Cassava Technologies moves to build a network of AI factories across Africa, with an investment of up to 720 million dollars targeting five countries and a first facility underway in South Africa. The initial site is specified to host 3,000 Nvidia processors, then scale toward 12,000 units across Nigeria, Kenya, Egypt, and Morocco over the next three to four years, aligning compute with domestic demand rather than export markets.
For Masiyiwa’s group, the facility is framed as critical infrastructure, not a single project, and is stitched into existing fibre, cloud, and data centre assets to deliver AI as a service. Timelines remain staged, but the sponsors are locking scope early to reduce slippage risk and to align with power and connectivity availability in each market. These are substantial claims, and they now carry dated milestones for 2025 delivery in South Africa and phased rollout thereafter.
Financing Signals Scale And Intent
On capital needs, the 720 million dollar figure sits at the top of the range and reflects more than GPUs, since civil works, power, cooling, and interconnect dominate build costs. By June 2025, South Africa is due to receive the first 3,000 GPUs, and management targets 12,000 units continent wide within four years, with unit prices cited at 45,000 to 60,000 dollars that only represent a portion of total outlays.
As CEO Hardy Pemhiwa put it, “If we do not take the first step to deploy our own capital, we cannot expect others to go first,” a statement that signals internal balance sheet commitment before outside project finance. Capacity will initially prioritise researchers and developers, then broaden to enterprise workloads as utilisation stabilises. For procurement, Nvidia’s stack is the reference design, which simplifies vendor risk but concentrates supply chain exposure.
Sovereign AI Shapes Delivery Choices
Across jurisdictions, policy design is steering the architecture, with Cassava positioning the build as sovereign AI capacity that keeps sensitive data within national borders and under local rules. In partnership with Accenture, the company is standing up GPU as a service inside secure, domestic data centres, an approach that aligns with data residency requirements while giving clients managed access to accelerated computing. “Building digital infrastructure for the AI economy is a priority,” said Strive Masiyiwa, underscoring the rationale for a federated network rather than a single hub.
Under the plan, the AI factories ride Cassava’s pan African, high speed fibre network and energy efficient data centres, which reduces latency and improves workload portability across metros. In practice, this configuration should allow ministries, hospitals, and banks to train and serve models locally, then replicate them regionally without exporting personal data.
Policy Watch and Read
Even with reference designs, delivery risk remains concentrated in power availability, import licensing, and GPU supply cadence, especially as global demand tightens. For context, Nvidia is advancing similar AI factory blueprints in Europe, signalling that African sponsors will compete with industrial economies for hardware and systems integration talent, which may stretch schedules if permitting or interconnection lags. Cassava has recently added balance sheet capacity, including a 310 million US dollar package anchored by development finance and a large cloud provider, which supports early phases while long term offtake and project level funding structures mature.
If sponsors can align grid upgrades, cooling, and cross connect capacity, the multi country footprint could anchor research, fintech, and public sector AI workloads within national borders, then scale regionally as utilisation and tariffs stabilise. The prize is clear, but so is the delivery discipline required to convert bookings into productive compute at predictable prices.