The News
Cisco released new research titled The Race to Agentic AI, finding that 96% of executives believe agentic AI requires robust networks and 78% expect it to reshape industries within 3–5 years. The report positions agentic AI as fundamentally an infrastructure challenge, with organizations allocating an average of 37% of technology budgets toward enabling these systems.
In parallel, Cisco amplified the conversation at its second annual AI Summit, featuring leaders from NVIDIA, OpenAI, Google, and Microsoft. Additional announcements at Cisco Live EMEA 2026 included the Silicon One G300, AgenticOps innovations, and updates to Cisco AI Defense. Cisco also reported strong AI and networking momentum during its Q2 FY26 earnings call and published its State of AI Security 2026 report highlighting emerging risks tied to autonomous AI systems.
Analysis
Agentic AI Becomes a Network and Data Fabric Problem
Agentic AI is shifting from experimentation to enterprise-scale deployment. Cisco’s research indicates 80% of executives believe agentic AI will be essential for survival by 2027, while 55% expect more than half the workforce to collaborate directly with AI agents within 24 months.
From an application development standpoint, this reinforces a structural reality: AI agents are not isolated LLM endpoints. They operate across APIs, SaaS platforms, hybrid clouds, identity systems, and data pipelines. Our Day 2 research shows 75.8% of organizations already operate across SaaS environments and 69.6% leverage public cloud IaaS/PaaS, with 25.8% using three cloud providers simultaneously. That multi-environment sprawl magnifies the infrastructure requirements behind agent orchestration.
Agentic AI therefore becomes less about model performance and more about network resilience, data locality, latency, and identity federation across distributed systems.
Infrastructure Spend Shifts Toward AI-Ready Architectures
Allocating 37% of technology budgets toward agentic AI signals a reprioritization of infrastructure modernization. Our Day 1 data shows 74.3% of organizations rank AI/ML as their top spending priority, while 60.7% prioritize cloud infrastructure and 68.3% security and compliance.
Cisco’s messaging aligns with this convergence. Silicon One G300 and AgenticOps innovations suggest that networking silicon, observability, and AI-driven operations are becoming foundational to scaling AI workloads. Meanwhile, the State of AI Security 2026 report highlights risks emerging from unmonitored “connective tissue” between AI systems, particularly as autonomous agents communicate with one another.
For developers, this means application architectures must account for east-west traffic visibility, zero-trust identity enforcement, encrypted data flows, and real-time policy evaluation. Agentic AI expands the attack surface, especially in hybrid and sovereign environments.
Scaling Agents in Hybrid Enterprises
Day 2 research shows 46.5% of organizations require deployment speeds that are 50–100% faster than three years ago. At the same time, 45.7% report spending too much time identifying root cause during incidents, and 23.7% cite data growth as a major observability challenge.
Agentic AI compounds these pressures. Unlike traditional applications, agents continuously interact with internal systems, external APIs, and sometimes other agents. This dynamic behavior introduces variability in network traffic patterns and execution flows that legacy infrastructure was not designed to handle.
Agentic AI introduces unpredictable east-west interactions, service mesh complexity, and identity propagation across microservices and cloud boundaries. Developers previously addressed scaling challenges by increasing compute capacity or optimizing code paths. With agentic systems, scaling often hinges on network determinism, policy enforcement, and distributed observability.
AgenticOps and AI Security as Development Primitives
Cisco’s framing of AgenticOps suggests a future where AI-driven operational automation becomes part of the core infrastructure stack. Given that 71.0% of organizations already leverage AIOps and 66.7% say it accelerates scaling, agent-level orchestration may represent the next iteration of operational automation.
However, the State of AI Security 2026 findings underscore a governance gap. Rapid deployment of agentic AI without securing identity layers, API gateways, and data fabrics creates exploitable seams. Our DevSecOps data shows APIs are considered the most susceptible cloud-native stack element at 36.2%, followed by identity and access management at 24.7%. Agentic AI amplifies both risk vectors.
Developers may need to treat secure network design, AI observability, and identity orchestration as first-class development concerns rather than downstream operational tasks. Infrastructure design decisions could increasingly influence how safely and reliably agents can operate at scale.
Looking Ahead
The industry narrative is shifting from “How do we integrate LLMs?” to “How do we operationalize autonomous systems at enterprise scale?” Cisco’s research and event announcements suggest that competitive differentiation may hinge on infrastructure maturity rather than model selection alone.
If agentic AI adoption accelerates as projected, we expect heightened investment in AI-ready networking silicon, hybrid connectivity, sovereign data controls, and agent-level observability frameworks. The organizations that align infrastructure, security, and development workflows around these realities may be better positioned to scale autonomous systems responsibly.
The race to agentic AI is not solely about intelligence. It is about whether the underlying infrastructure can sustain it.
Potpie Targets Enterprise AI Agent Context Gap
Red Hat Positions AI Native Telco Platform for MWC 2026
AMD and Meta Scale AI to 6 Gigawatts of GPU Power
Cisco Signals Agentic AI Infrastructure Reckoning
Axonis Targets the AI Trust Wall With Federated Decision Intelligence
