South Africa

AI-specific country records and supporting context for South Africa, kept separate. 3 of three AI pillars have a country record.

4 AI-specific indicators; 6 of six context measures reported.

Explore this data →

4AI-specific indicators held

3 of 3AI pillars with a country record

6 of 6context measures reported

30signals recorded

Jobs

Share of tasks economically exposed to AI

51.0 % of tasks

modelled Academic research · Tier 2 · 2026

Garg, Crosta and Baier, Global Automation Atlas (2026)

Tasks rated 2 or 3 on a 0-3 exposure scale, relabelled under each country's conditions. Not employment-weighted and not a share of jobs.

Recompute from data/jobs-atlas-exposure.csv.

Access

Illustrative AI workload as a share of a monthly minimum wage

55.8 % of a monthly minimum wage

modelled Derived from academic and official inputs · Tier 2 · 2026-03-01

Computed from Claude Opus 5 (Anthropic) rate card and the statutory minimum wage

A month of moderate use = 20 standard workloads at the median frontier rate card ($11.25 per workload), divided by the monthly statutory minimum wage.

Recompute from data/affordability.csv.

Illustrative AI-use basket as a share of monthly GNI per capita

43.1 % of monthly GNI per capita

modelled Derived from reported pricing and official economic data · Tier 4 · 2025

Computed from Claude Opus 5 rate card and World Bank GNI per-capita data

20 standard workloads at the median frontier rate card ($11.25 per workload), divided by monthly GNI per capita on the Atlas method (annual value divided by 12). Published only when the latest non-empty GNI observation is within the five-calendar-year freshness window.

Recompute from data/affordability.csv.

Official provider location listings

This records a country appearing on a provider's published geographic support list. It does not test signup, payment, every plan or feature, or service uptime. API and consumer products are separate.

Provider	Product	Published status	Evidence
OpenAI	ChatGPT web and mobile	Listed as supported	Official list
OpenAI	OpenAI API	Listed as supported	Official list
Anthropic	Claude.ai web and mobile	Listed as supported	Official list
Anthropic	Anthropic commercial API	Listed as supported	Official list
Google	Gemini web app	Listed as supported	Official list
Google	Google AI Studio and Gemini API	Listed as supported	Official list

Download the full source register.

Signup and payment observations in the 10-country sample

These records describe the stated signup, card and payment evidence for sampled countries. They are separate from official geographic support lists and do not calculate affordability.

Provider	Signup status	Evidence
OpenAI	yes	Source
Anthropic	yes	Source
Google	yes	Source
xAI	unclear	Source
Perplexity	unclear	Source
DeepSeek	unclear	Source

Infrastructure

Data centre capacity named in the register

40.0 MW named (all delivery statuses)

announced News reporting · Tier 4 · 2026

Observatory infrastructure register

Sum of published capacity in the register. Projects without a published capacity figure are excluded, never counted as zero. This inventory total includes all labelled delivery statuses; it is not operational supply. Use the status breakdown for operating capacity.

Recompute from data/infrastructure.json.

Named data-centre projects in the register

MTN AI-ready data centres — announced; 150.0 MW across multiple countries; national split not reported; 2026-08-26. Source
Cape Town hyperscale campus (two sites) — appeal_pending; No capacity figure published; 2026-07-14. Source
Equinix Cape Town campus — approved; No capacity figure published; 2026-08-04. Source
Teraco JB7 data centre expansion — under_construction; 40.0 MW stated; 2026. Source
Cassava AI Factory — operational; No capacity figure published; 2026-08-26. Source

Country context

Connectivity and power conditions help explain the environment for digital services. These are not AI-use, AI-adoption or data-centre-capacity measures. Each value keeps its source year.

Connectivity

Measure	Value / status	Year	Source
Individuals using the Internet (% of population) IT.NET.USER.ZS	78.4 % of population	2024	World Bank API
Mobile cellular subscriptions (per 100 people) IT.CEL.SETS.P2	179.3 subscriptions per 100 people	2024	World Bank API
Fixed broadband subscriptions (per 100 people) IT.NET.BBND.P2	5.3 subscriptions per 100 people	2024	World Bank API

Power and reliability

Measure	Value / status	Year	Source
Access to electricity (% of population) EG.ELC.ACCS.ZS	90.2 % of population	2024	World Bank API
Firms experiencing electrical outages (% of firms) IC.ELC.OUTG.ZS	92.0 % of firms	2020	World Bank API
Value lost due to electrical outages (% of sales for affected firms) IC.FRM.OUTG.ZS	0.0 % of sales among firms affected	2020	World Bank API

Download all country context values and source links.

