Cisco Live 2025 is being held this week in San Diego. I had a chance to speak with Jeetu Patel, Cisco’s president and chief product officer, before the event, and he dubbed it “the most consequential Cisco Live in the past decade,” as he emphasized the combination of interest in AI and Cisco innovations to help customers adopt it.
At last year’s Live event, Cisco indicated it was all in on AI, and it has continued to drive developments in that direction. Every IT pro and business leader has AI on their minds, but organizations struggle with adoption since there is no plug-and-play way of deploying the technology.
Most of the announcements at this year’s Cisco Live are focused on two aspects: AI being used to make Cisco technology smarter and easier to use, and Cisco innovations that will enable customers to adopt AI. An easy way to consider the two buckets is “AI for Cisco” and “Cisco for AI.”
Below are the key announcements at Cisco Live 2025.
Focus on AgenticOps
Cisco is debuting its strategy for the next phase beyond AIOps, that being AgenticOps. There is a massive change about to occur in AI. Today we have a workforce that’s all human, but soon there will be a mix of AI agents helping us with tasks and processes. Some industries will see a rise of physical AI that will bring robots, co-bots and humanoids to data centers, workplaces and industrial locations.
This AI evolution presents an entirely new class of risks and requirements for infrastructure. Another challenge is this shift is happening faster than customers can adapt. To help customers, Cisco announced the Cisco Deep Network Model. During his keynote (pictured above), Patel described it as “the most advanced networking LLM on the market.” Cisco is positioning the advancement not as a replacement for the engineer but as someone that can help networking professionals work faster. Patel urged the audience to embrace the concept of AgenticOps when he stated, “AI won’t take your job but the engineer that’s willing to use AI will.”
Key features of the Cisco Deep Network Model include:
- Purpose-built for networking: 20% more precise reasoning for troubleshooting, configuration, and automation.
- Trusted training: Fine-tuned on 40+ years of expertise and expert-vetted for accuracy.
- Continuous learning: Evolves with live telemetry and real-world Cisco TAC and CX services insights.
Another new Agentic AI capability from Cisco is AI Canvas, which supports the creation of generative dashboards using GenAI and, along with Cisco’s embedded AI assistant, enables users to solve cross-domain problems in ways that could never be done before.
AI Canvas has three important use cases: It provides the ability to troubleshoot and execute across multiple domains. It also enables collaboration across multiple users, NetOps teams, SecOps teams, and executives. And it provides a single dashboard with an embedded AI assistant. AI Canvas is built on the foundation of the Cisco Deep Network Model.
Secure global connectivity
Since early in the company’s history, Cisco has been a trusted partner to service providers. Cisco helped network operators build the Internet, transition to VoIP, and become part of the cloud ecosystem. Now, AI can be a massive tailwind for this audience. Earlier this year, at Cisco Live Amsterdam, Cisco introduced agile services as a new approach to building AI-ready transport networks.
Among the connectivity announcements in San Diego this week are two new 8000 series routers-built on Silicon One, Cisco’s own high-performance silicon optimized for network processing. The Cisco 8011, designed for converged access, enhances performance and efficiency at the access layer. The Cisco 8711, for edge routing, provides dense IPSec and MACsec services at the edge and will be available in November of this year. Many competitors poke at Cisco for building its own silicon, but Silicon One is purpose-built for the rigors of networking and performs better than general-purpose silicon designed for the masses.
The company is also introducing a new optical network product, the 400GB BiDi optic, that will enable transitions to 400G networks using existing duplex multimode fiber. Cisco is building multi-agenda capabilities into all parts of its platform, including Cisco Crosswork multi-agentic AI networking for service providers. The ultimate vision here is autonomous networking. It’s unlikely service providers will be fully autonomous any time soon, but this group of companies has always had a problem of escalating costs and falling revenues. AI presents an opportunity to bring costs down and sell services that help customers be AI ready, but they need the infrastructure to be ready.
