On 17 November 2025, Meta announced it had completed the core of 2Africa, a 45,000-kilometre subsea cable, the longest ever built, wrapping the continent from the UK down Africa's west coast to South Africa and back up its east coast to Djibouti and the Gulf. It carries 180 terabits per second across sixteen fibre pairs, enough capacity to serve the internet needs of more than 3 billion people. Thirty-three countries are wired into it, most of them African.
Not one of those African countries owns a strand of it. The cable is financed and controlled by a consortium of eight companies, Meta, China Mobile International, Vodafone, Orange, stc's center3, Telecom Egypt, WIOCC, and Bayobab, the renamed international arm of South Africa's MTN. Only two names on that list are African at all, and both are carriers leasing capacity on infrastructure whose capital, engineering, and routing decisions sit in Menlo Park, Beijing, London, and Amman.
2Africa is not an isolated case. It is the physical expression of a pattern running through every layer of the continent's digital stack: the cables data travels through, the satellites now relaying it, the towers broadcasting it, and the data centres increasingly storing and processing it. Look at each layer and the same fact repeats. Africa generates the demand. Capital, ownership, and operational control sit somewhere else.
Terrestrially, Huawei built an estimated 70% of Africa's 4G network, with ZTE responsible for a further 20–30% of the continent's earlier 3G rollout; the two Chinese vendors have deployed more than thirty national mobile networks continent-wide, and Huawei is expected to remain the primary builder as 5G arrives. In orbit, SpaceX's Starlink has gone from novelty to live service in 27 of Africa's 54 countries by early 2026, adding Liberia, Niger, Somalia, Guinea-Bissau, the DRC, Lesotho, Chad and Senegal within roughly a year.
In the cloud layer, Google switched on a Johannesburg region in 2025 after a $148 million build, Microsoft is spending roughly $300 million to expand South African cloud and AI capacity by 2027 on top of a separate $1 billion geothermal-powered data centre under construction in Kenya, and AWS has committed a further ZAR 30.4 billion, about $1.7 billion, to South Africa through 2029. Read those numbers against the denominator most coverage skips: the continent hosts under 1% of the world's data centre capacity despite holding roughly 18% of the world's population. The $2–3 billion in hyperscaler commitments above is not competing with African capital. It is filling a vacuum where none exists.
“The $2–3 billion in hyperscaler commitments is not competing with African capital. It is filling a vacuum where none exists.”
This is not a 2025 story; treating it as one understates it. Africa's teledensity was built in two waves of foreign capital, twenty years apart. The first, in the 2000s, was Eruopean and South African mobile operators, MTN, Vodacom, Orange, what became Airte, laying GSM networks under license from newly liberalising governments that had neither the capital nor, at the time, a policy framework demanding local ownership of infrastructure, only of spectrum. The second wave, from roughly 2010, was Chinese state-linked capital: China Exim Bank-financed loans built the towers Huawei and ZTE equipment now runs, bundled as concessional financing tied to Chinese contractors and equipment, a model criticised for opacity but rarely refused, because the alternative was no network at all. The current wave, American hyperscalers and satellite operators, is the same bargain restated for the cloud era. Three waves, one constant: African governments have negotiated the terms of access to infrastructure they do not own, not the infrastructure itself.
What is shifting now is not the ownership pattern but the leverage some African states are starting to exert against it. South Africa's fight with Starlink is the clearest case. Pretoria's Electronic Communications Act requires 30% ownership by historically disadvantaged South Africans before any network operator gets a licence, and Starlink, structurally unable to sell equity in a foreign-owned satellite constellation, has been shut out since 2023. Communications Minister Solly Malatsi's proposed workaround, letting Starlink satisfy the requirement through "Equity Equivalent Investment Programmes" such as rural connectivity donations rather than actual shareholding, has been read by the Communications Portfolio Committee as an attempt to erode transformation goals by stealth. The dispute, still unresolved in mid-2026, is a genuine test of whether a mid-sized African economy can set terms for a trillion-dollar satellite operator, or whether the operator sets terms for it.
The African Union is attempting the continental version of the same fight. Its Data Policy Framework, now in a 2025–2030 implementation phase with validation workshops running through December 2025, aims to harmonise data governance across all fifty-five member states into something resembling a single digital market by 2030, explicitly framed as a route to collective digital sovereignty. The AU's parallel AI Strategy, adopted in 2024, projects AI could add $1.5 trillion to African economies by 2030, but that figure is aspirational against a base of near-zero domestic compute. Without African-owned cloud and chip capacity, the AI layer risks becoming a fourth wave of the same pattern: American and Chinese models, run on American and Chinese infrastructure, trained partly on African data.
The stakes split unevenly. Consumers and small businesses win in the near term, 2Africa and Starlink both cut connectivity costs and reach places terrestrial networks never will, and hyperscaler regions cut latency for African users of everyday cloud services. Foreign operators win durably: they capture the margin on infrastructure, the data flowing across it, and the negotiating leverage that comes with owning something a government cannot easily replace or nationalise. African governments and domestic operators are the exposed party, dependent on infrastructure they cannot unilaterally alter, taxed on transactions across networks they do not control, and negotiating from a position where refusing a hyperscaler or satellite operator now risks losing the investment entirely, not merely delaying it.
Three things are worth watching. First, whether South Africa's Starlink resolution, expected sometime in 2026, sets a template other regulators copy or reject; a settlement through equity-equivalent programmes would likely become the continental default, since few governments carry South Africa's negotiating weight. Second, whether the AU Data Policy Framework produces anything with enforcement teeth by 2030, or becomes another aspiration individual states quietly ignore when a hyperscaler investment is on the table. Third, and least visible, whether any African state or regional bloc begins building sovereign compute or cloud capacity rather than only regulating access to foreign capacity, early data-centre incentive moves in Kenya and Rwanda are worth tracking as the first tests of that question.
For founders and investors, the sharper question is not whether foreign infrastructure will keep arriving. It will, because the capital gap is real and the alternative is no infrastructure at all. The question is where the defensible business sits on top of infrastructure you will never own, and nowhere in the stack is more contested, or more available, than the application and data layer riding on cables, towers, and clouds someone else built. The infrastructure question was answered decades ago, by default, because no one else was building. What is still open is whether the next decade is spent negotiating better terms of access to that infrastructure, or the one after that finally becoming an owner of it.