Signals

Signals whose headline names this country, newest first. Tags say what kind of source each one is.

AI skills now among South Africa's hardest to recruit as employers cool on hiring abroadSkills and talentiAfrica · Sep 30
Outsourcing body launches AI-focused job board for remote South African talentSkills and talentOutsource Accelerator · Sep 25
South Africans voice growing concerns over AI's rapid riseEmployment and outsourcingAfricanews · Sep 16
South Africa's AI Paradox: Best Prepared, Most ExposedEmployment and outsourcingIOL · Sep 10
AI is making South Africa's call centers pricierEmployment and outsourcingOutsource Accelerator · Sep 4
South Africa's AI Readiness Puts It in a League of Its OwnEmployment and outsourcingIOL · Aug 27
South Africa's Unemployment Rate Climbs to 33.6% in Q2 2026Employment and outsourcingRio Times · Aug 20
AI Job Squeeze Threatens 3.3 Million South African WorkersInvestment and fundingRio Times · Jul 29
South Africa's gig economy becomes a bargain with AI skills shortage loomingSkills and talentWeeTracker · Mar 31
Meta’s smart glasses are officially coming to South AfricaAccess and affordabilityTechCentral · Oct 2

About this page

Generated, not written

This profile exists for every African country in InferenceAfrica's geographic scope. The AI-specific indicator count excludes the context table. “No information” means no country-level value or record was available in the named dataset; it does not mean zero or prove that no information exists elsewhere.

Where the numbers come from

AI-specific figures come from the indicator registry and pillar registers. The separate context snapshot uses World Bank API indicators, with the originating ITU, ESMAP and Enterprise Surveys sources named in the downloadable file. Every context row links to its country-specific API query. Product geography is separately transcribed from the cited providers' supported-region lists, retrieved 25 September 2026. Read the matching method and limitations.

A country in one page

Download a country snapshot

Six selected measures, with dates and sources. Use the full country profile for the complete record.

Country

Portrait · 4:5 format
PNG: 2160 × 2700 pixels
SVG: editable vector artwork

Download SVG Download data CSV

Open full country profile

The export uses a fixed light background for readable sharing. Unknown values stay labelled; no country score is calculated.

AI and the economy

Work and economic conditions

annual % · 2025
reported

-1.16 percentage points from 2024 to 2025.

Explore annual values · Source

Definition and limits

Annual percentage change in consumer prices; national basket definitions differ. Economic context helps interpret AI activity. These figures do not measure AI’s effect on employment or growth. Observation years vary; ILO labour estimates are modelled. Missing observations remain missing. Retrieved 2026-10-06. Latest selection is limited to 2019–2025.

Download annual economic context. GDP per capita is not personal income. Low unemployment can coexist with informal or low-paid work.

Infrastructure · Facility directory

Where are the data centres?

Source-reported directory listings, not a national census. Includes commercial, government, enterprise and research facilities. Source claims have not been independently verified by InferenceAfrica. Campuses may contain multiple facilities. Do not add these counts or capacities to the project register.

Adapted from Data Landscapers, edition 2026-10-05-2, CC BY 4.0. Fields selected and counts calculated by InferenceAfrica. Download the full directory · Chart country counts.

Collection round · 2026-10-01

Evidence on use, work, costs and compute

This collection preserves each source’s population, period and definition. Estimates, survey responses, vacancies and provider claims remain separate. Coverage is incomplete.

AI use and adoption

Survey percentages describe respondents; telemetry estimates use a different denominator. The five-country employee sample has no country-level breakdown here.

AI use reported by online survey respondents — 1/1 records with information

Any AI tool or application, Q3. Online survey respondents aged 18+; not all residents, and not restricted to work use. Country sample approximately 1,000.

2025 fieldwork; use in preceding 12 months · survey · measured · Google / Ipsos Our Life With AI, 2025 fieldwork

Online non-probability survey, approximately 1,000 adults per country; fieldwork 22 September–10 October 2025. Self-reported use. Results describe the surveyed sample and must not be presented as a census of all residents.

