A project starting at 20MW and targeting 200MW shows that the next phase of Africa’s AI economy will be determined partly by where computing power and data are physically located.
Egypt’s latest digital-infrastructure project makes a useful distinction between using artificial intelligence and owning the infrastructure on which artificial intelligence runs. Vodafone Business, Elsewedy Electric and Cassava Technologies have announced partnerships to develop a major data-centre platform and sovereign AI infrastructure in Egypt, beginning with 20 megawatts of capacity and targeting 200MW over time.
Egypt’s State Information Service said the project is expected to attract about $200 million in foreign direct investment initially, rising to $1 billion at full build-out. The initiative includes Africa Data Centres Egypt and a sovereign AI centre using Nvidia technologies through Cassava Technologies. The government expects the programme to support local data hosting, cloud services and large-scale AI applications.
The phrase “sovereign AI” is the key. Most organisations can access AI models through global cloud platforms without owning any physical computing infrastructure. That is efficient, but governments and regulated industries increasingly care about where data is stored, who controls the compute environment and which legal jurisdiction applies. Sovereign AI infrastructure aims to provide high-performance computing inside the country so sensitive or strategically important workloads can be processed locally.
The mechanism is compute capacity. Modern AI requires specialised processors, large amounts of electricity, cooling, network connectivity and secure data-centre space. A country may have talented software developers but still depend entirely on foreign infrastructure to train or run advanced models. Data centres therefore become part of the industrial base of the digital economy.
Egypt has several advantages in this market. It has a large domestic population, a sizeable technology sector, extensive subsea cable connectivity through the Red Sea and Mediterranean, and a geographic position connecting Africa, the Middle East and Europe. Those factors can support both domestic demand and international data-centre services if power, regulatory and cost conditions remain competitive.
The planned scale is material. A move from 20MW to 200MW changes the project from a national facility into infrastructure capable of serving large enterprise and AI workloads. Data-centre economics are heavily influenced by power availability and utilisation. Operators need enough customers to fill expensive capacity while maintaining redundancy and service standards.
Energy therefore sits underneath the AI story. High-density GPU infrastructure consumes substantial electricity and produces substantial heat. Egypt’s ability to expand data centres will depend on dependable generation, grid connections and efficient cooling. The country’s broader energy strategy is thus directly linked to its ambition to become a regional AI hub.
The partnership structure also matters. Elsewedy Electric brings infrastructure and engineering capability, Vodafone Business brings enterprise customers and network services, while Cassava Technologies brings data-centre and AI infrastructure capability. Large digital projects increasingly require this type of stack: land, power, connectivity, compute, cloud services and customers organised together.
For African businesses, local sovereign compute can change procurement decisions. Banks, governments, healthcare institutions and large companies may be more willing to adopt AI services when data can remain within an approved jurisdiction. Latency can also improve for local workloads. Most importantly, local infrastructure creates a market for engineers, cybersecurity specialists, cloud architects, maintenance providers and software companies that build on top of the compute layer.
There is a risk of overbuilding. Data-centre announcements can be impressive, but capacity only creates value when customers use it. Global competition is intense, technology changes rapidly and AI hardware can become obsolete faster than traditional infrastructure. Operators therefore need credible demand, not only national ambition.
Skills are another constraint. A sovereign AI centre can host advanced GPUs, but productive use requires organisations capable of building and operating AI systems. Training programmes, university links and enterprise adoption must grow alongside physical infrastructure. Otherwise, the country risks owning expensive compute that is primarily consumed by external customers.
For policymakers, data sovereignty should also be implemented carefully. Local hosting can improve control, but restrictive rules can raise costs or isolate domestic firms from global cloud ecosystems. The strongest model is likely to combine sovereign capacity for sensitive workloads with open access to international services where appropriate.
The project could also reshape the economics of regional cloud services. African firms often host workloads in Europe, the Middle East or South Africa because local high-performance infrastructure is limited. A large Egyptian platform can compete for some of that demand, particularly from North Africa and the eastern Mediterranean. Lower latency, local support and jurisdictional certainty can become commercial advantages. But regional customers will still compare price, uptime and interoperability with global hyperscalers. Egypt therefore has to compete as an infrastructure market, not simply as a sovereign project. Scale will matter only if service quality and economics are strong enough to attract sustained external demand.
The decisive point is that Africa’s AI competition is moving below the application layer. Models and software attract attention, but the strategic assets are increasingly power, fibre, data centres and specialised compute. Egypt’s proposed 200MW platform is a bet that owning more of that infrastructure will translate into digital sovereignty, investment and regional influence. The return will depend on whether the country can turn megawatts of computing capacity into productive economic use.
