The Deep Edge Podcast

The Deep Edge Podcast covers the infrastructure behind telecom and cloud services, and it keeps asking what that infrastructure costs. The host is Ray Mota, who introduces himself as being from ACG Research. His guests come from the companies building the equipment and software. That puts vendor strategy on the record firsthand, from people describing their own companies' products. Will Eatherton, an engineering leader at Cisco, is a good example. His conversation with Mota covers Cisco's AI strategy, its NVIDIA partnerships, Silicon One, SONiC and Hyperfabric. He also describes a 50-terabit chip with deep buffers, pitched at backbone and data center interconnect traffic. Other episodes move outside the data center. Julius Francis and Jai Thattil of Juniper Networks discuss the cloud metro. They contrast a traditional model built on devices, manual operation and individual know-how with one where operations shift toward service experience and cloud-delivered management. Ritesh Mukherjee, SVP and General Manager for Enterprise Networks at Inseego, returns for a question Mota puts plainly: is the industry ready to run 5G as the main line in SD-WAN designs rather than as a backup? One episode, with Livio Silva of Red Hat, is in Spanish. It lays out access and aggregation edge sites, MEC applications such as facial-recognition payments, and Red Hat OpenShift as a common platform from edge to core. The economics are what make the show distinctive. Ivan Melia of Cisco, another returning guest, walks through an economic study of MSX that weighs deployment size, domain complexity and churn, along with service lifecycle and day-two management. In the Juniper conversation, Mota offers to link papers with the TCO findings for anyone who wants more detail. Mota makes the sharpest version of the argument himself. In an episode on the economics of AI infrastructure, he points out that the GPU cluster, not the network, often takes the largest share of the investment. By his account, a fabric that leaves those GPUs idle causes an economic loss, not just a performance problem. He says operators are shifting from judging cost after a decision to modeling it beforehand, and he names economic digital twins as the tool for that. The tone is friendly and upbeat. Mota calls returning guests 'my friend' and praises one answer as 'spoken like a true technologist.' The vocabulary is dense: DCI, MEC, OSS and BSS, SD-WAN, OpenRAN. A good way in is to pair the Eatherton conversation with Mota's economics episode, which together cover the AI data center from both the engineering side and the financial side. Anyone closer to operator networks may get more from the Juniper metro discussion or the Spanish-language edge episode.