Within the AI period, Gillis mentioned, the community performs the position the PCI bus as soon as performed inside a single server. Distributed AI techniques require reminiscence, compute, GPU, and storage to work collectively throughout bodily infrastructure at scale. The community is the backplane that makes that attainable. The end result, he mentioned, is that clients have come to see it because the one factor they’ll depend on.
“There’s a new working mannequin crucial for infrastructure,” Gillis informed Community World.
AI modified how Cisco builds its personal merchandise
The shift exhibits up inside Cisco’s personal growth group. Gillis runs a staff of roughly 12,000 software program builders, and AI coding instruments have essentially modified how that staff works.
Earlier generations of AI coding instruments produced important positive aspects on new initiatives, the place a staff of 5 to 10 builders might accomplish what as soon as required 100 individuals working for a yr. However these instruments hit a ceiling on advanced legacy merchandise. A Catalyst swap or Cisco firewall can include 50 to 100 million strains of code, extra context than prior fashions might deal with without delay. Newer AI coding instruments have eliminated that ceiling, permitting Cisco to speed up growth throughout its total product portfolio.