In partnership with

Check out our latest Next TMT Talks podcast with media entrepreneur Robert Rose, founder of AIM TV Group, who delves into such topics as broadcast-TV self-sabotage, the ‘untapped’ portion of the Latino TV market, the broadening horizons of world travel … and the need for FAST monetization to hurry up and get here already.

Nvidia works with Comcast and Charter to distribute AI compute closer to end users

Dispersing centralized data-center functions to the edge of cable’s HFC networks could solve a rang of problems, from power consumption to latency

First deployed by cable operators in the early 1990s to distribute television, hybrid fiber-coax (HFC) networks transitioned nicely over the ensuing decades as these same companies shifted network use to primarily distribution of high-speed internet service.

And now that fiber and wireless operators seemed to have caught up to these cable operators in home broadband, the distributed architecture of HFC networks is about to give Big Cable the edge once again in the AI era … on the edge of their networks, that is.

Convening its big annual “GPU Tech Conference” (GTC) in San Jose this week, Nvidia had already been moving aggressively into telecom. Anyone attending Mobile World Congress 2026 in Barcelona two weeks ago could see that the world’s most valued company (market cap: $4.38 trillion) wants to position its powerful, machine-learning-capable graphics processing units (GPUs) into places like 5G cell towers and satellite Wi-Fi access nodes.

Dispersing the burden of the GPU from one central power-hungry data-center location to myriad edge positions presents a range of advantages, cutting latency by milliseconds and dispersing electrical needs.

And at GTC 2026, it was made clear that Comcast and Charter Communications’ distributed HFC networks are well suited for an era of “edge AI.”