Cisco report: AI strains Australian networks as upgrades lag

Australian organisations are accelerating network modernisation as artificial intelligence increases infrastructure demands, but only 17% believe their networks are fully prepared for expected growth.

Cisco and Foundry research found 93% of Australian organisations said AI had accelerated their network modernisation plans. Of those, 29% said it had significantly accelerated their plans.

Budget pressures are constraining those efforts. Nine in 10 Australian respondents said limited budgets were restricting their ability to modernise networks in response to AI growth.

The Australian findings form part of a global study of 3,472 CIOs and leaders responsible for networking, end-user computing and technology. Respondents came from Asia-Pacific, Europe, the Middle East, Latin America and North America and worked for organisations with at least 500 employees.

Upgrade pressure

More than four in five Australian organisations, or 81%, said network upgrades would be required to support projected AI growth.

The results show a gap between AI adoption plans and the infrastructure needed to support them. Australian organisations are also more confident in their AI strategies than in the networks carrying those workloads.

Two-thirds of Australian organisations, or 66%, said they were more confident in their AI strategy than in their network's ability to support it. The global figure was 75%.

Globally, 76% of organisations said their campus and branch environments required upgrades to support current and future AI workloads. More than one-third were already reporting increased capacity demands linked to AI.

Campus and branch networks connect employees, devices, applications and systems outside centralised data centres. The research found these networks were experiencing changes in traffic volume and behaviour as AI adoption expanded.

Capacity limits

Capacity is becoming an immediate concern for Australian technology teams.

Almost three-quarters of Australian organisations, or 74%, expect AI workloads to push their networks to capacity. Nearly two-thirds, or 63%, expect this to happen within the next two years.

The global findings show similar pressure. Some 73% of respondents said they were already facing, or expected to face, campus and branch capacity limits within 24 months.

Global respondents reported an average 34% increase in campus and branch network traffic tied to AI workloads over the previous 12 months. They expected traffic to grow another 96% over the following year. The research projected traffic associated with generative, agentic and physical AI to rise by a compounded 209% over three years.

AI is also changing how traffic moves across corporate networks.

Globally, 67% of respondents reported increases in east-west traffic, which covers communication between internal devices, servers and applications. Another 61% reported growth in continuous automated traffic generated by AI systems.

Performance demands

Australian organisations also identified bandwidth and latency as constraints on AI performance.

Some 64% said AI performance was highly sensitive to bandwidth availability, while 60% said latency, or network delay, was a major factor.

AI systems can generate persistent traffic between systems, require rapid responses and operate across distributed environments. Agentic systems can also communicate autonomously across applications and infrastructure.

The global research found AI workloads were becoming more distributed, dynamic and sensitive to latency and reliability than many traditional enterprise applications. Half of AI-driven demand was concentrated in wireless networks within campus locations.

These performance requirements are becoming more important as organisations move from isolated AI trials to applications embedded in operational processes.

AI workloads

Generative AI is expected to place the greatest pressure on Australian networks.

Some 73% of Australian organisations expect generative AI to create the strongest network demand. Agentic AI followed, with 68% expecting it to place major demands on infrastructure.

Agentic AI systems can perform tasks and initiate actions with less direct human involvement. Their network requirements differ from conventional software because agents may continuously exchange information with other systems and applications.

The global survey found organisations expected deployment of generative AI, agentic AI and physical AI to increase over the following 12 to 24 months. Organisations with broad generative AI deployments were also more likely to report traffic growth, operational complexity, security concerns and infrastructure readiness challenges.

Only 30% of these more aggressive AI adopters said they were fully prepared to support projected AI growth across their networks.

Security pressure

Network capacity is not the only challenge emerging as AI use expands.

Globally, respondents identified security complexity as the most significant challenge associated with AI-driven network demand. Some 78% expected security risks to increase as adoption expanded beyond generative AI. Another 77% said AI had already expanded their attack surface over the previous 12 months.

While 86% of global respondents had added security controls for AI workloads, 61% were holding back from scaling AI initiatives until they had greater confidence in their security posture.

Visibility is another challenge. AI projects can be deployed across individual teams and business units without central oversight. This can make it harder for network and security teams to identify workloads and understand how traffic moves through their environments.

Business risks

Delaying network upgrades could also increase business costs.

Globally, 75% of IT leaders agreed organisations could face higher long-term costs from reactive upgrades or remediation. About 73% cited an inability to meet customer expectations as a risk of delayed modernisation. Another 72% identified missed business opportunities and increased security risks.

Australian organisations are bringing forward network plans, but the local results indicate budgets could limit how quickly those projects proceed.

The study covered organisations with an average of 3,292 campus and branch locations. It also included six in-depth executive interviews across Asia-Pacific, Europe and the United States.