Cloud dependence exposes African AI startups to hidden cost risks – Expert

African AI startups face hidden financial and operational risks due to their dependence on cloud computing providers, particularly when they rely heavily on a single provider, according to Oluwaseyi Ayodeji, senior programme leader in AI and cloud infrastructure and founder of Regal Stack. Renting computing capacity is not inherently problematic for startups, but the issue lies in how much growth and margins are exposed to the terms set by the cloud provider.

The cost of building a single megawatt of AI-ready capacity in Africa ranges between $600,000 and over $10 million, primarily due to backup power requirements and currency risks associated with equipment imports. This makes renting a more practical option for most startups, as it allows them to provision capacity quickly and conserve cash for growth.

However, the trade-off is that startups do not control the terms of their infrastructure relationship with the provider. Investors should assess how concentrated a startup's computing expenditure is with a single provider. If 100% of a startup's workload is through one hyperscaler, the provider holds significant leverage over the company's cost base.

Ayodeji recommends examining the percentage of computing spend going to a single vendor and the impact of a 20% price increase from that provider on the company's runway. The global cloud market's concentration amplifies this exposure. Nigeria, the second-largest data-center market in Africa after South Africa, is strengthening its domestic cloud infrastructure in response to growing demand for cloud computing and AI services. However, power challenges remain a concern as the country's data-center footprint expands.