South Africa Must Build Sovereign AI Capability to Avoid Dependency
Published by AINave Editorial • Reviewed by Ramit
HPE South Africa Managing Director President Ntuli argues that South Africa must treat sovereign artificial intelligence as a strategic economic priority to avoid becoming dependent on global providers . The country is making progress through initiatives like the University of Cape Town’s African Compute Initiative and Cassava Technologies’ R3.6 billion investment, but regulatory uncertainty remains a critical factor for long-term investment in compute, models, and AI products.
Why Sovereign AI Matters Now
Sovereign AI, in Ntuli’s framing, does not mean technological isolation. It means maintaining sufficient control over strategically important data, local computing infrastructure, cybersecurity capabilities, AI skills, and regulatory frameworks . Ntuli draws a parallel with South Africa’s earlier cloud computing adoption, where reliance on a small group of international providers created long-term dependencies. “We cannot afford to repeat that mistake with AI,” he says. President Cyril Ramaphosa has also emphasised digital sovereignty, stating at the July 2026 Google Cloud Summit that “sovereignty in the digital age increasingly depends on a country’s ability to secure its data, develop domestic digital capabilities and exercise meaningful control over technologies on which its economy depends” .
Building Sovereign AI Capability: Infrastructure Investments Take Shape
South Africa is not starting from zero. One significant development is the University of Cape Town’s African Compute Initiative (ACI), announced in March 2026, which aims to establish Africa’s largest GPU-intensive computing cluster dedicated to AI research at a higher education institution . The cluster will combine modern GPUs, multi-petabyte secure storage and high-speed networking, enabling local development, fine-tuning and testing of AI systems. It is part of the AI for Development programme, a $58 million partnership co-funded by the UK Foreign, Commonwealth and Development Office and Canada’s International Development Research Centre. UCT expects the cluster to support about 100 active users in its first year and reach 300 users across at least five institutions by year three .
Private-sector investment is also expanding. In April 2026, Cassava Technologies pledged R3.6 billion in South Africa through Liquid Intelligent Technologies, Cassava Intelligence South Africa and Africa Data Centres, planned over 24 months to expand infrastructure and technology capabilities .
Constraints and Policy Gaps
An International Monetary Fund analysis from 2026 estimated that wider AI adoption could raise Sub-Saharan Africa’s GDP by around 4% cumulatively over the coming decade under stronger adoption scenarios . However, the IMF identified electricity supply, digital infrastructure, internet connectivity, technical skills and institutional capacity as major constraints. Productivity gains could range from 0.2% under limited adoption to 2.1% under broader scenarios .
Regulatory uncertainty adds another layer. South Africa’s Cabinet approved a Draft South Africa Artificial Intelligence Policy for public consultation in March 2026, but it was later withdrawn after problems with references in the document . Communications and Digital Technologies Minister Solly Malatsi then announced an Independent Expert Review Panel to assist with revising the policy. For businesses considering major investments in computing infrastructure, models and AI products, this regulatory uncertainty becomes an important component of long-term investment decisions.
A Practical Model: Global Technology, Local Control
Siemens Energy’s deployment of a dedicated global high-performance computing platform through HPE GreenLake in June 2026 illustrates how organisations can combine advanced global technology with greater control over strategically important systems . The sovereign private-cloud environment, deployed in the United States and Germany, supports engineering simulations, digital twins and predictive maintenance, and HPE says some accelerated simulation workflows are expected to reduce processes from days or weeks to hours. Ntuli suggests similar thinking could apply to South African infrastructure operators like Eskom. The key distinction: participate in global technology ecosystems from a position of choice, not structural dependency. As Ntuli puts it, “The goal is not isolation from global technology ecosystems. It is ensuring that we are building enough of a foundation that we create choice, not dependency” .
