Africa Needs a Grid Upgrade to Support its AI Ambitions

As artificial intelligence transforms economies globally, African nations face a critical question: are they building the right infrastructure to support this technological leap? While governments often focus on generating more electricity, companies like Schneider Electric argue that upgrading existing grids is equally essential.

“The traditional grid as we know it is already becoming obsolete,” says Ifeanyi Odoh, president of Schneider Electric East Africa. “We need digitally enabled grids from the start because how electricity is generated and consumed has fundamentally changed.”

Africa currently possesses less than 1% of global data centre capacity—just 360 megawatts across 217 facilities in 33 countries. This gap highlights a fundamental challenge: AI infrastructure demands exponentially more power than traditional systems.

The Changing Economics of Electricity

AI racks already draw 40-120kW, with next-generation systems approaching 200kW per cabinet—significantly higher than conventional servers (5-15kW). This requires continuous, uninterrupted power and sophisticated cooling solutions.

Moreover, data centres are increasingly located near population centers rather than industrial parks. For instance, most of Lagos, Nigeria’s 18 data centres operate within residential and commercial areas.

Three Paths to AI Readiness

Different African nations are pursuing unique approaches:

  • South Africa: Following improvements in grid reliability, Microsoft announced a R5.4 billion ($300 million) expansion of its cloud region.
  • Kenya: Leveraging abundant geothermal energy (over 70% of electricity mix), developers are locating data centres directly beside power plants—Microsoft’s $1 billion AI campus with G42 exemplifies this strategy.
  • Nigeria: With significant gas reserves, Schneider proposes a “gas-to-compute” model where dedicated power infrastructure is integrated into data centers themselves.

Each approach addresses the same fundamental need: reliable, modern electricity to support AI workloads.