US software giant Microsoft last week announced plans to invest R5.4-billion in building cloud computing and AI infrastructure in South Africa.
The move was lauded by President Cyril Ramaphosa, who spoke at the briefing where Microsoft’s plans were announced and described the software company’s “longstanding presence” in South Africa as a “vote of confidence in our economy”.
The announcement was well received by stakeholders in the broader IT sector as well, including International Data Corp (IDC) associate research director for IT services in sub-Saharan Africa Jon Tullett.
“Microsoft’s investment in South Africa is certainly significant and very welcome but must be considered in light of the company’s broader investment commitments in AI and cloud, which amount to US$80-billion in 2025 alone, according to Microsoft,” Tullett told TechCentral. “South Africa’s $300-million over three years is a relatively minor share of that but still a welcome one.”
The investment is significant for two reasons: the first is directly related to the impact that access to AI computing power will have on the economy. The second has to do with the political implications of bringing AI-capable graphics processing units (GPUs) into South Africa.
To the first point: Microsoft president and vice chairman Brad Smith shared the company’s perspective on the impact of AI on the economy at the announcement event. As Microsoft sees it, access to AI tools is going empower South African businesses, entrepreneurs, students and innovators to use AI tools to solve problems.
Economic impact
A good example of this is a start-up called Lelapa AI. Lelapa creates large language models of indigenous African languages. These tools allow citizens for whom English is not a first language to interact with critical services like banking apps in their home language using text or voice interaction. The idea is that with more access, South Africa will see more Lelapa-type solutions.
The result, said Smith, will be the development of new value chains and even industries that will have a direct impact on South Africa’s GDP.
Last August, JSE-listed Naspers – in collaboration with the Mapungubwe Institute for Strategic Reflection – released a report predicting that digital platforms will add R91-billion to the South African economy over the next 10 years. In the report, AI is identified as one of the main drivers of this digital-led boom.
Read: Data centres are straining electricity grids worldwide
The second reason the investment by Microsoft is significant relates to the increasingly important role that AI hardware and software plays in geopolitics. Nvidia is the world’s leading producer of chips used in AI applications. Access to Nvidia’s AI-capable GPUs has become a bone of political contention, with the US wanting to bar its geopolitical enemies, especially China, from using American-made tools to advance AI technology in ways that could threaten US leadership in the field.
One of the last actions by the Biden administration, prior to Donald Trump’s inauguration in January, was to enact the AI Diffusion Framework, which divided the world into three tiers and placed limitations on the AI compute that US firms could export.
Tier-1 counties are considered to be US allies and enjoy unrestricted access to AI hardware and software developed by American technology firms, including Nvidia. Tier-2 countries, which include South Africa, have limits on how much AI GPU compute they can buy. Tier-3 nations are restricted from importing any American-made AI tools, including GPUs and software.
Read: Microsoft’s eye-popping data centre investment plans
How this relates to Microsoft is that the framework defines a “Validated End User” scheme, which sets out which American companies are allowed to export US-made AI technologies abroad. The gist of it is that if a non-US company wants to build an AI-powered data centre outside the US, the likelihood it will be allowed to buy the Nvidia chips it needs is much lower than if Microsoft asks to buy them.
One of the major challenges threatening to get in the way of South Africa’s ability to capitalise on AI is the country’s ongoing energy crisis. AI workloads consume orders of magnitude more power than the CPUs traditionally used in hyperscale data centres, meaning they place additional strain on the electricity grid.
“Investment in AI infrastructure is likely to escalate the power demands of large data centre facilities, which are growing in consumption all the time. We expect to see that increasing the focus on independent power producer agreements. It is a natural evolution, and AI is certainly power-hungry. At the same time, more efficient AI models like China’s DeepSeek are coming to market, which may change the power equation,” said the IDC’s Tullett. – © 2025 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.
