiation of obligations between developers and deployers. Not all network infrastructure must bear the same AI-related obligations.
• Urgency of spectrum: New AI-driven devices will change traffic patterns, especially with 6G, requiring large contiguous spectrum blocks with favorable propagation (e.g., 400 MHz+ needed/channel). Current spectrum may carry early 6G, but not the full future load.
• International competition: Europe, China, and others already have 6G roadmaps; the U.S. must continue to secure spectrum and develop infrastructure to stay ahead.

Technical Deep Dive: AI’s Impact on Networks

Panel: Iain Gillott (WIA), Tony Sabatino (Diamond Communications), Akhil Gokul (Ericsson), Jonathan Lester (Airwavz)

1. AI Impacting Demand and Traffic

• Data demand is growing steadily, but a tipping point is coming:
  • AI-backed services are increasing uplink demand (image/video editing, wearables streaming).
  • Anticipated 20% uplift in uplink traffic due to AI, with compound annual growth rates for certain segments approaching 98% over five years.
• New use-cases: Edge AI—video analytics, predictive maintenance, healthcare, and smart manufacturing—drive both uplink and latency-sensitive traffic, previewing what’s to come.

2. Preparing Networks for the AI Age

• Infrastructure design:
  • Fiber-deep architectures inside buildings are necessary for future-proofing, especially in critical sectors like healthcare.
  • Upgrading to support more spectrum (e.g., C-band, 2.5 GHz) is essential but not yet universal indoors.
  • Network flexibility: Must allow for slicing and bandwidth allocation strategies specific to AI use-cases.
• Role of latency and uplink:
  • Latency is as important as bandwidth, especially for applications like robotics and remote healthcare (robotic surgery).
  • The AI revolution likely will reverse the traditional downlink bias, with uplink capacity becoming the limiting factor.
• Carrier engagement:
  • Wireless carriers may push more responsibility for in-building network investment onto enterprise/customers (except for high-value venues).
• Investment needs: Advanced radio hardware, edge compute integration, and dynamic network resource allocation (slice for AI traffic, even dynamic reallocation based on real-time traffic patterns).

3. AI in Network Operations & Engineering

• Existing AI tools:
  • Network design now informed by geofenced, real-time analytics (billions of points daily), replacing thousands of manual engineering tasks.
  • AI now assists with:
    • Detection of performance problems at the individual customer/location level.
    • Automated maintenance and self-optimization (e.g., antenna beamforming, interference reduction, traffic loading).
• AI in the RAN (Radio Access Network):
  • Native link adaptation for modulation/coding schemes, yielding 20% improvements in downlink throughput, 10% gains in spectral efficiency.
  • Centralized and distributed AI (in SMOs, O-RAN) used for load balancing and traffic optimization.
  • Proactive and predictive network support is evolving toward fully autonomous/cognitive networks.
• AI in operations/business:
  • Accelerates design workflows, automates support (customers and operations), and serves as a “force multiplier” for lean engineering teams.

4. Anticipated Evolution

• Near-term: “Human assisted → AI assisted” networks; increasing automation but with human oversight.
• Longer-term: Cognitive, self-healing, fully autonomous networks—optimization and maintenance with minimal or no human intervention.
• Full realization: Dynamic spectrum allocation and real-time traffic steering, including for high-density events.
• AI as workforce enabler, not replacer: Productivity and safety improvements, not full replacement of human roles.

Spectrum, 6G, and Policy

Policy Panel: Josh Koenig (SBA), Arpan Sura (FCC), John Kuzin (Qualcomm), Matt Pearl (CSIS)

1. Wireless Infrastructure as AI Infrastructure

• Consensus: Networks underpin AI’s edge, not just cloud; edge connectivity (low-latency, ubiquitous) is vital for future applications.
• Device-side intelligence: Hybrid models—device/edge and network-based AI—are necessary. Edge AI reduces network load and enables privacy.

2. Spectrum Roadmap and Urgency

• One Big Beautiful Bill: Recent legislation provides a framework and begins studies on wide bands, but this is just step one.
  • Current U.S. planning lags international efforts—other regions already have 6G deployment roadmaps.
• Action is urgent: Need spectrum allocations of 200–400 MHz per carrier for 6G by 2029–2030.
• Industry-federal collaboration is necessary to clear and reassign federal spectrum.
• Threats: Potential legislative language (e.g., NDAA amendments) could take needed bands off the table, hampering 6G planning.

3. Policy Priorities

• Codification of pro-infrastructure rules: FCC has advanced policies for efficient deployment (e.g., NEPA reform, local preemption), but Congress could further strengthen and clarify these.
• Infrastructure deployment bottlenecks: Regulatory “sclerosis” (NEPA delays, litigation) has slowed data center and wireless builds.
• Workforce development: Building an “AI ready” workforce is critical, both for telecom and broader economic leadership.

4. Regulatory Fragmentation

• Federal-state-local tension: FCC has strong preemption powers for telecom, but AI may trigger new jurisdictional disputes as infrastructure blends with AI obligations.
• Desire for a federal framework (via Congress) to avoid a regulatory patchwork that slows rollout or innovation.

5. Global Competition and National Security

• Race with China (PRC):
  • The competition is primarily with China, which can mandate technology deployment and spectrum assignment, accelerating their progress.
• National security at stake:
  • 6G and AI are both geostrategic and essential to military capability (commercial wireless for defense applications).
  • U.S. leadership is not just economic, but crucial to shaping global standards and geostrategic stability.

6. Metrics of “Winning”

• U.S. and allies set standards, deploy spectrum first, and dominate the global app/innovation ecosystem—akin to the U.S. 4G triumph.
• Leadership proven by domestic adoption, ecosystem scale, global exports, and geostrategic position.

7. Key Challenges & Debates

• Spectrum scarcity and allocation (who gets it, how quickly can it be deployed, international harmonization).
• Uplink vs. downlink demand: AI applications threaten to overwhelm uplink and edge-to-core capabilities.
• Latency and edge computing needs: Where should AI be processed—on-device, at the edge, or in central data centers? Each path has trade-offs.
• Trust and user adoption: “iPhone moment” for AI/AR will depend on public trust and seamless integration, not just technology readiness.
• Cybersecurity and data integrity: Increasing network AI increases attack surface and demands new governance.
• Policy agility: Pacing regulations and social adaptation to technology; risk of over- or under-correction.