The AI-ready data center
Cisco announced a unified Nexus dashboard that converges the management policy and data planes across the company’s Application Centric Infrastructure (ACI) and Nexus-based EVPN (Ethernet VPN) and Virtual eXtensible LAN (VXLAN) fabrics. It also includes SAN and campus EVPN environments. The much needed unification enables Cisco to streamline operations across all these domains. It also integrates with CI/CD pipelines via DevOps-ready APIs, which is a big deal for customers looking to modernize their operations. This is long overdue as customers have been asking for simplification of management tools for years.
Simplified full-stack AI and unified intelligence
Cisco will begin taking orders for the RTX 6000 Pro GPU in its C845 server. This new GPU will be available in its AI PODs, which will be configurable so customers can tailor configurations for the specific kind of AI work they’re running, enabling them to start small and then scale up when needed.
New secure firewalls
Adding to the product payload, Cisco is introducing two new secure firewall models. The Secure Firewall 200 series is directed at branches that need on-box decryption and threat inspection. It also has integrated SD-WAN capabilities optimized for Cisco SASE and Cisco ZTNA architectures.
The second, the 6100 series firewall, sets a new high-water mark for performance and density for the Cisco firewall product line. It is meant to address the emerging security challenges of AI workloads that most customers are now starting to see as they scale up their AI initiatives. With these updates, the entire firewall line at Cisco has been refreshed in the past couple of years.
In more firewall-related news, Cisco will be the first company in the industry able to push policy into third-party firewalls. The addition of an intent-based mesh policy engine into Security Cloud Control will enable customers to push policies into competing firewalls. Historically, managing firewalls from multiple vendors was a headache as policies and rules quickly were out of sync. Migrating to another vendor isn’t often practical as customers can’t just rip and replace the existing infrastructure. Now, Cisco can meet customers where they are, and they can swap enforcement points without having to rewrite policies.
Industrial Ethernet
Cisco is introducing 19 new industrial Ethernet products, including small form factors that can be placed on robots. Cisco has beefed up the power over Ethernet (PoE) on the switches, and the products can now handle up to 720 watts per switch. PoE is powering things like AI vision for automated quality inspection in manufacturing. The switches include embedded cyber vision security designed specifically to protect critical operational infrastructure and integrate with IT security operation centers.
Every keynote the Nvidia CEO Jensen Huang has done this year has included a discussion about physical AI, where everything that moves will eventually be autonomous and AI enabled. Cisco’s revamped industrial portfolio addresses the connectivity and security requirements to shift physical AI from a vision to reality.
In a related announcement, Cisco announced the new CW9179F access point, an enterprise grade Wi-Fi 7 product designed specifically for high-density environments such as public venues, stadiums, airports and conference centers. These environments have historically been outfitted with APs designed for carpeted offices and often provide suboptimal experience in large venues. The new CW9179F has software configurable so the coverage can be adapted the unique attributes of venues where coverage patterns aren’t uniform.
Digital resilience
Splunk, Cisco’s big data platform, is gaining new capabilities for monitoring AI infrastructure and trace-level integration for AI-enabled applications in the company’s Observability cloud, as well as LLM monitoring within Cisco AppDynamics. The goal is to help organizations get the visibility they need of AI metrics across AI stacks, LLMs, and infrastructure, all correlated with observability and business context.
Other areas where Splunk is accelerating value realization for customers include Splunk Machine Learning Toolkit version 5.6, which will enable customers to use LLM integrations to ask questions about their Splunk data. This will help analysts do their jobs more effectively and faster. Also, Splunk Search Processing Language (SPL) now includes an improved AI model with the addition of personalization, so SPL can better understand the customer’s environment for a higher degree of accuracy.
I believe this is my 25th Cisco Live, and I can’t remember one where the set of product announcements were as broad and deep as this one.
As his portion of the keynote ended, CEO Chuck Robbins said: “This is the most consequential time for us to come together given what’s happening in the industry. It made me reflect to the early says of the Internet, but this is moving faster, is more complicated, and upside is greater.”
The massive set of product announcements is designed to help customers adopt AI faster and securely. The biggest risk to Cisco is that the company’s innovation might outpace its customers’ and channel partners’ ability to adopt the new technology. Trying to figure out where and how to use all the new technology can be overwhelming. This is where Cisco needs to focus on business outcomes, such as help companies reduce costs or use AI to improve customer service. These can be used as insertion points for the new technology to prove the value and then expand from there.