Mmaki Jantjies had a pointed message for her country: South Africa risks remaining a consumer of foreign algorithms rather than a builder of its own. Speaking Thursday at the CNBC Africa AI Summit in Johannesburg’s Sandton Convention Centre, the Telkom SA Group Executive for Innovation and Transformation framed the challenge not as a technical problem but as a question of national will.
Her remarks came alongside those of Sid Wahi, Vice-chairman of CNBC Africa, who opened the day by describing a paradox sitting at the heart of current AI development. The cost of inference for near-frontier-level intelligence has fallen sharply. Yet overall spending keeps climbing, because organizations are using AI far more intensively than before. Cheaper processing has not meant cheaper bills. It has meant more usage.
Wahi laid out how AI systems are evolving toward a model in which humans manage fleets of AI agents. One orchestrator oversees multiple agents, each of which can itself manage hundreds of sub-agents handling specific tasks. For enterprises, two things matter most in that architecture: a context window large enough to process tokens, and cheap inference. More complicated tasks require more tokens, creating a direct relationship between task complexity and cost.
The capital requirements are staggering. Hyperscalers are investing billions in data centre expansion, and the pace is accelerating. Wahi noted that the industry borrowed more in the first seven months of 2026 than in all of the previous year. Google, for the first time in its history, reported negative free cash flow as a result of capital expenditure. Computing power, electricity, cooling, bandwidth and inference budgets are all emerging as hard constraints on growth.
Wahi contrasted the approaches of China and the United States. The United States has relied heavily on private-sector leadership. China treats AI as part of a national industrial strategy, deploying cheaper models into real-world systems rather than chasing artificial general intelligence. Chinese firms have limited access to frontier chips and capital, and they are attempting to compensate for a shrinking workforce through automation and humanoid robots, embedding AI across manufacturing, commerce, surveillance, biotechnology and robotics.
The economics of that divergence are sharpening. Lower-return work will increasingly go to smaller models or cheaper workflows. Frontier intelligence will be reserved for complex problems where its capabilities justify the cost. Adoption is becoming less about what frontier models can do in principle and more about the price and scarcity of the inputs required to run them at scale.
This is where human judgment becomes essential. Wahi argued that the most durable productivity gains will come from using AI to complement human labour, not replace workers outright. Humans will increasingly be responsible for allocating intelligence, deciding which tasks to delegate to AI and which require frontier-level capability or cheaper inference. The danger, he cautioned, is not that AI becomes smarter. It is that humans become less practiced at thinking without it. Organizations that outsource too much thinking, writing, remembering and problem-solving to AI may become highly productive while slowly weakening the very skills needed to judge whether the output is any good.
AI performs well in predictable situations but becomes dangerous when organizations allow systems to make final decisions in circumstances requiring additional context, exceptions and common sense. Wahi’s consulting company, Thinker, was designed not to remove humans from the loop but to remove repetitive work around the human loop. Let the machine handle what is repeatable. Let the person handle what requires judgment.
By contrast, Jantjies focused on what it would take for South Africa and the continent to move beyond passive consumption of AI. The markers of an innovative nation, she said, include domestic research capacity, sovereign digital infrastructure, local data ownership and deliberate investment in homegrown solutions. South Africa’s gross domestic expenditure on research and development stands at about 0.6% of GDP and is declining. Pure emerging and global economies invest between 2% and 4%, with the OECD average around 2.7% and leading innovative economies investing up to 5%.
Telkom is investing in the broader ecosystem through digital skills, an AI institute and enterprise programmes, because a competitive AI economy cannot be built by large corporations alone. Africa should develop technologies and solutions tailored to South African languages and local needs. Telkom is already building language translation technologies and investing in tools for South African and African industries, infrastructure and communities.
Jantjies closed on two interconnected foundations she called non-negotiable: digital infrastructure and human capability. Innovation without infrastructure cannot scale. Infrastructure without skills cannot build capability. The task beyond the summit, she said, is to connect those pieces and move from AI as an experiment to intelligence that drives real economic progress for businesses, communities, the country and the continent. Whether South Africa’s research investment can reverse its decline before that window closes is the question neither speaker left fully answered.