A proposed R100mn ($5.7mn) AI start-up fund for South Africa has a first backer: investor Andile Ngcaba pledged 25% of it at a working lunch with President Cyril Ramaphosa in Silicon Valley.
The fund would pay for computing power, the biggest cost for young artificial intelligence companies. The other R75mn has no committed source yet, and no development finance institution has publicly responded to Ngcaba’s challenge to match him.
What did Ngcaba propose?
Ngcaba, executive chair and founding partner of Convergence Partners and Solcon Capital, tabled the idea on 18 November at a lunch between Ramaphosa and South African micro, small and medium enterprises, held on the margins of the South Africa and the AI Opportunity Conference in California, according to the government news agency SAnews.
He proposed that development finance institutions (DFIs) and private partners set up the AI start-up fund together, and committed to provide a quarter of it himself, about R25mn ($1.4mn).
“I want to challenge my colleagues from DFIs. I will put 25% into a R100 million start-up fund for you [President Ramaphosa] to launch during [the] State of the Nation Address,” he said. “If the [DFIs] are able to come along with us, we will commit that here and now.”
The next State of the Nation Address is normally delivered in February. The business delegation on the trip was led by Stella Ndabeni, the minister of small business development.
Why does an AI start-up fund focus on computing power?
Training and running AI models requires graphics processing units (GPUs), which are scarce and costly, and start-ups usually rent them from cloud providers. “If you are an AI company, access to compute is one of the most expensive parts. This is what burns cash quickly,” Ngcaba said, citing India’s support for its start-ups as an example. He added that the AI start-up fund could grow “depending on consumption”.
Nomso Kana, a nuclear scientist and telecoms entrepreneur at the lunch, called for a bigger budget for the Department of Small Business Development so that emerging firms have “bridge capital” to build their own models, ITWeb reported.

How much money reaches African AI start-ups?
The sum is small by global standards but would not be trivial locally. African start-ups raised about 12% less between January and October 2026 than the $2.7bn of the same period in 2025, and only 343 companies secured at least $100,000, down 22%, according to the tracker Africa: The Big Deal, as reported by The Will.
South Africa already has public seed capital, including a R300mn fund for tech start-ups launched in October 2024 with the SA SME Fund. Minister Ndabeni told reporters in San Jose that entrepreneurs rely too heavily on government and commercial banks and that venture capital is “underutilised”, SAnews reported.
What happens next for the AI start-up fund?
Several questions remain open. Nothing published says who would manage the AI start-up fund, how companies would apply or which providers would supply the computing capacity, and no DFI has confirmed it will contribute, the business site Innovation Village noted.
The proposal also lands in a 90-day window. Ramaphosa has told the roughly 300-member delegation to turn the visit into practical projects within three months of returning, DNE Africa reported, a timetable that would end before a February address. During the Silicon Valley mission he also offered investors ageing power-station sites for AI data centres, though his spokesman said no technical details had been worked out, TechCentral reported.

FAQ
What is the proposed AI start-up fund?
It is a R100mn ($5.7mn) fund, proposed by investor Andile Ngcaba, to help South African AI start-ups pay for computing power. It does not exist yet.
How much has Ngcaba pledged?
He committed 25%, about R25mn, and challenged development finance institutions and private partners to provide the rest.
When could the AI start-up fund launch?
Ngcaba suggested Ramaphosa launch it during the State of the Nation Address, normally held in February. No date or manager has been announced.
Why is computing power so costly for AI firms?
AI models need specialised chips (GPUs) that are expensive to buy or rent, which Ngcaba said “burns cash quickly” for young companies.
