Cisco Live 2025: Unlocking the Future of Compute & AI
In this blog
Key takeaways from Cisco Live 2025
This year's Cisco Live 2025 event made it clear that Cisco is doubling down on its AI strategy, compute infrastructure and secure networking. With a focus on enabling enterprises to embrace scalable AI-ready infrastructure, Cisco's messaging and product updates signal a transformative shift in how organizations should approach the future of digital operations.
From June 8–12, thousands of IT professionals, partners and enterprise leaders convened in San Diego to explore the future of AI and infrastructure. Major sponsors such as NVIDIA, AMD, Intel, Red Hat and Splunk helped reinforce Cisco's ecosystem-driven approach. Keynote sessions and product launches emphasized the ongoing shift toward "agentic AI" architectures and the importance of building converged platforms that unify compute, security and observability.
Compute announcements: UCS C845A M8 and C885A M8 servers
Among the most impactful hardware announcements were the UCS C845A M8 (NVIDIA MGX) and C885A M8 (NVIDIA HGX) rack servers. These systems are built on AMD EPYC processors and feature PCIe Gen5 and DDR5 memory, underscoring Cisco's focus on delivering scalable, high-performance infrastructure for AI and data-intensive workloads.
The UCS C845A M8 is a 4RU air-cooled rack server, designed with flexibility in mind. It supports configurations of 2, 4, 6 or 8 GPUs, allowing customers to choose from NVIDIA H100 NVL, H200 NVL, L40S, or AMD MI210. This scalability makes it ideal for a wide range of use cases — from inference and analytics to full-scale model training. Its modular approach lets enterprises right-size their environments, with future-ready support for upcoming GPU generations.
Start small and scale up with the C845A M8.
The UCS C885A M8, by contrast, is a larger 8RU rack server engineered for maximum GPU density. It supports up to 8 GPUs in a single chassis and is optimized for compute-heavy AI workloads such as generative AI (GenAI) training, large-scale inference, and multi-model orchestration. Despite its physical footprint, the 8RU design allows for dense GPU deployment without compromising on power delivery or cooling.
These compute platforms are designed to integrate seamlessly with the new Cisco UCS 6600 Series Fabric Interconnects, providing the high-throughput and low-latency backbone required to fully unlock the performance of GPU-intensive workloads.
Both servers are designed for seamless integration into AI-ready data center architectures and can be validated through WWT's Advanced Technology Center (ATC) to help organizations benchmark performance and optimize deployments before scaling into production. The C885A, in particular, will be making its way to the AI Proving Ground later this summer, offering customers a chance to evaluate its performance in AI-specific environments.
Architecting the AI-ready data center
Cisco's vision for the AI-ready data center was one of the most forward-thinking elements of the event. Central to this vision is the concept of AI Infrastructure PODs — modular, scalable building blocks designed to support AI workloads ranging from inferencing to large-scale model training. These Cisco UCS AI PODs combine compute, storage and high-speed networking into pre-validated, reference architectures that can be deployed quickly and expanded easily.
The AI PODs are built around Cisco UCS X-Series servers and C-Series GPU-dense systems, such as the C845A M8 and C885A M8, integrated with NVIDIA GPUs, NVMe storage, and Cisco Nexus networking. Each POD is optimized for specific AI use cases and performance tiers, enabling customers to start with a single POD and scale horizontally as demand grows. Importantly, the architecture is workload-aware, supporting both inferencing and training requirements with tailored power, cooling and connectivity profiles.
Cisco also incorporates observability and automation across these AI PODs via Intersight and Nexus Dashboard, giving customers full visibility and lifecycle management from a single pane of glass. These tools simplify ongoing operations, optimize performance and reduce time-to-value.
For organizations exploring GenAI or domain-specific models, the modularity of AI PODs presents a practical path forward. This approach is further strengthened by Cisco's AI Factory with NVIDIA— an ecosystem initiative designed to accelerate the adoption of enterprise AI by bringing together infrastructure, software and expertise into a unified framework. At WWT, we are enabling customers to test and validate these architectures in our ATC, simulating real-world workloads to fine-tune AI performance before moving into production.
