This content is part of the Conference Coverage: Cisco Live 2024 conference coverage and analysis

Latest Cisco products show measured approach to AI

Cisco has launched a SaaS product for applying policy controls to AI model-bound data and an Nvidia partnership to bolster Cisco UCS servers for AI at the edge and data centers.

Cisco has added infrastructure and software to its growing product portfolio for cloud providers and enterprises utilizing AI applications and models.

Cisco launched a SaaS product called Motific that applies policy controls for using corporate data in generative AI large language models. Other steps Cisco took in the AI market include partnering with chipmaker Nvidia to improve Cisco hardware for AI at the edge and in data centers, and validating designs to build data center network fabrics underneath AI models and applications.

The networking giant launched Motific at its Cisco Live Amsterdam user and partner conference this week. The service, developed within Cisco's tech incubator Outshift, fits the company's focus on protecting and managing private data that companies use to train AI models in the cloud or private data centers.

In an interview with Yahoo Finance last month, Cisco CEO Chuck Robbins said  successful AI companies will have product features  that help users take advantage of their data securely.

Cisco Motific

Motific targets organizations using AI models in the cloud or private data centers. The online software provides a central dashboard with multiple services:

  • Automatic configuration of APIs to access models, AI assistants and retrieval-augmented generation services for fetching private data to fine-tune model outputs.
  • Controls for applying policies that determine model access, control misinformation from models and restrict the use of sensitive data, such as personally identifiable information.
  • The ability to track model usage and provide cost analyses to help maintain a return on investment. Motific can establish cost budgets and estimates to help control spending.

Cisco plans to make Motific generally available in June.

Nvidia comes to Cisco M7

Cisco's latest partnership with Nvidia makes the latter's Tensor Core GPUs available in Cisco M7 rack and blade servers, including the Unified Computing System (UCS) X-Series and X-Series Direct. The GPUs make the data center servers useful for AI applications at the edge and the data center.

Also, Cisco made the Nvidia AI Enterprise portfolio, which includes software frameworks, pre-trained models and development tools for AI production, available on its global price list.

The deal opens a significant enterprise sales channel for Nvidia. Cisco plans to offer Nvidia products through its channel partners in the second half of the year.

They're not rushing with AI; they're doing things one step at a time.
Vijay BhagavathAnalyst, IDC

Cisco's partnership with Nvidia is part of its push into the GenAI market. According to Robbins, some of the largest cloud providers are testing Cisco's Silicon One processor for AI workloads.

Cisco's products cover every networking layer, so its approach to AI must be measured but steady as it navigates current market hype, IDC analyst Vijay Bhagavath said. "They're not rushing with AI; they're doing things one step at a time."

Ethernet for AI

Cisco is working with rivals Arista and Juniper Networks to make the Ethernet standard capable of replacing InfiniBand as the network protocol for running AI models in enterprise data centers. The three companies, which provide hardware and software for only Ethernet networks, are members of the Ultra Ethernet Consortium, which is developing the standard for the most data-intensive workloads.

"The greatest weaknesses of Ethernet has long been the fact that its general-purpose nature meant that you had to go through a lot of hoops to have it perform properly for specific types of workloads," said J Metz, the UEC steering committee chair, in a recent interview. Metz added that he expects the UEC to release its first full specification by the end of 2025.

If the UEC is successful, Bhagavath said, Ethernet would likely replace InfiniBand first in hyperscale data centers and later in large enterprise facilities. "It's a graduated, stepwise approach" he said.

In other networking news, at Cisco Live, the company made more Catalyst hardware manageable from the online Cisco Meraki dashboard. The added hardware includes the Catalyst 9800 Wireless Controllers; Wi-Fi 5 Wave 2; Wi-Fi 6 and 6E access points; and the 9300-M switches for the branch and campus.

Other releases include an AI-powered assistant in Secure Access that uses natural language to create network access policies. Secure Access is part of Cisco's cloud-based security service edge, which protects against network threats from any location where employees work.

In December, Cisco introduced an AI assistant for firewall policies within its cloud-delivered Firewall Management Center and Defense Orchestrator, a platform for managing policies in Cisco firewalls and intrusion prevention systems.

Antone Gonsalves is an editor at large for TechTarget Editorial, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco.

Next Steps

Cisco commits to GenAI with HyperFabric, $1B fund

Dig Deeper on Cloud and data center networking

Unified Communications
Mobile Computing
Data Center