Can Africa supply it's own digital infrastructure?

The previous article in this series concluded that most African countries can never satisfy data localisation, resilience and sovereignty at once within their own borders, but that regional trusted zones could solve this problem. This solution faces two major challenges: whether, given the apparent dominance of the hyperscalers (Google, Amazon and Microsoft), sovereign facilities can deliver; and whether countries are committed to making coordination and collaboration work.

This article tests the first challenge and it finds that the sovereign supply side is not an aspiration. It is already in the ground, profitable in places, and operating across the very borders the regional model needs it to span.

Who owns the racks?

Data Landscapers maintains a register of Sub-Saharan Africa’s data centres: 306 verified facilities across 46 countries, republished this month with a corrected ownership classification. Filtering for those that are operational, carrier-neutral (tenants can choose between competing network providers) and offer colocation or hyperscale capacity — excluding captive server rooms, university facilities and edge nodes — leaves 154. For each of these we asked who is ultimately in control, tracing ownership through Mauritius holding companies and London registrations to the people, institutions and states at the top.

The answer does not support the narrative of hyperscaler dominance:

• 103 of the 154 facilities — two thirds — are under African control: owned, ultimately, by African founders, families, listed companies, pension funds or states.
• 31 countries now have at least one operational, African-controlled, carrier-neutral facility.
• US-controlled operators run 33 facilities, about a fifth, dominated by Digital Realty’s estate (Teraco and iColo) — 82% of them in three countries: South Africa, Nigeria and Kenya.
• Not a single facility in the register has a Chinese ultimate owner. China’s exposure runs through a different layer: Huawei and ZTE equipment is present in more than half of all facilities. This is a real dependency, but it is a procurement problem rather than an ownership problem, and it deserves a paper of its own.

One caveat needs stating plainly: facilities are not megawatts. By capacity the US-controlled share is far larger; Teraco’s South African campuses are the deepest on the continent. African-controlled operators run the majority of facilities and by far the widest footprint, but US-controlled operators run the deepest capacity. For regional backup pairing — one good facility in each partner country — breadth matters more than depth. The breadth is African.

How much of the stack is African-owned?

The pessimist’s picture has Africa renting every layer of the stack: the submarine cables, the stations where they come ashore, the buildings, and the cloud services that run inside them. How true is this? Our research identified twelve companies (see the table below) — each at least 40% African-owned and with facilities in more than one country — that contradict that perception. Profiles of these companies can be found in this accompanying compilation.

Submarine cables

• Angola Cables, wholly owned by the Angolan state, owns 33,000km of submarine cable outright — including SACS, the South Atlantic’s only direct link between Africa and the Americas — and ranks 21st in the world among internet interconnection operators, the only African company in the top 100.
• WIOCC, the most African-owned company in the set — its shareholders are ten African telecom operators, from Dalkom Somalia to Zimbabwe’s TelOne — is a core investor in the EASSy cable and the landing party for Meta’s 2Africa at Amanzimtoti.
• Seacom, 75% African-owned, ended the east coast’s dependence on satellite in 2009.
• Safaricom co-owns TEAMS, with its Daraja cable due this year. MTN’s Bayobab is the African partner with the most landing sites on 2Africa, the longest submarine cable ever built.

Landing stations and terrestrial fibre

• CSquared began life as Google’s “Project Link”. In 2023 Convergence Partners, Andile Ngcaba’s investment firm, bought Google out: Google built the network, and African capital now owns it. CSquared runs 7,500km of open-access metro and backbone fibre and has declared an ambition to connect all fifteen ECOWAS states.
• Paratus operates the landing stations for the Equiano and WACS cables at Swakopmund, and a terrestrial network licensed in fifteen countries. Bayobab wholesales fibre across and beyond MTN’s markets.

Data centres

• WIOCC’s Open Access Data Centres is the single largest operator in the register by site count, with a stated 13 core facilities and more than 30 edge sites.
• Paratus runs five centres across Angola, Namibia and Zambia.
• Onix, majority-owned by AIIM under South Africa’s Old Mutual, operates Ghana’s only facility certified at Tier IV (the highest standard of certified resilience) in Accra, and a second Tier IV site at the 2Africa landing in Dakar.
• Wingu anchors the Djibouti–Ethiopia corridor, with Djibouti Telecom as a state joint-venture partner.
• Axian’s Stellarix is building out Madagascar, Senegal and Côte d’Ivoire.

Cloud and services

This is where the stack narrows.

• ST Digital, founded and majority-owned by the Cameroonian entrepreneur Anthony Samé, markets a “100% African cloud” across Francophone West and Central Africa; it is the only company in the set founded, owned and led end-to-end by a single African national.
• Cassava is building the continent’s first serious sovereign AI capacity, as an NVIDIA Cloud Partner with 3,000 GPUs deployed in South Africa since mid-2025 and AI-as-a-service aimed at African governments. Angola Cables and Safaricom both run cloud businesses.

This is a start, not parity — but five years ago the layer barely existed. At every layer of this infrastructure, with the single exception of chip manufacture, there is now at least one credible African-owned operator. That sentence could not have been written in 2020.

Are they where the regional model needs them?

