Vodacom is pursuing a regional approach to becoming an “AI‑first” company, prioritising investments in digital infrastructure that combine internal efficiencies with enterprise offerings across Africa, Business Day reported.
AI strategy driven by efficiency and market demand
Dejan Kastelic, Vodacom group chief technology officer, told Business Day the operator is using artificial intelligence for network optimisation, power management and other internal processes while also developing external services for enterprise customers. The shift forms part of a broader push to convert business units into AI‑centred organisations, a change Kastelic said affects “people, processes, tools”.
“What we’re looking at now is how we convert our companies into being AI‑first. And that’s the piece that’s keeping us busy, not only me but all of us … literally people, processes, tools,” Kastelic said, according to Business Day.
Growth in digital infrastructure has accelerated since the advent of cloud computing, and has been further turbocharged by demand from generative AI workloads. Those workloads are expensive to run and require specialised hardware such as GPUs. Vodacom is therefore navigating a strategic choice: rely on hyperscale cloud providers or deploy its own on‑premises GPU capacity.
Building versus buying: a strategic fork
Kastelic described the industry debate as almost binary: either enterprises go “left” and use public cloud services — which demands tight cost and efficiency management — or they invest in building their own compute estates. He highlighted the practical trade‑offs facing operators and enterprises as AI compute costs rise and some data‑centre projects worldwide are paused.
“It’s almost like two beliefs, right? As an enterprise now, should we go left and use the cloud, but we’d need to be very efficient ... and also manage the cost, versus building it,” Kastelic said, according to Business Day.
Vodacom has engaged partners including Nvidia and regional systems integrators to shape its roadmap. The group has also moved from concept to construction on at least one major project outside South Africa. In Egypt, through Vodafone Business, Vodacom and partners Elsewedy Electric and Cassava Technologies have started work on a large facility under the Africa Data Centres brand that will host both internal and external GPU capacity.
According to Business Day, the Egypt facility will be divided into two operating spaces: one for Vodacom’s internal use and one for third‑party customers. The operator said it plans to run its own GPUs in that site.
What this means for South Africa and the region
The debate between building local compute capacity and relying on multinational cloud providers has practical implications for sovereignty, latency, cost and industry development. For South African enterprises and public bodies weighing AI deployments, the choices include:
- outsourcing to hyperscalers such as Amazon, Microsoft or Google, which may offer scale and managed services;
- using local partners such as Teraco for colocation or regional data‑centre providers that might be closer to users and data;
- building proprietary GPU farms to retain control over data, models and costs for targeted, sovereign AI use cases.
| Option | Main consideration |
|---|---|
| Public cloud | Scale and managed services vs ongoing cost control |
| Colocation / regional providers | Local presence, potentially lower latency, partnership model |
| Build own GPU capacity | Greater control and sovereignty but high capital and operational cost |
For mobile operators such as Vodacom, the choice also affects product strategy. Owning GPU capacity can enable tailor‑made, sovereign AI services for specific African markets — an approach Kastelic said is under consideration in countries such as Egypt and Kenya — while cloud reliance can speed time to market for some services but may expose customers to higher recurring costs.
As local demand for AI grows, enterprises and infrastructure providers will face pressure to balance cost, performance and data governance. Vodacom’s regional moves, partner engagements and the Egypt project signal that telecommunications companies are looking beyond connectivity to play a larger role in the AI compute stack.
What remains uncertain is how widely operators will commit to costly on‑premises GPU builds in the face of volatile AI economics, and whether regulatory or industrial policy in South Africa and across the region will favour local infrastructure investment. For now, Vodacom’s strategy emphasises a pragmatic mix of partnerships and selective build‑outs to serve both internal needs and paying customers, Business Day reported.
