South Africa Visualization of Digital Infrastructure Priorities

Geography

The concentration, mapped

Interact: pan · zoom · pin markers · toggle layers.
Overlays: dam storage · firm grid headroom: the two siting constraints.

Marker area is proportional to facility count; landing stations are sized by systems landed. Overlay data. Water: Department of Water and Sanitation provincial dam levels, 29 June 2026; electricity: Eskom grid-connection assessment 2025 (January 2024 addendum).

What the map argues

Johannesburg is the hub: 36 of 63 data centres · 5 of 12 exchanges · 7 of 8 cloud. Inland, coast-dependent.
Nine of eleven systems are single-coast. Only SAFE and 2Africa land on both.
Six landlocked SADC states route mainly through South Africa: Botswana, Zimbabwe, Lesotho, Eswatini, Zambia, Malawi. Alternatives run via Namibia and Mozambique (ITU broadband maps); see the comparison below.

South Africa vs selected Southern African states, July 2026

Country	Subsea systems	Landing points	Data centres	Internet exchanges	Cloud data centres
South Africa	11	7	63	12	8
Angola	5	3 (+4)*	10	2	0
Mozambique	4	3	6	1	0
Namibia	2	1	2	1	0
Botswana	0	0	5	1	0
Zimbabwe	0	0	3	0	0
Zambia	0	0	3	0	0
Lesotho	0	0	3	1	0
Malawi	0	0	1	1	0
Eswatini	0	0	1	0	0

*Angola: 3 shared landing points, plus 4 used solely by the domestic Unitel North Submarine Cable. All nine neighbours combined: 34 data centres (fewer than Johannesburg alone, 36), 7 exchanges, and not one cloud facility. Counts from TeleGeography Submarine Cable Map, Data Center Map, PeeringDB and the Cloud Infrastructure Map.

Risks

Concentration amplifies three risks

Geographic concentration

─▶ bigger blast radius per fault

─▶ load lands on the driest, tightest provinces

Cable security

Energy & water

Geographic concentration

Johannesburg36

Cape Town13

Durban7

Pretoria4

3 others3

78% of data centres in two metros; Gauteng alone 63%
Every cable landing sits in one of two coastal clusters, plus one at Gqeberha
South Africa: 40.8% of Africa’s data-centre market, the continent’s largest (Mordor Intelligence, 2025)

Cable security

KwaZulu-Natal7

Western Cape6

Eastern Cape1

9 of 11 systems single-coast: only SAFE and 2Africa land on both
Mar 2024: suspected undersea landslide degraded 13 countries
May 2024: suspected anchor drag off KwaZulu-Natal cut SEACOM + EASSy for weeks
ITU: most faults worldwide are routine maritime activity

Energy & water

Fleet availability69%

Dams (national)96%

Dams (Western Cape)76%

~30% of the generation fleet unavailable (Dec 2025); load shedding returns past 13 000 MW of winter breakdowns (Eskom 2026 winter outlook)
0 MW firm grid headroom in Eastern, Northern & Western Cape (GCCA 2025)
Dams full today, but Western Cape lowest; long-term drying projected (World Bank)

Dependency

Who owns the stack

Ownership by infrastructure layer

Internet exchanges · 12 of 12 Domestic governance, foreign premises

INX-ZA · 6

NAPAfrica · 5

? · 1

Cloud data centres · 8 of 8 Three of eight are Digital Realty

MTN | Vodacom

ADC | OADC

Digital Realty · 3

? · 1

Subsea cables · 11 operational Planned: Umoja (Google) · Waterworth (Meta), both single-owner

South African participant · 6

No South African participant · 4

Google · 1

+ 2 planned

South African operator Other African or multi-party Single foreign owner Planned (under construction) ? (ownership unspecified)

NAPAfrica, Africa's largest exchange, runs entirely inside Teraco, majority-owned by Digital Realty, a US real-estate trust, since 2022 (a deal valuing Teraco near $3.5 billion)

Cloud provider presence × facility owner: the geopolitical seam

Provider \ Facility owner	Digital Realty (US-owned)	African operators (ADC \| OADC)	South African telcos (MTN \| Vodacom)
AWS	3	—	—
Microsoft Azure	3	—	—
Google Cloud	3	1	—
Oracle	1	—	—
Huawei Cloud*	2	1	2

AWS, Azure and Oracle presence is listed only in Digital Realty facilities; Google 3 of 4 sites
Huawei Cloud is the only provider on South African-operated floors (MTN, Vodacom); 4 of its 6 sites sit outside Digital Realty. *Its sixth Johannesburg site has no publicly specified operator
The structural risk: US cloud presence sits in US-owned facilities; Huawei’s spans South African telcos. A US-China shock would split this market along its ownership seam, with South Africa on both sides.

Policy mitigation

Four policies, one network

Distribute the infrastructure

Siting incentives (tax credits, expedited permitting) so new data centres and landings break the two-metro pattern.

Candidate regions:

Metro	Existing base	Grid headroom	Dam storage
Durban (KwaZulu-Natal)	7 data centres · 2 exchanges · 3 cable systems	5 500 MW	89%
Bloemfontein (Free State)	1 data centre · backbone midpoint	1 420 MW	100%
Gqeberha - East London (Eastern Cape)	2 data centres · exchange · landing	790 MW, curtailment-only	89%

Trade-off: Redundancy over metro efficiency: distributing new builds splits Johannesburg’s peering density.

Southern Africa resilience pact

Regional coordination pact: new landing stations beyond South Africa, shared monitoring, mutual-aid routing.

Routes for six landlocked states

Today→ South Africa: Western Cape cluster South Africa: KwaZulu-Natal cluster

In the future→ South Africa × 2 clusters + expand Swakopmund (Namibia) + expand Maputo (Mozambique) + mutual-aid routing

Trade-off: South Africa gives up hub rents in exchange for shared regional resilience.

Data centres that give back

Approval tied to contribution: contracted renewables, wheeling, water disclosure.

The deal per new hyperscale facility

New DC approved⇄ adds clean megawatts (purchase agreement + wheeling) publishes water intensity siting weighted by province stress

Precedent (2024)→ Africa Data Centres 12 MW Free State solar → wheeled to its Cape Town facility (2024)

Trade-off: Approvals become slower than rival markets, the price of grid and water headroom.

Protect & respond

Harmonised cable protection zones + a stationed regional repair vessel.

faster repair: weeks → days fault report < 2 h zones gazetted in every SADC coastal state mariner awareness program

Trade-off: A levy on landing operators, buying repair in days instead of weeks.

Coverage & connections

What covers what, and how the policies connect

Risk \ Policy	Distribute	Regional pact	Data centres give back	Protect & respond
Geographic concentration	Strong	Strong	Partial	None
Cable security	Partial	Strong	None	Strong
Energy & water	Partial	None	Strong	None

Strong: acts on the mechanism · Partial: contributing factor · None: no effect. Author-judged from the risk and dependency evidence, not computed.

Sources