Cisco Accelerates Shift From Copper to Optical Interconnects in AI Data Centers
As GPU-to-GPU connections inside AI racks push copper to its limits, Cisco is driving the industry toward optical networking through new silicon designs and denser pluggable modules developed alongside Nvidia.
The infrastructure supporting artificial intelligence systems is undergoing a fundamental transformation as electrical copper connections between processors inside server racks reach their performance ceiling at higher signaling speeds. Transitioning to light-based connections stands to reshape how the industry defines the physical boundaries of a single AI deployment.
Cisco Systems Inc. is advancing this shift through multiple initiatives, including new processor designs and optical transceiver modules, as well as a standard created jointly with Nvidia Corp. The industry now finds itself at a critical juncture between traditional electrical and emerging optical technologies, according to Rakesh Chopra, senior vice president of silicon and systems architecture and Cisco Fellow at Cisco.
Right now, in-rack copper, in-system copper … is not a durable place to stand. It will switch to optical. My guess is what you'll see is in a handful of years, that'll be the norm, and that allows you to take your scale-up domain from a single rack to row-scale or beyond.
Rakesh Chopra, Cisco
Why AI infrastructure networking needs a new playbook
The immediate priority involves deploying more compact optical transceiver modules. Cisco and Nvidia revealed their strategy around USFP, which stands for Ultra-Dense Small Form Factor Pluggable, a module capable of handling 400 gigabits per lane. Current optical modules max out at 200 gigabits per lane, Chopra noted.
Our goal with that is to drive an open ecosystem enabling people to deploy high bandwidth, highly efficient optics into their infrastructure. So it's about four times denser than today's OSFP 1.6T and about 1.3 times more power-efficient.
Rakesh Chopra, Cisco
This optical advancement builds on Cisco Silicon One, the unified chip platform Cisco introduced in December 2019. Rather than creating separate silicon designs for different market segments—an approach that creates isolated technical complexity—Cisco engineered a single flexible architecture that now powers devices ranging from 100 gigabits to 100 terabits, Chopra explained.
https://www.youtube.com/embed/RSxrKnX-a4k?feature=oembed
This notion that locality defines the function is actually a fundamentally flawed premise. It's the premise of yesterday and not the premise of tomorrow. What really matters is that you can secure and deploy it in different sizes anywhere you need to.
Rakesh Chopra, Cisco
Read more
Speed to operational status now matters more than GPU counts in AI infrastructure buildouts
As organizations race to deploy AI factories, the bottleneck has shifted from securing graphics processors to managing data pipelines and storage. Infrastructure vendors are rethinking how they approach these projects from the ground up.
SpaceXSI's Grok Voice 3.0 Tops Industry Benchmarks in Speech-to-Speech Showdown
Elon Musk's SpaceXSI has unveiled Grok Voice 3.0, a speech-to-speech model that outperforms competitors from OpenAI and Google across multiple independent testing frameworks.
Microsoft Pursues Composable Memory to Ease AI Data Center Bottlenecks
As memory scarcity constrains AI workloads across data centers, Microsoft is exploring composable memory architectures that dynamically allocate capacity across servers rather than locking it within individual machines.