Retired supercomputers power Africa's problem-solver pipeline

Big science hardware gets second life as students train and compete in high-performance computing  

In February 2026, CSIR senior engineer Samuel Mathekga delivered decommissioned scientific computing resources to students at the University of the Witwatersrand (Wits). These research-grade M1000 servers are especially suited for superfast data processing and will be used for artificial intelligence (AI) supercomputing and high-performance computing infrastructure.

 

Image removed. Image removed.


The CSIR delivered decommissioned scientific computing resources to students at the University of the Witwatersrand. These research-grade M1000 servers are especially suited for superfast data processing and will be used for artificial intelligence supercomputing and high-performance computing infrastructure. 

 

Mathekga is a lead mentor within the HPC Ecosystems Project, an initiative of the CSIR-hosted Centre for High Performance Computing (CHPC) that redeploys decommissioned advanced research computing hardware from global big science and big data projects to deserving universities and research institutions in Africa. With a background in aeronautical engineering and computational fluid dynamics, he helps students across the continent set up and use local supercomputing hubs.  

 

“The decommissioned hardware comes from various places,” says Mathekga. “Some are from the Square Kilometre Array project, some are from in-house at the CSIR, and some are from other high-performance institutions across the world.”  (While the SKA has not provided any hardware directly, it has funded the redeployment of the hardware.) 

He says the goal is to democratise access to high-performance computing and AI, so that Africa can become a major player with its own tailored models and solutions.  

"Africa needs high-performance computing because we need to start doing research on African-focused problems,” says Lily de Melo, a student at Wits. She explains that global supercomputer-assisted research is often dominated by Europe and America, so it may not be as relevant to African contexts.  

De Melo is a member of the Wits HPC and Big Data Interest Group, a university club in which students get to practice and compete in global competitions to answer some of the world's biggest questions around climate change, AI, astronomy, biological mechanisms, engineering and more.

In one recent international competition, de Melo applied her supercomputing skills to create a fluid dynamics model of how harmful fumes or smoke from a cook fire could spread inside a home.   

At another competition, fellow Wits student Kevin Ebrahim tackled the protein folding problem, which was considered one of the most difficult computing problems before the AI era, and therefore an excellent supercomputing training ground.  

“The protein folding example was exciting for me because it was the first time I saw my visualisation at one of these competitions,” he says. “It was very eye-opening to see all the work that went into just a simple video clip of the protein folding.” 

HPC Ecosystems Project lead Bryan Johnston, a senior technologist at the CSIR, says Wits is just one hub in a continent-spanning network of partner sites across North, West, East, Central and Southern Africa. These sites bring practical supercomputing capability to those who would otherwise have limited access to this kind of infrastructure.  

“Over its first decade, the project redeployed 35 entry-level HPC systems across 11 African countries, pairing each installation with skills transfer through more than 30 formal training workshops for over 700 participants,” he says.  “The result is a set of locally run HPC hubs, linked by a shared software stack and an active community of practice. All in all, this strengthens regional cyberinfrastructure readiness and enables African teams to participate in data- and compute-intensive research and global HPC competitions.”  strengthens regional cyberinfrastructure readiness and enables African teams to participate in data- and compute-intensive research and global HPC competitions.”  

The referenced media source is missing and needs to be re-embedded.
A screenshot of a map showing the African footprint of the HPC Ecosystem Project. The full interactive map is available here: https://www.google.com/maps/d/viewer?mid=1Hw4DOd4QovlQrMNy5bOl2TXYxiEPA08&femb=1&ll=-8.709176995796714%2C22.469989946017954&z=3   

Johnston says strong student-led programmes like the one at Wits signal that the model is working. “They are an example of how, being provided with access to HPC, they’ve been able to fly and do their own thing. They've now built a pipeline of talent.”  

Reinhard Jansen van Vuuren, a senior member of the Wits HPC group, says his team has once again qualified for a prominent supercomputer student competition in Asia to be held later this year. The team relies on the decommissioned hardware received from the CSIR to power their preparations and to train undergraduates. 

“Often at these competitions, we do weather simulations that are important for climate change, weather prediction and storm prediction, early warnings and so on,” he says. “In these simulations, you're able to predict heatwaves, storms, rain, different pressures in certain areas; and all of these are then combined to provide not only visuals, but also data to predict natural disasters.”  

The Wits HPC group has already published two peer-reviewed papers (available here and here) about how they use the hardware to learn and compete in supercomputing challenges.  

 

Image removed. Image removed.


Members of the Wits HPC and Big Data Interest Group set up supercomputing hardware that has been decommissioned from global big science projects. Groups like these exist across Africa thanks to the HPC Ecosystems Project. The hardware enables students to practice and compete in global competitions to answer some of the world's biggest questions around climate change, artificial intelligence, astronomy, biological mechanisms, engineering and more. From left: Kevin Ebrahim, Lily de Melo, Reinhard Jansen van Vuuren, Samual Mathekga.

 

“The collaboration with the CSIR, the HPC Ecosystems Project and the CHPC is the backbone of our group; it's the reason we can exist and the reason we can do the teaching and the competitions we do,” says van Vuuren. “The resources we have from them are what we use to teach, to train and to learn, and it allowed us to build a little ecosystem within Wits.”  

On a larger scale, Johnston says the HPC Ecosystems Project has evolved into a worldwide community and won international awards, with buy-in from collaborators that help carry African work into global forums.  

With the Wits group currently preparing for their competition in China, Mathekga and Johnston are getting ready for the next hardware deployment in Africa.  

Johnston says they have their sights set on the Democratic Republic of Congo and Angola, after a need for supercomputing infrastructure was identified there. His team is also working on funding proposals to expand supercomputer access and skills further across the continent. 

“I'm really looking forward to seeing Africa's young people tackle complex problems with great solutions and ideas that are generated here at home,” says Mathekga. 

The CHPC is part of the National Integrated Cyber Infrastructure System, a national initiative of the Department of Science, Technology and Innovation that is implemented by the CSIR. The CHPC’s role is to strengthen Africa’s high-performance computing capacity by developing skills and supporting big science. It has redeployed infrastructure to partners in South Africa, including regional partner countries Madagascar, Mauritius, Ghana, Malawi, Lesotho, eSwatini, Botswana, Namibia, Zambia and Kenya. 
South African oil and gas, mining, engineering designs and weather forecasting industries also rely on the CHPC for stable and reliable high-performance computing systems, world-class consultancy in the areas of fluid dynamics, materials science, finite-element analysis, discrete element modelling and general design of high-performance systems.

Published 25 May 2026

Country