Ex-Nvidia engineer starts company to bring AI computing power to Africa

Alexander Tsado helped set up early cloud chip systems at Nvidia, the first $5 trillion company. He has now left to start Udu Technologies, which offers flexible, affordable access to AI computing power for startups and institutions in six African nations.

What Udu does

Udu sells access to cloud GPUs. A GPU is a powerful computer chip that AI systems depend on. The company also supplies computing hardware and AI advice. Its customers are businesses and institutions in the United States, Africa, the Caribbean and beyond.

Why he left Nvidia

Tsado was a product manager on cloud computing at Nvidia. From 2016 to 2020 he helped set up some of the company's early cloud GPU systems for Amazon Web Services. He says he kept noticing that Africa was barely part of the AI conversation.

He recalls two prominent African businessmen who contacted Nvidia and got no reply. He then wrote to the company's CEO, who answered within 10 minutes and had the sales team follow up. Tsado says the lesson was bigger than that one reply.

"There was nothing set up to serve our continent, even if our continent got interested," he said.

The spending gap

• Five of the largest AI data centre companies, including Amazon and Google, are on track to spend $900 billion on AI in 2026.
• That is roughly as much as they spent in the previous three years combined.
• China plans about $295 billion over five years on data centres.
• The US announced plans for up to $500 billion in private investment in AI infrastructure.
• Africa has not made comparable investments, according to the article.

Cost and caution

Powerful computing equipment is expensive, and the electricity to run it adds more cost. Tsado also argues that African investors and governments tend to wait until an idea has proved itself. Other regions, he says, spent two to four years and hundreds of millions of dollars on research before seeing returns.

Udu wants to shorten Africa's adoption period from 10 years to two or three. Its focus is what Tsado calls homegrown AI, where organizations use their own data to build tools for their own needs. "It's not something you copy and paste and it will work well," he said.

What is not known yet

The article text provided does not give Udu's prices, customer numbers, funding or the six countries it serves. It also cuts off before fully explaining how Udu makes access cheaper. For smaller businesses, the idea is that renting computing power as needed could allow AI trials without a huge upfront purchase.