The Everything Feed - All Packet Pushers Pods NB549: Startups Take on Switch, ASIC Incumbents; Google Claims Quantum Advantage
Oct 27, 2025
In this discussion, Kevin Egan, Senior Director at Equinix, dives into the company's vital role in AI infrastructure. He highlights emerging trends in edge workloads and the significance of colocation as a third option for businesses. Kevin explains how Equinix facilitates low-latency, private connectivity for AI applications. He also touches on the increasing network demands from inferencing workloads, advocating for observability and integration. Discover how these innovations shape the future of AI networking!
AI Snips
Chapters
Transcript
Episode notes
ASIC For Scale-Up GPU Pods
- Upscale AI targets scale-up GPU pods with a custom ASIC (Skyhammer) for deterministic latency and high bandwidth.
- The ASIC targets UA-Link and ESUN rather than NVIDIA's NVLink, narrowing the initial market to non-NVIDIA GPU stacks.
Private 5G With Local Breakout
- T-Mobile's edge control lets private on-prem 5G keep traffic local via local breakout for low-latency and data-locality use cases.
- This appeals to customers needing strict data locality, ultra-low latency, and edge AI/IoT control.
Validate Network AI Agents Before Trusting
- Vet AI agents in networking carefully and build trust before automating critical tasks.
- Verify outputs because agentic LLMs are non-deterministic and can produce incorrect actions.