The Fifth Layer: What Will it Take for Africa to Make its own Chips ?
The Layers of the AI Stack, and Why Africa Can Currently Only Participate in Four out of Five
Two weeks ago, at the Global Africa Business Initiative Kenya watch party at the Serena Hotel in Nairobi, I gave a keynote on how Africa can build, own and shape Artificial Intelligence (AI) rather than simply consume it. I borrowed Jensen Huang's picture of AI as a five-layer cake: energy at the bottom, then chips, then infrastructure, then models, and applications at the top.
My argument was that Kenya, and much of Africa, can credibly compete in four of those five layers. The continent has clean, firm power: Kenya's geothermal, the great hydro rivers of the Congo and Nile basins, and solar across the north and south. We have two decades of building subsea cables, cross-border fibre and data centres. And the models and applications at the top are about brains, and Africa's are as good as anyone's.
The fifth layer was chips. Nobody expects Africa to fabricate GPUs any time soon, so I said we should simply buy the best ones available, from West or East. I still believe that is right for this decade. But it left a question in the room, and in my own head: what would it take for Africa to start making chips at all, for our own markets and for the world's?
A tower of small, fragile pieces
A chip is not one industry. It is a stack of dozens of specialised inputs, and many of them come from a single company or a single country.
The best map of this I have seen is The Tsumiki Report from Beebolt. It ranks the 54 supply nodes the semiconductor industry cannot afford to lose, scoring each one on fragility, importance and interdependence. Tsumiki is the Japanese word for wooden stacking blocks, and the image fits: pull the wrong block and the tower falls. My thanks to the Beebolt team for an evidence-based list the industry has needed for a long time.
What struck me reading it was how many of those 54 blocks are not machines or factories. They are minerals, chemicals, gases, power, water, shipping routes and skilled people. Those are exactly the things Africa either already supplies or could.
What Africa already brings
At Digital Economy Advisors we mapped African contributions against the Tsumiki 54. We found more than 60 across 22 countries, touching 24 of the 54 entries. A few stand out:
• Minerals. The DRC mines roughly 70% of the world's cobalt and is a major tantalum source. Rwanda produces tungsten. Zambia and the DRC hold the Copperbelt, and South Africa holds most of the world's platinum-group metals.
• Strategic openings. Guinea is the world's largest bauxite exporter, and gallium is a by-product of refining bauxite. Namibia's Lofdal deposit is rich in the heavy rare earths China restricted last year. South Africa and Kenya have fluorspar, the feedstock for hydrogen fluoride.
• Process inputs. Algeria supplies helium. Morocco's OCP is among the world's largest producers of sulphuric and phosphoric acid.
Footholds in the chip process itself. STMicroelectronics has run an assembly and test plant near Casablanca for decades. Cairo hosts large chip-design and verification teams.
In East Africa, Semiconductor Technologies Limited (STL) has been quietly building since 2018. It runs a nanotechnology and semiconductor facility at the Dedan Kimathi University science park in Nyeri, with around 100 staff, 70% of them women. It is designing IoT chips for livestock monitoring for the Kenyan government. A US Trade and Development Agency grant was announced to conduct a feasibility study for a larger fab processing silicon and silicon carbide wafers.
Nor is this entirely new. South Africa's SAMES ran a semiconductor fab east of Pretoria for more than 25 years. Africa has made chips before. The question is whether we can do it again, at scale, and keep the value.
What it will take
The mistake would be to start at the top, with a leading-edge fab. A single EUV lithography machine costs over $200 million, and the world makes about 50 a year. That is not where Africa enters. We climb the chain from where our strengths already are.
1. Refine at the mine. Exporting raw ore keeps us at the bottom of the value chain. Zimbabwe and Namibia have already banned raw lithium exports to force local processing. The same logic applies to gallium from bauxite, germanium from mine tailings and rare earth separation. Each refinery moves a Tsumiki block closer to home.
2. Start where chips are packaged and tested. Back-end assembly and test needs far less capital than a fab and far more people. Morocco shows it works on this continent. It is also where AI hardware is most constrained right now: the Tsumiki Report names advanced packaging as a critical bottleneck.
3. Choose the right fab. The realistic first fabs are mature-node and specialty plants: power chips in silicon carbide and gallium nitride, sensors, and application-specific chips for agriculture, energy and IoT. These are the chips Africa's own economy needs, and STL's plans point in exactly this direction.
4. Lead with power and water. Fabs run around the clock and use enormous volumes of ultrapure water. The Tsumiki Report lists both power grids and water as critical dependencies. Africa's clean, firm power, from Kenya's geothermal to Ethiopian and Congolese hydro and North African solar, is the same advantage I argued for AI data centres, one layer down.
5. Train the engineers. Skilled fab engineers are one of the 54 dependencies, with fewer than 50,000 advanced fab engineers worldwide. Africa's universities, from Cairo to Kigali to Johannesburg, and programmes like Kenya's AI Skilling Alliance show we can train at scale once we decide to. Semiconductors need the same deliberate effort, and it has started. ChipMango, founded in 2022 by Nigerian-born engineer Ola Fadiran and Jovan Andjelich, trains African engineers in chip design and then puts them to work on commercial design and verification projects for global customers. It is an Arm Approved Training Partner, works with Nigerian universities including the University of Lagos, Obafemi Awolowo University and Miva Open University, and raised a $1.9 million seed round in September.
6. Bring patient capital and anchor buyers. Fabs take years to pay back. They need development finance, partners from West and East, and governments may not be natural offtakers but need to create incentives. In Kenya, IoT hardware designer and maker Geviton and smartphone assembler EADAK are natural first buyers for African-made chips.
7. Build regionally. No African country will own the whole chain alone. The African Continental Free Trade Area lets minerals from one country, refining in another and assembly in a third add up to one industry.
That continental coordination has now begun. On 30 September, AUDA-NEPAD launched the Africa Semiconductor Technical Advisory Group (ASTAG) in South Africa, with the CSIR, the African Academy of Sciences, South Africa's Department of Science, Technology and Innovation, and Nina Jojer Africa. Its aim is to turn fragmented national efforts into one continental ecosystem.
Over the next 12 months, ASTAG will work on almost exactly the agenda above: critical minerals, skills, packaging and testing, chip design and RISC-V, policy, investment, and applications in energy, health, agriculture, mobility and fintech. It will produce an Africa Semiconductor Roadmap for 2026–2030. AUDA-NEPAD puts Africa's share of the world's known critical mineral reserves at around 30%. Its stated ambition, to go from minerals to microchips, is the right one.
Three questions for the fifth layer
In my keynote I offered three questions to test any AI spend. Here is the version for chips. Was the mineral refined in Africa? Was the chip designed, packaged or tested here? Is an African firm part of the value chain? Today, almost every answer is no.
An African chip industry will take some time to realise, perhaps a decade or more. That is the reason to start now. Begin with refineries, assembly lines, specialty fabs, trained engineers and anchor contracts, and build on those who have already begun, from STL in Nyeri to the assembly lines in Casablanca and the design teams in Cairo.
If you are an investor, a policymaker or a company looking at this opportunity, and most of all share this vision for African AI "Full Stack" Made and Operated in Africa Digital Economy Advisors would be glad to talk. Our map of where Africa fits into the Tsumiki 54 is available for download, and we look forward to contributing to ASTAG's roadmap.
Download our Africa Semiconductor Value Chain Report Here
Comments