Beyond Bandwidth

Mike Connaughton talks to Iain Farquharson, the Head of Pre-construction at Red Bear Technologies. Iain has a background in cognitive science and AI, as well as extensive experience in data center infrastructure projects. The two discuss:

• The history of AI and recent exponential jump in AI advancements  
• The human problem solving required between data center trades for AI data center planning 
• Eliminating waste and rework from data center construction

In this episode, Mike Connaughton sits down with Vladimir Kozlov, the founder and CEO of LightCounting Market Research. They discuss some of the developing optical trends attributed to the growth of AI clusters, and challenges predicting trends in a fast-moving industry. They touch on:

• Transceivers and the current uncertainty surrounding trade and the business climate
• The power efficiency requirements driving new approaches to optics and connectivity 
• Examples of AI applications that generate revenue now and in the future 
• Forecasting in a constantly changing world

This week, Mike Connaughton talks to Robert Henke of WestCal Technologies about the exiting and dynamic state of AI networks. They discuss:

• The early “wild west” era of AI infrastructure
• The diverse group of companies interested in AI networks
• The frequency of upgrades to AI chips and GPUs
• Structured cabling and lasting solutions for quickly evolving tech

David Potts, the Director of National Data Center Business a Graybar, brings a unique perspective when supporting data center customers with their AI infrastructure. He talks to Leviton’s Mike Connaughton about: 

• Transitioning data center space to AI and finding power grid capacity
• The role distribution partners play in supporting AI data center deployments
• The demand for inventory and shorter lead times 
• The leaps that markets and industries will make with the help of AI

Host Roy Chamberlain talks to network specification engineer Peter Helfrich about the design and implementation of structured cabling for AI data centers. They discuss:

• The difference between AI data centers and traditional data centers
• Challenges repurposing existing data center space for AI networks
• The big players in AI for GPUs, servers, switches, and applications
• How Leviton specification and product experts help clients with AI network design, deployment and education

In the wrap-up to our 4-episode AI panel discussion, Leviton’s Mike Connaughton hosts a Q&A session with panelists Jonathan Jew, Brooke Ford, and Nick Taylor. Topics include: 

• Power infrastructure for AI data centers
• Moving to very small form factor connectivity
• Structured cabling vs point-to-point cabling
• Dispelling myths around latency 
• Trends in cooling distribution

Our panel discussion continues as Nick Taylor from Networks Centre talks about the cooling technologies used for AI networks. Nick discusses the challenges for deploying AI clusters and high performance computing in existing data centers designed with air cooling. He also covers the latest cooling technologies, including cold aisle and hot aisle containment, rear door heat exchangers, direct to chip, and immersion cooling.

In a continuation of the AI panel discussion at the 2025 BICSI Winter Conference, Leviton’s Brooke Ford discusses some of the emerging fiber infrastructure solutions that address density, including very small form factor (VSFF) connectors and multicore fiber. He also outlines potential technologies that reduce power consumption, such as co-packaged optics and linear drive pluggable optics. Lastly, Brooke explains how hollow core fiber is a unique (but expensive) option for improving latency in AI networks.

Leviton hosted an AI panel discussion during the 2025 BICSI Winter Conference, which included telecommunications industry luminary Jonathan Jew, president of J&M Consultants. In this episode, Jonathan provides an overview of the major AI chip manufacturers and their architectures for GPU networks. He also covers the GPU, storage, in-band, and out-of-band network connections in AI clusters, and explains why structured cabling is key to handling the sheer volume of fiber connecting AI networks.

In this first episode of our series on artificial intelligence, host Roy Chamberlian talks to data center network expert Mike Connaughton about the larger trends in the AI space and what they hope to learn from this season. They discuss:

• How AI growth is requiring solution providers and standards bodies to keep up with rapid changes
• The range of disciplines collaborating to address AI network challenges and best practices
• What does an AI data center look like? Who is building them?