Why the diagnosis holds
Manamela argued that AI makes the university’s traditional work more urgent: questioning arguments, testing evidence and taking responsibility for knowledge. AI systems can produce content without establishing whether it is true.
The labour market gives that argument weight. Graduates recorded 12.4% unemployment in the second quarter of 2026, against 39.2% for people without matric, and youth unemployment rose to 47.4%. But the protection is not guaranteed. In the first quarter, unemployment fell year on year among those with below-matric education but rose among matriculants and graduates. A degree still protects, but the margin is not fixed.
Building requires inputs still being assembled
Policy. Cabinet approved the draft National AI Policy on 25 March 2026, with public comment open until 10 June, and full implementation is expected in 2027 or 2028. Its six pillars include capacity and talent development. The 86-page draft also proposes seven new oversight bodies. In my view, the risk is that governance structures arrive before funded programmes do.
Language data. Manamela asked whose languages AI represents. None of the top 34 languages used on the internet is African. Progress is real but largely community- and donor-driven. The Africa Next Voices initiative captured 9,000 hours of speech across 18 languages, including isiZulu and Tshivenda. The Masakhane hub, supported by Google.org, the FCDO, IDRC and the Gates Foundation, is funding datasets for 50 languages. What appears to be missing is a durable public funding line.
Compute. Capacity is arriving, mostly through private operators. Stratos Lab is deploying 400 Nvidia B300 GPUs at a Centurion data centre, and Cassava has taken delivery of 3,000 GPUs in South Africa. Even so, Africa accounts for less than 1% of global data centre capacity. Dependence is visible at the research level too: Lelapa AI obtained compute credits from Microsoft’s AI for Good Lab to train its InkubaLM model. For universities, the terms of access matter as much as availability.
What universities should decide
Manamela urged universities to move beyond general discussion and make practical choices about teaching, assessment and research capacity. Three seem most consequential, in my analysis:
Assessment that tests judgement. Oral defences and supervised problem-solving make it harder to outsource thinking than take-home essays do.Joint programmes pairing computing with the humanities and social sciences. He noted that questions about bias, power, labour, law and human dignity cannot be settled by computing power alone. Negotiated compute access and data stewardship. Researchers should not be renting capacity entirely on others’ terms. Universities could take stewardship roles in language datasets.He also said technological change will bring people back to learning throughout their working lives, while qualifications must keep their depth and standards. Modular, stackable credentials would be one practical answer. The real measureThe speech sets the right test, but speeches do not fund GPUs or datasets. Four things will show whether it translates into capacity:the final AI policyfunded compute access for researcherssustained support for African-language datawhether graduate unemployment keeps rising in coming labour force surveys.
Written by:
*Dr Iqbal Survé
Past chairman of the BRICS Business Council and co-chairman of the BRICS Media Forum and the BRNN
*Sesona Mdlokovana
Associate at BRICS+ Consulting Group
Africa Specialist
**The Views expressed do not necessarily reflect the views of Independent Media or IOL.
** MORE ARTICLES ON OUR WEBSITE https://bricscg.com/ (https://bricscg.com/)
** Follow @ (https://x.com/brics_daily)brics_daily (https://x.com/brics_daily)on Twitter for daily BRICS+ updates and instagram @brics_daily (https://www.instagram.com/brics_daily?igsh=bmhvbTd0YzA4a2wx)
