South African data centre providers are warning that cloud computing and web hosting costs in the country are set to rise as infrastructure expenses increase.
This, as surging global demand for artificial intelligence (AI) infrastructure has pushed up prices for key server components, such as solid-state drives (SSDs) and DRAM memory.
Escalating electricity costs and the need for backup power are also raising operational expenses for local data centres.
Industry players say these pressures could soon translate into higher cloud and hosting prices for businesses.
According to Ziaad Suleman, CEO of Cassava Technologies in South Africa and Botswana, global trends suggest infrastructure costs across the technology sector are under pressure.
“South African companies should prepare for sustained increases in cloud and hosting fees throughout 2026. However, the impact on South African organisations will vary depending on factors such as where their workloads are hosted and the efficiency of the underlying infrastructure. For companies hosting their services in highly-efficient, locally-based facilities, the effect may be more contained,” says Suleman.
He adds that the explosive demand for AI infrastructure has driven data centre construction costs to an average of $10 million per MW.
“This cost, coupled with rising enterprise system prices, is fuelling a strategic market correction in which South African organisations are shifting their priorities from ‘cheap’ cloud models to ‘resilient’ cloud environments.”
Businesses are increasingly opting for high-redundancy facilities to mitigate financial risks of downtime and regulatory non-compliance, Suleman says.
Constrained hardware supply
Andrew Cruise, MD of evoila Africa (formerly Routed), explains that ultra-high-value agreements between major public cloud providers such as Microsoft Azure and Oracle, and AI firms including Anthropic and OpenAI, have overwhelmed global hardware supply chains.
He says these deals, tied to AI training and inferencing infrastructure over the next three to five years, are driving massive demand for memory chips in high-bandwidth memory for graphics processing units (GPUs) and tensor processing units, as well as storage systems for vast AI datasets.
According to Cruise, supply constraints are worsened by the dominance of only three major memory manufacturers – SK Hynix, Micron Technology and Samsung Electronics – whose output is largely committed to AI projects. AI servers also rely heavily on CPUs from Intel and AMD, he notes.
“The South African market is so small that we have no purchasing power, and we are already seeing substantial price increases across both consumer and enterprise hardware. It is important to understand that this is not a temporary ‘blip’, but an ongoing structural change, which will perpetuate and, likely, escalate,” Cruise says.
“Further, the issue is not solely cost of hardware but, critically, availability of inventory – many OEMs [original equipment manufacturers] such as HPE and Lenovo just don’t have stock and don’t know when they will. Clearly, some OEMs did not have solid supply agreements with the memory manufacturers.”
Cruise warns that cloud providers with insufficient scale may be forced to restrict growth on their platforms due to limited infrastructure capacity.
“Performance and reliability could deteriorate if older equipment is kept in operation beyond its intended lifecycle instead of being refreshed. Cloud prices will rise broadly across the market. Only providers with sufficient scale, strong growth and reliable access to server inventory will be able to mitigate these pressures and limit or avoid price increases, particularly when adjusted for inflation.”
Suleman adds that high inflation in core materials like steel and cement, currency fluctuations in imported tech, and cautious staged construction are contributing to rising costs.
He notes that providers now compete on fibre diversity and access to large-scale renewable power agreements, not just space.
“Several global supply chain and technology trends are contributing to the upward pressure on infrastructure costs. Demand for high-performance computing hardware is increasing rapidly, driven largely by artificial intelligence workloads and the expansion of cloud services worldwide.
“Components such as DRAM, NAND flash memory, GPUs and specialised processors are experiencing strong demand, which can impact pricing and availability. Energy costs are also a significant factor globally, as data centres require reliable and large-scale power to support digital services. These combined pressures are influencing the cost structure across the global cloud and hosting ecosystem.”
Cruise urges that, with an unstable power grid, data centres must invest in diesel reserves, redundant transformers and private power agreements, yielding premium pricing.
Mounting pressure
Derrick Chikanga, research manager for IT services at IDC, says cloud service costs are mostly driven by enterprise demand, accelerated by AI workloads, with hosting prices influenced by ISPs seeking to reduce IT expenditure.
“The prices will translate to increased service fees for both enterprises and consumers. This will be more prevalent in some sectors, such as telecoms and ISPs, compared to others,” he says.
Ofentse Dazela, director of pricing research at Africa Analysis, adds that global AI demand, rising energy and infrastructure costs, and a shift toward profitability are driving up cloud and hosting prices after years of decline.
“Global cloud price increases will place significant financial pressure on South African consumers and businesses, compounded by a volatile rand and high electricity costs. While consumers may experience the impact indirectly through higher subscription fees, enterprises will face more direct cost increases, requiring improved cloud cost management, contract renegotiation, and stronger financial discipline to mitigate rising expenses.”
Chris Ngwenyama, director of Africa Analysis, is of the view that the huge AI infrastructure buildout has led to a shortage of higher-performance GPUs as OEMs try to meet demand, mainly by hyperscalers and data centre operators.
He says cloud providers and hyperscalers such as AWS, Google Cloud and Microsoft have announced large-scale AI infrastructure buildout plans with high memory requirements, which OEMs may not be able to meet.
According to Ngwenyama, this also has the potential to disrupt the supply chain, which may result in delays in completing the buildout of these facilities.
“The supply chain disruption is also compounded by geopolitical factors, including higher tariffs imposed by the US and the war in the Middle East. With all these factors – the shortage of high-performance memory, disruption of the supply chain and geopolitics – demand and supply constraints are converging, leading OEMs to steadily push their prices upward by anything between 5% and 30% or more.
“These increases will be passed through to consumers as cloud service providers can no longer absorb the operational costs alongside power, network, facility and staff costs. The impact on consumers may lag a bit by a couple of months, especially from vendors that have large inventories, and also due to the complex contract structure between vendors and enterprises. But consumers will ultimately feel the pinch, possibly by the third or fourth quarter of this year.”
Share