Securing the AI era
Security was also a central theme, particularly as Cisco unveiled enhancements in Zero Trust Network Access (ZTNA), Hybrid Mesh Firewalls, and quantum-resistant cryptography. These technologies were introduced as foundational pillars of the "secure infrastructure for the AI era," blending traditional threat defense with new requirements for securing autonomous workloads and distributed compute fabrics.
The concept of AgenticOps — Cisco's AI-powered operations layer — represents a clear push toward self-operating IT environments. This includes Deep Network Models, AI-powered observability with ThousandEyes and Splunk integration, and unified security policy enforcement. These features are especially relevant for enterprise customers looking to balance agility with compliance and governance.
AI Canvas + Deep Network: Driving smarter compute operations
At the heart of Cisco's AI-driven network transformation lies the Deep Network Model — a domain-specific large language model (LLM) purpose-built for networking. Unlike general-purpose LLMs, this model is trained using decades of Cisco expertise, leveraging more than 40 years of CCIE-level certification material, TAC case histories and real-world network configurations. This deep specialization enables up to 20 percent greater accuracy in troubleshooting, configuration recommendations, and automation tasks compared to more generic AI models.
This precision intelligence is foundational to Cisco's emerging AgenticOps paradigm. Rather than merely surfacing insights, AgenticOps enables the network to take intelligent action: hypothesizing root causes and proposing remediations or configuration changes using a natural-language, telemetry-aware workflow. These recommendations are grounded in real-time operational data from Cisco ThousandEyes, Splunk and Cisco Intersight.
The Deep Network Model is delivered through Cisco AI Canvas, a unified, multi-user interface where engineers engage with AI via conversational prompts. AI Canvas dynamically builds context-aware dashboards and enables secure action from within the same interface, maintaining human control while improving response time and operational accuracy.
Together, the Deep Network Model and AI Canvas don't just alert operators to anomalies—they diagnose root causes, recommend next steps, and initiate remediation. In early use cases, this collaboration has helped reduce incident resolution times from days to minutes, setting the stage for more autonomous, intelligent compute environments.
Ecosystem synergies
What made this year's announcements particularly effective was Cisco's focus on partnerships. From GPU integration with NVIDIA to support for Red Hat OpenShift, VMware Tanzu and Nutanix, Cisco showcased a deep awareness that AI transformation requires an ecosystem, not a siloed stack. Improved public cloud connectors and multicloud telemetry further solidify Cisco's positioning as a hybrid-first infrastructure provider. For WWT, these integrations are critical. They align with our customer engagements and allow us to deliver full-stack architectures that bridge on-premises and cloud-native technologies.
Real-world impact
Cisco highlighted several in-progress deployments in verticals like healthcare and automotive. In these industry use cases, customers were able to accelerate AI onboarding, reduce infrastructure provisioning time, and gain real-time insights through AI observability stacks.
These kinds of outcomes — shortened AI lifecycle deployment, tighter security perimeters and seamless scalability — are what enterprises are demanding. Through our ATC, WWT is uniquely positioned to help customers replicate these success models.
Leadership insights
Cisco Live 2025 made it clear: Infrastructure is no longer just about speeds and feeds. It's about operational intelligence, scalability and embedded security. For executive stakeholders, this means rethinking investment strategies around infrastructure to ensure readiness for AI-enabled operations. TCO and ROI are no longer measured just by capacity and utilization, but by how quickly and securely a business can act on data.
The era of agentic AI, as Cisco terms it, will be defined by architectures that are dynamic, observable and secure by design. And for organizations aiming to lead in this next phase of innovation, partnering with a strategic integrator like WWT — one of Cisco's largest and most trusted global partners — offers unmatched value. WWT provides far more than just access to products; we bring deep technical expertise, end-to-end architectural support, and the ability to test and validate solutions through our ATC. This ensures that our customers gain not only the technology but the guidance and insight needed for successful deployment and long-term impact.