104. AI Networking Will Never Make It To The Enterprise

21.4.2026

Tech Field Day Podcast

0:00

26:42

Learn more about ⁠Networking Field Day 40 here⁠.

AI networking is driving innovation in the market like no other time in history. But the likelihood of all these advances filtering down into the enterprise is very slim. In this episode of the Tech Field Day Podcast, host Tom Hollingsworth and a panel of IT experts debate whether high-performance AI networking will ever truly penetrate the traditional enterprise. The discussion centers on the premise that the specialized, low-latency fabrics and ultra-high-speed Ethernet modules (like 800GbE) may be overkill for the average organization. While Ethernet continues to evolve to meet these demands, they question if the typical enterprise workload, primarily driven by human users rather than GPU clusters, actually requires the “trickery” of distributed flows and SmartNICs.

The conversation shifts toward the practicalities of adoption, highlighting that while the technology exists, its integration into the enterprise is often limited by physical constraints like power, cooling, and the lack of a killer app. However, the guests suggest that AI networking might arrive in smaller, rack-optimized pods or self-contained black boxes as companies begin to train private models on their own sovereign data. Ultimately, the experts advise that while engineering teams must become literate in high-performance computing, organizations should be wary of over-provisioning for a future they may not yet need, lest their expensive high-speed infrastructure ends up collecting dust.

Panelists:

Jeff Ello, Head of AI at Purdue University

Mike Witte, World Wide Technologies

Vincent Celindro, Director of Strategic Sales & Technology at Quantum Foundry

Hosts:

⁠⁠⁠⁠⁠⁠Tom Hollingsworth⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, Event Lead for Tech Field Day

⁠⁠⁠⁠⁠⁠⁠Alastair Cooke⁠⁠⁠⁠⁠⁠⁠⁠, Event Lead at Tech Field Day

⁠⁠⁠⁠⁠⁠⁠Stephen Foskett⁠⁠⁠⁠⁠⁠⁠⁠, President and Organizer of ⁠⁠⁠⁠⁠⁠⁠⁠Tech Field Day⁠⁠⁠⁠⁠⁠⁠⁠