AI training isn’t just about GPUs, it’s about the network that ties them together. Host Phil Gervasi sits down with Vijay Vusirikala to unpack why job completion time is the true metric of success, how optical interconnects shape AI datacenter performance, and why power efficiency and observability are becoming mission-critical.

In this episode of Telemetry Now, Phil and guest David Cliffe explore “neocloud”—what it is, why it’s growing fast, and how it’s reshaping AI infrastructure. From GPUs and high-speed interconnects to orchestration layers and energy constraints, we discuss what makes neocloud different from traditional cloud, and what it means for data center operators, service providers, and AI architects.

In this Telemetry News Now episode, Phil Gervasi and Justin Ryburn tackle Intel’s plan to spin off its Network & Edge Group (NEX) and what it means for high‑performance Ethernet NICs, examine Cisco’s new partnership with Hugging Face to scan every open‑source AI model for malware, and break down Broadcom’s Jericho4 fabric router—bringing 3.2 Tb/s “hyper‑ports” to distributed AI clusters. The hosts also discuss the real‑world state of SD‑WAN adoption, Palo Alto Networks’ $25 B CyberArk acquisition, Cisco’s quantum‑networking research, and upcoming industry events. Plus: a personal tale of basement‑automation triumph to kick things off.