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	72 % of respondents	Any AI tool or application, Q3. Online survey respondents aged 18+; not all residents, and not restricted to work use. Country sample approximately 1,000.

Generative AI use reported by online survey respondents — 1/1 records with information

Generative AI tools or generation of images, video or audio, Q4. Same online survey and fieldwork as the all-AI measure; a different question.

2025 fieldwork; use in preceding 12 months · survey · measured · Google / Ipsos Our Life With AI, 2025 fieldwork

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	70 % of respondents	Generative AI tools or generation of images, video or audio, Q4. Same online survey and fieldwork as the all-AI measure; a different question.

Estimated generative AI use — H1 2025 — 1/1 records with information

Generative-AI diffusion estimate adjusted from Microsoft telemetry for device/OS market share, internet penetration and population. Working-age denominator; not a direct population survey or a measure of workplace adoption. Appendix pp. 15–17.

H1 2025 · vendor_estimate · modelled · Microsoft AI Economy Institute, Global AI Adoption in 2025

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	19.3 % of working-age population	Generative-AI diffusion estimate adjusted from Microsoft telemetry for device/OS market share, internet penetration and population. Working-age denominator; not a direct population survey or a measure of workplace adoption. Appendix pp. 15–17.

Estimated generative AI use — H2 2025 — 1/1 records with information

H2 2025 · vendor_estimate · modelled · Microsoft AI Economy Institute, Global AI Adoption in 2025

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	21.1 % of working-age population	Generative-AI diffusion estimate adjusted from Microsoft telemetry for device/OS market share, internet penetration and population. Working-age denominator; not a direct population survey or a measure of workplace adoption. Appendix pp. 15–17.

Employment outcomes and hiring demand

Vacancy advertisements indicate hiring demand. Comparable national wage, hiring and net-employment effects remain a research gap.

Growth in AI-related vacancies on Pnet — 1/1 records with information

H1 2025 versus H1 2024 · platform_records · measured · Pnet Job Market Trends, August 2025

Change in AI-related vacancies observed by Pnet. A vacancy is an advertisement, not a hire or net job creation. The report includes international and remote listings; it is not a census of domestic employment.

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	77 % change	Pnet report, page 7: year-on-year growth in AI-related advertisements. Platform scope includes international/remote listings. This is a demand signal; it does not measure hires, net employment, wages or a causal effect of AI.

A career-guidance experiment was announced. The reviewed announcement supplies no completed employment-impact result. Research source.

Practical access costs

These are ITU’s 2025 observations, including the new 5 GB mobile basket. Local device prices and verified payment conditions remain to be collected.

Mobile data basket (5 GB minimum) — 1/1 records with information

ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. The 5 GB mobile basket must not be spliced into the older 2 GB series. Connection charges and device costs are not added to this monthly figure.

2025 · official · measured · ITU ICT Price Baskets, December 2025 release

Retail basket prices collected under ITU rules; country prices are published in USD and as a share of monthly GNI per capita. Basket definitions changed in 2025.

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	6.43 USD per month	ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. The 5 GB mobile basket must not be spliced into the older 2 GB series. Connection charges and device costs are not added to this monthly figure. Selected operator: Vodacom SA. Plan allowance: 5 GB. Tax rate as published: 15.

Mobile data basket / monthly income — 1/1 records with information

ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. The 5 GB mobile basket must not be spliced into the older 2 GB series. Income ratio uses ITU’s own GNI denominator and release, which can differ from this Observatory’s AI-price income base.

2025 · official · measured · ITU ICT Price Baskets, December 2025 release

Retail basket prices collected under ITU rules; country prices are published in USD and as a share of monthly GNI per capita. Basket definitions changed in 2025.

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	1.27 % of monthly GNI per capita	ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. The 5 GB mobile basket must not be spliced into the older 2 GB series. Income ratio uses ITU’s own GNI denominator and release, which can differ from this Observatory’s AI-price income base. Selected operator: Vodacom SA. Plan allowance: 5 GB. Tax rate as published: 15.

Fixed broadband basket (5 GB minimum) — 1/1 records with information

ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. Connection charges and device costs are not added to this monthly figure.

2025 · official · measured · ITU ICT Price Baskets, December 2025 release

Retail basket prices collected under ITU rules; country prices are published in USD and as a share of monthly GNI per capita. Basket definitions changed in 2025.

