ERSA Says Africa Must Shift Focus Beyond Expanding Network Coverage

A new policy paper from Economic Research Southern Africa makes an argument that deserves a wider African audience than the South African policy community it was written for. Lukasz Grzybowski of the University of Cape Town, in Digital Infrastructure as a Driver of Inclusive Growth (Policy Paper 57, July 2026), concludes that South Africa’s digital infrastructure challenge is no longer primarily one of basic network coverage. The binding constraints have moved: to affordability at the lower end of the income distribution, to effective adoption, to regulatory modernisation, to energy reliability, and to the institutional capacity to implement complex reforms.

That conclusion travels. And it has a direct implication for how the continent should think about the layer of infrastructure that sits above the network – compute, data residency, and the AI capacity that will determine whether connectivity converts into economic value locally or exports that value offshore.

What the ERSA Paper Shows

The evidentiary base is worth taking seriously, because it is unusually strong for this topic.

On the supply side, the South African picture is one of substantial progress. 3G population coverage is essentially universal; 4G exceeds 95 percent on the larger operator networks. Mobile penetration sits at roughly 200 percent of population on a subscriber basis, reflecting multi-SIM ownership. Market concentration has fallen steadily – the Herfindahl–Hirschman Index has dropped from above 0.5 in 2000 to below 0.3 in 2025 – as Telkom Mobile and Rain built share against Vodacom and MTN. Average revenue per user has collapsed from above USD 30 per connection per quarter in the early 2000s to roughly USD 5 today, with all four major operators converging within a narrow band.

On the demand side, the picture is less comfortable. The paper’s most pointed finding concerns the distribution of price reductions. Drawing on ICASA’s published data, no prepaid bundle in the regulator’s basket recorded a price decrease between 2022 and 2025. The 2 GB prepaid bundle actually rose. Over the same period, the cheapest 50 GB post-paid allocation fell from ZAR 907 to ZAR 389, and the 30 GB from ZAR 605 to ZAR 215. The gains went to high-volume users concentrated in higher-income metropolitan segments. The low-volume prepaid users at the bottom of the income distribution – who face the highest effective per-megabyte price – got nothing.

This is not a failure of coverage. It is a failure of the economics one layer up.

Connectivity Still Delivers Economic Benefits

The paper’s literature review is the most useful synthesis of the African connectivity-and-growth evidence currently available in one place. Hjort and Poulsen’s work on submarine cable arrival documents employment growth of roughly 6.9 percent in connected locations relative to comparable unconnected ones, with gains extending to less-educated workers and little evidence of displacement. Bahia and co-authors show 3G coverage in Tanzania raising household consumption and reducing poverty through labour-market channels, with a companion study finding the same pattern in Nigeria. Goldbeck and Lindlacher estimate that early submarine cable arrival raised nighttime-light intensity in mid-sized sub-Saharan towns by around 10 percent.

But the paper is careful – and this is what makes it credible rather than promotional – to include the counterpoint. Azam and colleagues, using Indian data, find no effect at the extensive margin: labour-force participation and employment rates unchanged, with gains only in hours, occupational quality, and household expenditure. The lesson the paper draws is the right one. Connectivity does not mechanically produce employment. Its labour-market effects depend on whether workers and firms have the skills, the electricity, and the institutional support to use it productively.

Anyone building infrastructure in African markets should internalise that qualifier rather than quote around it.

Where the Analysis Stops

Policy Paper 57 treats data centres and cloud as an adjacent layer, explicitly outside its scope. It notes that Teraco operates roughly 189 MW of critical IT load across four campuses, that three hyperscale cloud regions are live in South Africa, and that data centre demand is projected to approach 1,000 MW by 2030. Then it moves on.

That boundary is analytically defensible for a telecommunications policy paper. It is not defensible as a description of where the constraint now sits.

If the diagnosis is correct – that coverage is largely solved and the problem is now effective use – then the question of where the workload runs becomes a first-order economic question rather than a technical detail. A connected user in Lusaka or Kampala whose applications, data, and inference all execute in Frankfurt or Northern Virginia generates local consumption of bandwidth and local payment of licence fees, while the value capture, the employment, and the data itself sit elsewhere. The connectivity is real. The economic transmission mechanism is partial.

The paper’s own framing supports this reading. It emphasises that digital infrastructure operates through firm productivity, labour-market matching, financial inclusion, and trade facilitation. Every one of those channels runs on applications, and applications run on compute. Coverage is a necessary input to all of them and sufficient for none.

Four Lessons for African Policymakers

Distribution Matters More Than National Averages

South Africa’s experience demonstrates that falling average prices can conceal widening inequalities.

The same principle applies to cloud infrastructure. National investments mean little if small businesses, universities and public institutions cannot afford to use them.

Better Targeting Beats Bigger Subsidies

The paper cites research estimating that roughly €902 million of France’s €2.2 billion broadband subsidy programme funded deployments that private companies would likely have completed within three years anyway.

The lesson is straightforward. Public funding works best when it fills genuine market gaps rather than replacing private investment.

That principle is equally relevant for cloud infrastructure, AI facilities and national data centre strategies.

Interoperability Requires Investment

Research cited in the paper found that mobile money interoperability reduced fees but also weakened incentives for tower investment in poorer and rural communities.

Greater openness remains valuable, but regulators must balance access obligations with investment incentives if they want networks to continue expanding.

The same challenge is likely to emerge as countries pursue federated cloud and AI infrastructure.

Energy Remains the Common Constraint

Reliable electricity cuts across every part of digital infrastructure.

Vodacom reported spending roughly ZAR4.7 billion on electricity during FY2024 and invested around ZAR4 billion in batteries and generators between 2020 and 2023.

MTN South Africa operated more than 2,000 backup generators and consumed between 400,000 and 450,000 litres of diesel every month during severe load shedding.

Industry executives have acknowledged that these investments diverted capital away from expanding 5G coverage and rural infrastructure.

Data centres and AI infrastructure consume even more power than mobile networks, making electricity policy inseparable from digital policy.

The Hardest Phase of Digital Infrastructure Is Just Beginning

There is a temptation, for anyone in the sovereign infrastructure business, to read a paper like this as validation. It is more useful as a warning.

Grzybowski’s central point is that the easy work is done and the remaining work is institutional: regulatory frameworks drafted for a different technological era, regulators whose mandates have outgrown their capacity, subsidy programmes that displace the investment they intend to catalyse, and merger remedies whose effectiveness will only be visible after five years of monitoring data. None of that is solved by building more capacity. It is solved by getting the ownership structures, the access terms, the targeting, and the enforcement right – and by accepting that this takes a decade rather than a funding round.

AfriQloud’s position is that sovereign compute and edge AI capacity, deployed with local majority ownership and federated across markets, is the layer that converts connectivity into retained economic value. That thesis is not proven by the ERSA paper. What the paper does establish is the prior condition: the coverage argument for African digital infrastructure has largely been won, and the next set of arguments will be harder, slower, and more institutional than the last.

Which is a reason to start on them now, not a reason to be optimistic about the timeline.

Credit | ERSA Policy Paper 57, “Digital Infrastructure as a Driver of Inclusive Growth” by Lukasz Grzybowski (School of Economics, University of Cape Town), was published in July 2026 as part of ERSA’s Economic Growth Policy Paper Series. All data cited above is drawn from that paper and the sources it references.

Download the free Kaspersky SMB Cybersecurity Guide here to learn how businesses can move beyond traditional antivirus and build a more resilient approach to cybersecurity.