The regional localisation model needs operators already working across borders within each bloc, so that a Kampala ministry’s backup in Nairobi sits with a single accountable provider under regional law. Mapped onto the regional economic communities:

One row deserves special attention. Onix already operates Tier IV facilities in both Accra and Dakar: two carrier-neutral sites in two ECOWAS member states, 2,000km apart, under one sovereign operator. The previous article asked who would host Abidjan’s backup in Dakar. The infrastructure required is not a proposal. It is built, certified and selling capacity today.

How African is “African-owned”?

Before we get too carried away we need to acknowledge that “African-owned” describes a range, not a badge. The twelve companies run from Angola Cables (100% state-owned) and Axian (wholly family-held), through WIOCC (roughly 80–90%) and Paratus (75–85%, anchored by Namibian institutions), to Seacom (75%), down to Cassava (25–40%) and MTN (28–32%, with South Africa’s Public Investment Corporation the largest single shareholder).

And three further distinctions need to be taken into consideration.

Debt is not dilution

Lending from development finance institutions — EAIF behind Paratus’s expansion, RMB behind Wingu’s — finances growth without transferring control. Equity with board seats is what to watch, and there the picture is dominated by African pension funds (the PIC, the GIPF) rather than foreign principals.

Hyperscaler entanglement runs in both directions

Both Google and NVIDIA hold equity in Cassava. Safaricom resells AWS. A hyperscaler is present as a tenant in over half of the African-controlled estate. None of this transfers legal control, though an operator whose anchor tenant is Microsoft is sovereign in a real but bounded sense. Selling to hyperscalers is a revenue line; depending on their jurisdiction is a sovereignty failure.

Jurisdiction is the test, and it must be applied evenly

The CLOUD Act logic of the previous article cuts within this list too. Cassava is headquartered in London; Onix is registered in England; CSquared and WIOCC sit under Mauritius holding companies. None of this creates US legal exposure, but it is a reminder that “African-owned” needs a working definition — ultimate control by African persons, institutions or states, with no US legal nexus — rather than a purity test. By that definition all twelve qualify. By a purity test, almost no global company anywhere would.

Swings and roundabouts

In May 2026 the plan to build a $1 billion Microsoft–G42 cloud region at Olkaria north of Nairobi collapsed into indefinite delay after Microsoft demanded that the Kenyan government guarantee annual payments for capacity, used or not. President Ruto had to do the energy arithmetic: powering the campus would mean switching off half the country. This was not a procurement hiccup. It was the hyperscaler model revealing its terms: builds at a scale African grids cannot feed and African treasuries are now declining to underwrite.

Nine days later in Lagos, Kasi Cloud commissioned LOS1: West Africa’s first AI-capable, carrier-neutral hyperscale campus — 7.2MW now, designed for 100MW, adjacent to six subsea cable landings, and powered by its own gas–solar–battery hybrid rather than a grid guarantee. Its foundational investor is the Nigeria Sovereign Investment Authority (NSIA), the first African sovereign wealth fund to anchor a data centre — a model it is now promoting to its peers. One project needed a sovereign guarantee and froze. The other had sovereign capital and opened. Modular, energy-realistic and African-owned is not a consolation prize; on the evidence of May, it is becoming the working model.

What is still missing?

Four gaps temper the optimism.

Capital

The twelve companies combined deploy in a year roughly what a single hyperscaler spends in weeks. The NSIA’s move shows where the next tranche should come from, but African pension and sovereign money must follow at scale. And there are encouraging signs. In June 2025 the 16 members of the African Sovereign Investors Forum (ASIF) signed a cooperation agreement to establish the ASIF Investment Platform to pool African sovereign wealth resources for large-scale development projects across the continent.

Software

The cloud-platform layer remains the thinnest, and ST Digital and Cassava cannot carry it alone. The big three hyperscalers rarely build their own data centres on the continent, but, as we showed in our work on Google in Africa, they are dominating the rest of the stack.

Compute

Cassava’s GPUs and Kasi’s AI halls are a beginning, not a base. What sovereign AI might look like for Africa remains barely conceived.

Demand

Most importantly, African governments still procure US cloud by default, even where certified sovereign capacity sits, partly empty, in the next country. Nigeria’s National Cloud Policy — mandating domestic residency for regulated data, and guaranteeing Kasi a home market — shows what closing the gap looks like. The supply side has turned up. The demand side is the remaining work.

Where to next?

There are three options already on the table for African governments to embrace.

Firstly, the regional free-flow agreements argued for in the previous article — recognising cross-border transfers within economic regions as compliant with localisation rules — would unlock demand for an industry that already exists.

Secondly, procurement preference: when governments become anchor tenants of sovereign operators within their bloc, as Djibouti does with Wingu and Nigeria with Galaxy Backbone, policy is converted into revenue and revenue into capacity.

Thirdly, redirect the capital: every dollar from the AfDB, the development finance institutions and the sovereign funds that strengthens an African-controlled operator buys sovereignty twice — once as infrastructure, once as jurisdiction.

The previous article concluded that the hardware for regional sovereignty was largely already in the ground. The register and the twelve profiles allow us to say something more precise: we know whose names are on the title deeds, and most of them are African. What remains is for African governments to become customers of their own continent.

The full company profiles, with sources, accompany this article as a single PDF. The underlying data centre register, including the revised ownership classification, is published in the Data Landscapers Lab.