Andrew Fox will be delivering the keynote address at HPE Networking Day in Sydney on 7 October 2026.
.jpg&h=420&w=748&c=0&s=0)
The first is that AI is actually improving the network itself. Known as “AI for networking”, this refers to using AI to make the network itself more autonomous and intelligent. The second is that we need to build networks that can support growing AI workload demands, referred to as “networking for AI” Although easy to conflate, the two concepts are clearly distinct, yet essential to consider together in the AI era.
“AI for networking” makes networking smarter. We see AI being embedded into the network itself, transforming it into a more autonomous, self-driving platform that can predict issues, strengthen security and automate operations across increasingly complex environments. It enables networks to adjust rapidly and keep pace with heavier system demands and growing security challenges.
These self-driving networks don’t need to wait for a user to find and report a connectivity issue, AI-native networks can identify abnormal behaviour and take action before it impacts the business. Increasingly digitally-reliant organisations will be relieved, as they cannot afford system downtime or security failures, particularly those in critical industries like healthcare and banking. Here, we see network shift from being reactive to becoming proactive, freeing up IT teams to focus on governance and strategy rather than time consuming repetitive fixes.
These proactive AI-native networks act as vital security sensors for organisations grappling with an ever-expanding attack surface that comes with the proliferation of “edge” devices, such as smart sensors and routers. As the number of devices connected to the internet offering a pathway into enterprise networks grows, even unskilled malicious actors are making trouble for organisations. Recent research by the Australian Signals Directorate’s Australian Cyber Security Centre (ASD’s ACSC) identified 17.9 million devices that were visible to the public internet over a two month period, including roughly 212,000 “edge” devices With improved visibility across users, devices, applications and traffic, AI-native networks become a vital first line of defence that can help automate rapid threat responses.
“Networking for AI”, on the other hand, relies on AI-native networking, but focuses on the foundations needed to turn AI ambition into operational reality. As AI becomes embedded across the enterprise, networking is shifting from a supporting function to a determining factor in how successfully organisations can scale AI.
AI workloads behave very differently from traditional enterprise applications. They move larger volumes of data, generate new traffic patterns across environments, and place far greater demands on latency, throughput, reliability and visibility. As organisations connect AI models, data sources, applications and users across distributed environments, the network increasingly determines how effectively those systems perform. It is no longer simply carrying traffic between systems. It is becoming a critical part of how organisations deliver AI outcomes at scale. The urgency becomes clearer when we look at the pace of AI adoption.
Australian Bureau of Statistics figures show the share of Australian organisations using AI increased from 1% in 2022–23 to 12% in 2024–25. As organisations move from experimentation to production deployments, attention is naturally shifting from AI models themselves to the infrastructure that supports them. Every AI initiative ultimately depends on the ability to move data efficiently, connect distributed resources and deliver consistent performance at scale.
Networks built for a previous generation of traffic patterns were not designed for the volume, latency requirements and data movements that AI workloads generate. Over time, that gap becomes a constraint on how far organisations can scale their AI ambitions.
There is no one-size-fits-all approach to solving that challenge. Rather than assuming everything will move into a single environment, organisations need architecture and network systems that can support workloads across the entire business as requirements evolve.
Where AI workloads run will depend on more than performance and cost. Security, business continuity, digital sovereignty and latency requirements will also shape where data resides, how infrastructure is managed and who maintains control.
This makes adaptability essential. Organisations need network infrastructure that can support AI workloads across different environments and evolve as their requirements change.
The same principle applies beyond AI. A network that can accommodate new environments and use cases without having to be rebuilt also gives the broader business more room to grow.
We’re already seeing this play out across distributed environments in Australia. Mercy Health, for example, is modernising connectivity across more than 40 sites through a cloud-managed network designed to improve visibility, reliability and performance across its operations.
Importantly, the network is not only supporting what Mercy Health needs today. By simplifying operations across a large, distributed environment, it also gives the organisation a foundation for future digital health capabilities and continued growth, without adding operational complexity.
It demonstrates a broader principle. The right network solves what is in front of the business today, while leaving room to adapt as technology, workloads and requirements change. AI for networking and networking for AI are not separate projects. We must treat them as one strategy, enabling organisations to manage growing complexity now while preparing their infrastructure for what comes next.
My advice, build for what the business needs now, but don't let today's requirements dictate a network that can't flex tomorrow. That's the standard AI is going to hold every network to.
