Cisco shifts AI racks from copper to optics as speeds strain links
Cisco is phasing out copper interconnects inside AI racks, replacing them with optical links to handle the signaling speeds required by next-generation GPU clusters. The move, reported by SiliconANGLE, responds to a physical bottleneck: copper’s electrical resistance degrades signal integrity at the terabit-per-second rates AI workloads now demand. Optical fibers, which transmit data as light pulses, avoid this constraint, potentially redefining what counts as a single AI system.
The shift suggests a broader industry trend. Earlier this year, CScale raised $145 million in a Series C round led by Nvidia and Intel Capital, reflecting investor confidence in optical solutions for AI networking. While Cisco’s efforts may accelerate adoption, the transition will likely unfold over time. Copper remains cheaper and simpler to deploy, and most existing AI racks still rely on it. But as GPU clusters grow denser and models demand lower latency, the case for optics will strengthen.
For now, the question is how quickly the industry will pivot—and whether incumbents or startups will lead the charge.
Sources: siliconangle.com
“Cisco’s push into optical interconnects signals a structural limit in AI scaling—and a market opening for startups like CScale that have already drawn Nvidia and Intel’s capital.”
Read the original reporting
The outlets below did the original reporting.
- Copper’s days are numbered inside AI racks as Cisco bets on optics — siliconangle.com
Related briefs
- September unicorn surge led by semiconductors, space tech
- ScaleFlux targets AI memory bottleneck with smarter SSD controllers
- Vast Data targets AI data sync gap for sensitive enterprise workloads
- Beijing AI chip startup SUNMMIO closes Pre-A2 and Series A
- Intel pitches "right-sized" AI systems at OCP Summit
This brief was drafted automatically from the sources above and published under our editorial policy. Spotted an error? Tell us.