on Posted on Reading Time: 3 minutes
Artificial intelligence is entering a new operational era. As AI systems evolve from discrete, task-based models into agentic systems that reason, coordinate, and act continuously, the infrastructure assumptions that supported earlier AI deployments are rapidly becoming obsolete. While much of the industry conversation has centered on GPUs, models, and cloud capacity, Mplify’s NaaS for Agentic AI Market Brief identifies a more fundamental constraint emerging across deployments: the network.
Agentic AI systems are inherently distributed. They operate persistently across clouds, data centers, and edge environments, generating continuous, latency-sensitive traffic flows that depend on sustained context and coordination. Unlike traditional application workloads, these systems cannot tolerate unpredictable performance, best-effort routing, or static network configurations. As organizations move from AI experimentation into production-scale deployments, network determinism, automation, and operational responsiveness are increasingly becoming the critical factors for success.
This conclusion is already being validated with operator experiences. In a recent Bits, Bytes & Beyond podcast episode with Divesh Gupta, Vice President of Professional Services and Technology, PCCW Global and Pascal Menezes, CTO, Mplify, they discussed how AI workloads are exposing structural limits in existing networks. Unlike previous infrastructure transitions that allowed gradual adaptation, AI is advancing at a pace that networks were never designed to absorb. In many environments, the network has shifted from an AI enabler to a critical bottleneck.
The move from centralized training workloads to distributed inference further amplifies this challenge. AI inference increasingly occurs closer to users, devices, and data sources at the edge. These workloads demand ultra-low latency, predictable performance, and local processing, often under strict data sovereignty requirements. Traditional WAN architectures, designed for bursty, transactional traffic, were never built for this reality.
As a result, the industry is reframing the role of the network. Across enterprises, service providers, and cloud ecosystems, connectivity is becoming an active, programmable control plane that must dynamically adapt to AI workloads. This is driving renewed strategic focus on Network-as-a-Service (NaaS) models that integrate high-capacity connectivity, lifecycle automation, observability, and security into a unified, orchestrated platform.
The Mplify market brief highlights how deterministic networking capabilities, including Carrier Ethernet, wavelength services, and performance-assured interconnection, are becoming foundational to AI-ready infrastructure. Equally critical is end-to-end automation. Agentic AI systems cannot wait weeks for provisioning or rely on manual troubleshooting processes. Standardized APIs, aligned service payloads, and real-time telemetry are now essential for enabling on-demand capacity, rapid fault isolation, and continuous optimization across multi-provider environments.
PCCW’s Gupta also highlighted a critical duality emerging across the industry: AI for NaaS and NaaS for AI. On one side, AI is transforming how networks are operated through AIOps, predictive analytics, and automated diagnostics. On the other, networks must be re-architected to support AI agents operating at the edge, within enterprises, and across sovereign AI environments. This shift places enormous importance on open standards and interoperability. Proprietary integrations do not scale in an AI-driven world. Instead, standardized APIs, consistent data models, and lifecycle automation are essential to enabling global, multi-provider AI connectivity.
These capabilities also raise new questions around trust, governance, and security. As agentic systems gain autonomy, networks must enforce strong identity, authorization, and auditability frameworks. Zero-trust architectures, richer cybersecurity telemetry, and policy-driven controls are now prerequisites for safely operating AI across jurisdictions and sovereign environments. Emerging approaches such as the Model Context Protocol (MCP) point toward a future where AI agents interact with network APIs dynamically, increasing both opportunity and responsibility for robust guardrails.
Taken together, these trends signal a clear inflection point for the industry. Supporting agentic AI requires rethinking the network as a strategic foundation – programmable, deterministic, automated, and deeply integrated with AI operations.
The full NaaS for Agentic AI Market Brief explores these dynamics in-depth, including architectural implications, market drivers, and the standards and operational alignment required to move from AI experimentation to AI at scale. We invite you to download the brief and engage in the ongoing industry dialogue shaping the future of AI-ready networking.
Learn More
- Download the Mplify Market Brief: NaaS: The Automated Network Supply Chain for Agentic AI.
- Learn more about Mplify’s vision for NaaS.
- Read the Mplify NaaS Industry Blueprint.