The story Africa likes to tell about itself is digital: mobile leapfrogging, a young online population, and a start-up surge. A World Bank policy working paper from May 2026, "The Economics of Cloud Infrastructure" (No. 11391), introduces an awkward physical fact into that narrative. As cloud computing turns from a bespoke service into a standardised, commoditised global market, the advantage in hosting it goes to wherever the underlying factors are cheapest, and the decisive factor is energy.

The authors apply a classic trade framework, Heckscher-Ohlin factor proportions, to the cloud. They model cloud output as a standardised data-processing unit produced with globally uniform technology, then ask where it can be made most cheaply. The answer: locations with low-cost, reliable energy; robust connectivity; deep pools of skilled labour, and limited regulatory friction or geopolitical risk. That is why data centres cluster where they do and why so much of Africa, for all its connectivity gains, fails the test on the one factor that matters most.

The implication cuts against the prevailing enthusiasm captured in work such as the Observer Research Foundation's "Digital Africa: Tales of Transformation" (2026), which charts genuine progress across the continent. Connectivity and talent are necessary; they are not sufficient. Without firm, affordable power, the compute that trains and runs artificial intelligence will be sited elsewhere, and Africa will export its raw data to be processed abroad and re-import the finished intelligence a digital echo of its commodity trade.

This reframes the continent's AI debate around electricity. The African Development Bank's 2025 Africa Industrialisation Index already ranks economies on the industrial fundamentals, energy among them, that determine who can host heavy production. Data centres are simply the newest heavy industry, and they are pitilessly mobile: they go where the megawatt-hour is cheap and uninterrupted.

There is a path through. Africa's renewable endowment, solar above all, is precisely the kind of low-cost generation the cloud rewards if it can be made reliable at scale and paired with grids that do not fail. The countries that solve firm clean power first will have a credible claim on data center investment; those that do not will watch it route around them.

For governments chasing a digital economy, the message is to stop treating power as a separate file from technology policy. For investors, the logic is simple: follow the reliable electron. Africa's AI ambition is real, but in 2026 it does not die for lack of code or talent. It dies, quietly, at the power plant.