The old new thing: Cisco and retro tech are back and forming the AI buildout’s backbone

As the industry accelerates construction of infrastructure to support artificial intelligence, some previously sidelined technologies are reclaiming a central role. Equipment and systems that many considered obsolete have been re-evaluated for their reliability and compatibility with the massive networking demands of AI workloads. Companies such as Cisco, long known for enterprise networking gear, are being cited as part of that revival.

Engineers and data-center builders are turning to well-understood networking hardware and established platforms to stitch together the complex environments that AI training and inference require. The preference for proven components is driven by the need for predictable performance and interoperability as organizations assemble heterogeneous stacks of accelerators, storage, and software.

Observers note that this is less a return to the past than a pragmatic reuse of mature technologies in a new context. The quote, "Everybody forgot about them for 20 years. And now suddenly they’re important again," captures how quickly market priorities can shift when new compute paradigms—like large-scale AI—emerge and put a premium on stable infrastructure.

The shift highlights how infrastructure ecosystems evolve: tools and vendors that fell out of favor can regain prominence when their strengths match fresh technical challenges. For suppliers of legacy networking gear and for operators sourcing hardware, the trend suggests renewed attention to compatibility, longevity, and the role of tried-and-tested platforms in modern AI deployments.