Every AI company in the world needs more computing power than it can find, and most of that capacity sits in a handful of countries. One European-founded startup has found a faster way to supply it, and it is the kind of business an African founder could build.
Runware makes AI data centres in a factory, as self-contained units that ship to a site and switch on within 48 hours, needing only power and an internet connection. And it’s already looking at mass-producing them on an assembly line, like cars. Africa needs this capacity, and there is no reason the company building it here has to come from somewhere else.
Watch: think data centre in a pod…
Insights on building AI data centres in Africa
The model is what makes it a fit for this continent. A traditional data centre takes years to plan and build on site, with heavy approvals at every stage (a Cape Town one was recently stalled in court and Terraco commissioned a new one, set to only break ground years from now).
Runware's Sonic Inference Pod is built in a factory instead, so most of the work never happens on the site at all. Capacity is added simply by adding more pods, from one unit at a university or bank to dozens on a large campus. Every pod runs as part of one network, so work goes wherever there is spare capacity, and if one pod fails, the rest keep running. The pods use a closed-loop cooling system with air-cooled radiators, so they need no water supply, which matters in dry regions where big data centres face pushback.
Customers can rent computing power from the pods or supply their own chips and pay a monthly fee. The demand is real. Runware says it has had $ 1 billion in committed interest, about R16.5 billion, since it began offering pods to outside customers a month ago.
Building one pod used to take three weeks. Its second factory in Romania, due online early next year, is meant to cut that to a day, and chief executive Flaviu Radulescu is targeting 30 megawatts of capacity a day by 2028, close to a gigawatt a month. The company has raised a little over $ 60 million so far, roughly R1 billion, which shows this does not start with a billion-rand budget.
Next-level: now learn how to mass-produce them…
What others are saying about building AI data centres in Africa
TechCrunch's Dominic-Madori Davis reported Runware's factory plans, noting that at its target pace it would add roughly as much capacity in a year as the whole US was expected to add this year, while warning that execution is the real test. She pointed to electric vehicle startup Arrival, once valued at more than $ 13 billion, which struggled to mass-produce affordably. Her earlier report quoted Radulescu: "Demand for inference is growing faster than facilities can be built." Others, including Crusoe and Schneider Electric, also sell prefabricated data centre pods, which shows the model is spreading.
This could be you
African governments want their own computing capacity rather than renting it from far away, a point underlined when Ukraine's Starlink was throttled by a foreign provider. Banks, universities and telcos want AI capacity close to home. A factory-built pod supplies that at any size, without years of construction on every site. It also spreads the risk, the same reason we argued one giant facility is a single target.
South Africa already has the engineering skills, steel and fabrication, and an electricity utility that is selling less power than it used to, which means sites with spare power exist. The hard parts are real: capital, chips from Nvidia, reliable power and cooling that copes with African heat.
But you do not need to build all of it on day one. Assemble pods for African buyers. Own the siting, installation and operations. Or build the cooling and power systems that go inside them. The demand for compute is not slowing down, and someone is going to own this on the continent.
Go build it.
You might also like our piece on self-driving excavators, HisWay Labs' rail sensors, and Alphabet's AI spending.
Get more SA tech and business news and subscribe to The Open Letter.


