Components that are set to replace South Africa's Lengau supercomputer have been waiting to be deployed at the CSIR since last year. They are believed to be worth nearly half a billion rand.
Components of a next-generation high-performance computing system, believed to be valued at almost half a billion rand, have been sitting in storage at the CSIR for a year.
The components will be used to replace the existing Lengau supercomputer at the Centre for High Performance Computing (CHPC).
Lengau is South Africa’s current public supercomputer, which the Minister of Science, Technology and Innovation, Blade Nzimande, described as a “legacy environment.”
Despite it being more than 10 years old, Lengau is still used by South Africa’s scientific and research community and organisations across the continent.
Lengau supported scientists on the continent in performing massive computational operations relatively quickly, including weather and seismic modelling, drug discovery and fluid dynamics.
It represents the most powerful supercomputer in the South African government’s stable, and replacing it with a superior system will greatly benefit the country’s science, research, and technology efforts.
The Department of Science and Innovation, through the CSIR and the CHPC, is set to replace Lengau with a new 4 petaflop system, which will be approximately 300% more powerful.
A ring-fenced budget of R292 million for the 2026/27 financial year has been set for the National Integrated Cyberinfrastructure System (NICIS) to oversee and finance the project.
The CSIR said that this budget will support technologies, equipment, infrastructure, operations, and human capacity, with payments towards the high-performance system forming part of it.
Alongside the allocated budget for 2026/27, Nzimande also revealed a R342 million budget for 2027/28 and R349 million for 2028/29, which were still indicative and could be adjusted at a later date.
Components that will allow the CSIR to replace Lengau with the more powerful system are already in South Africa, but sitting in storage.
A source close to the matter told MyBroadband that the components are valued at nearly R500 million, funded by the South African taxpayer.
“The CSIR confirms that components of the next-generation HPC system were delivered approximately a year ago and are currently being prepared for deployment,” a CSIR spokesperson said.
“Key power and cooling readiness requirements have been completed, and the remaining deployment, migration and decommissioning activities are in progress.”
The Lengau high-performance computing system at the CHPC
The spokesperson said the deployment has taken longer than initially anticipated due to several factors influenced by the complexity of the operation.
This included site readiness, procurement, governance, and infrastructure dependencies, while also ensuring that Lengau remains operational during the transition.
The CSIR does not want to switch off Lengau and risk extended downtime of its national computing services as it sets up the new 4 petaflop system.
It rejected the idea that the new system could not be deployed because it lacked the technical expertise to do so.
“The assertion that the system remains undeployed because CHPC lacks technical expertise is incorrect,” the spokesperson said.
“The CHPC has demonstrated over many years that it has the capability to operate, maintain and support large-scale national HPC systems.”
The CSIR would also not provide a straightforward value for the high-performance system sitting in storage. It said instead that the assertion that it was “more than half a billion rand” was incorrect.
“The replacement HPC system was acquired through a lease-to-own model under which contractual payments are made annually over an agreed period,” it said.
“It was therefore not purchased through a single upfront capital payment of more than R500 million.”
CSIR said that spending on the system was accounted for on relevant NICIS budgets and government processes, as it followed an open procurement process through government tender platforms.
They said that a national-scale HPC system required a longer period for preparation, testing, commissioning and migration, while also ensuring a controlled decommissioning of the previous system.
This included six to 12 months to prepare for site readiness, including power, cooling, and network preparation, and a further two to four months to actually get the system up and into its location.
It could then take another three to six months for workload migration and the decommissioning of the legacy system, the spokesperson said.
The replacement system is expected to become operational before the end of the current financial year, with the decommissioning of Lengau handled in a controlled manner.
