President Ntuli, managing director of Hewlett Packard Enterprise South Africa. (Photograph by Lesley Moyo)
South African organisations risk undermining their AI ambitions if they fail to modernise their networks, according to executives from Hewlett Packard Enterprise (HPE).
Speaking at an event in Rosebank yesterday, President Ntuli, MD of HPE South Africa, and Mandy Duncan, country manager for HPE Aruba Networking South Africa, said while AI adoption is gaining traction, underlying infrastructure, particularly networking, remains a key constraint.
“Networking is under-weighted and under-prioritised, yet it is the foundation of everything,” Ntuli said.
He added that globally, organisations are increasingly adopting generative AI tools, with uptake estimated at around 30%. In Africa, however, adoption remains lower, with roughly 27% of organisations deploying some form of AI.
Ntuli noted that while interest is high, many organisations are adopting AI reactively.
See also
“There’s pressure from boards asking what organisations are doing about AI, but often there isn’t a clear strategy behind the deployments.”
This is leading to fragmented implementations across business units, with limited coordination and unclear long-term value, he noted.
HPE argues that while much of the AI focus is on compute, GPUs and cloud, the network is the most overlooked component.
AI workloads are data-intensive and require high bandwidth, low latency and reliable connectivity across distributed environments. Legacy networks, designed for basic connectivity, are not equipped to handle these demands.
Mandy Duncan, country manager for HPE Aruba Networking South Africa. (Image: Supplied)
Duncan pointed out that current enterprise networks are “not built for AI”, limiting organisations’ ability to scale beyond initial use cases.
“If the underlying infrastructure is not stable, the AI layer cannot perform,” she said.
According to HPE, companies need to move beyond basic connectivity and build networks that are intelligent, secure and scalable.
Duncan said HPE is embedding AI into networking platforms to enable predictive analytics, automation and improved performance.
“We are moving towards autonomous networks that can identify and resolve issues before they impact users.”
In a recent ITWeb TV interview, Ntuli said Africa can write its own narrative regarding investment in AI and secure its share of a projected $15.7 trillion in the AI global economic value chain.
According to Ntuli, for the continent to benefit from AI, role players like the African Union can take the lead from Europe and foster collaboration between key stakeholders.
According to Statista, the global AI market was valued at $244 billion in 2025 and expected to exceed $800 billion by 2030.
Meanwhile, IDC estimated that global AI spending will reach $235 billion in 2024, increasing to more than $630 billion by 2028.
Ntuli said countries that are leading in AI have already invested significantly in infrastructure, including data centres, connectivity and skills.
He highlighted that in South Africa, while investment in data centres is increasing, gaps remain in networking, data architecture and skills development.
“There is strong intent from both the public and private sectors, but infrastructure investment still needs to catch up.”
He added that data readiness is also a key issue, with many organisations collecting data but not effectively using it to generate insights.
Concerns around the cost of AI infrastructure remain, particularly for smaller organisations. However, Ntuli said AI is becoming more accessible through alternative consumption models.
These include renting GPU capacity on demand, using cloud-based AI services and deploying pre-configured solutions for specific use cases.
“AI is affordable, depending on what you are trying to achieve,” he said.
Skills and operational challenges
Skills shortages continue to be a constraint, particularly in networking and AI.
Duncan noted that traditional networking roles are evolving, requiring new skills in automation, AI and data analysis. At the same time, organisations are under pressure to simplify operations and reduce costs.
AI-enabled networking can help address some of these challenges by automating routine tasks and reducing the need for manual intervention.
HPE estimates that automation can significantly reduce IT support tickets and improve operational efficiency.
While upgrading network infrastructure requires investment, Ntuli said the cost of not doing so is higher.
AI initiatives are likely to stall without the necessary foundation, limiting organisations to pilot projects rather than scaled deployments.
“AI is network-intensive and data-intensive. Without the right infrastructure, organisations will not realise its full value,” he said.
HPE recommends a layered approach to building AI-ready environments, starting with reliable connectivity, modern network infrastructure, compute and storage capabilities, integrated security and a clear data strategy.
These elements must be aligned to support AI workloads and business outcomes.
As organisations move from AI experimentation to scaled deployment, the focus is shifting to what enables real outcomes. For HPE, that means rethinking the network as a strategic asset rather than a background utility.
Without the right foundation, AI risks remaining fragmented and underdelivering. But with intelligent, secure and scalable networks in place, organisations are better positioned to turn AI investment into measurable business value, according to HPE.