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	17.03 USD per month	ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. Connection charges and device costs are not added to this monthly figure. Selected operator: Telkom. Plan allowance: Unlimited GB. Tax rate as published: 15.

Fixed broadband basket / monthly income — 1/1 records with information

ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. Income ratio uses ITU’s own GNI denominator and release, which can differ from this Observatory’s AI-price income base.

2025 · official · measured · ITU ICT Price Baskets, December 2025 release

Retail basket prices collected under ITU rules; country prices are published in USD and as a share of monthly GNI per capita. Basket definitions changed in 2025.

Make a chart from this measure

Scope	Value	Meaning and limits
South Africa	3.36 % of monthly GNI per capita	ITU 2025 retail basket, selected under its representative-operator rules. At least 5 GB; actual allowance may be higher. Historical observation, not a current quote. Income ratio uses ITU’s own GNI denominator and release, which can differ from this Observatory’s AI-price income base. Selected operator: Telkom. Plan allowance: Unlimited GB. Tax rate as published: 15.

Operating facilities and compute services

A starter review of operating assets and services. The Infrastructure project pipeline tracks a different set of announcements; this is not a national or continental capacity total. Rated MW, GPU inventory and actual load are different measures.

Critical IT capacity at reviewed completed sites — 1/1 records with information

Selected site evidence, not a national inventory. Rated critical IT capacity is not actual load, GPU capacity or an Africa total.

Operator completion statement, 2025-08-12 · operator_statement · measured · Teraco JB4 expansion completion

Operator reports completion of a 30 MW addition, bringing the JB4 campus to 50 MW critical IT capacity. Total and increment must not be added. Rated capacity is not actual electricity consumption or AI-specific compute.

Make a chart from this measure

Scope	Value	Meaning and limits
Teraco JB4, South Africa	50 MW of rated critical IT capacity	50 MW is total critical IT capacity after a 30 MW addition. Actual utilisation and AI-specific capacity are not reported. JB4 is a different facility from JB7 in the project pipeline; do not add its incremental 30 MW again.

AWS EC2 G4dn — Cape Town · GPU cloud service

Status: Provider-listed regional service
Rated critical IT capacity: No information in reviewed source
Installed GPU count: No information in reviewed source
GPU hourly rental price: No information in reviewed source
Evidence date: Listed when reviewed on 2026-10-01

AWS lists G4dn in Cape Town. Listing does not verify spare capacity or account access. The reviewed regional listing gives no GPU fleet count, power capacity or hourly price. Inf1 is also listed, but is an inference accelerator and is not counted as a GPU.

Operator evidence · reviewed 2026-10-01

Cassava AI — South Africa · AI compute deployment

Status: Operator-reported deployment and service offering
Rated critical IT capacity: No information in reviewed source
Installed GPU count: No information in reviewed source
GPU hourly rental price: No information in reviewed source
Evidence date: 2026-03-18 operator update

South Africa deployment is stated by the operator. Nigeria, Kenya, Egypt and Morocco remain expansion plans in this source. No MW, installed GPU total or numerical rental rate is published here. Linked to the existing Cassava AI Factory project; not an additional project.

Operator evidence · reviewed 2026-10-01

Teraco JB4 Bredell campus · General data-centre facility

Status: Operator-reported completed expansion
Rated critical IT capacity: 50 MW
Installed GPU count: No information in reviewed source
GPU hourly rental price: No information in reviewed source
Evidence date: 2025-08-12 completion statement

50 MW is total critical IT capacity after a 30 MW addition. Actual utilisation and AI-specific capacity are not reported. JB4 is a different facility from JB7 in the project pipeline; do not add its incremental 30 MW again.

Operator evidence · reviewed 2026-10-01

Research coverage and remaining gaps for this country

No information in reviewed sources means the checked sources supplied no figure. Not yet researched means a country-specific search is still pending. Neither means zero.

Country	Area	Review state	Remaining work
South Africa	AI use and adoption	Survey evidence available	National business-adoption and usage-frequency measures remain uncollected. Pooled employee survey is not allocated to countries.
South Africa	Employment outcomes and hiring demand	Vacancy evidence only	Comparable country-level wages, hires and causal net employment outcomes have not been established in this collection round.
South Africa	Practical access costs	Price evidence available	ITU 2025 baskets checked. Current local device retail prices and verified payment checkout conditions remain uncollected.
South Africa	Operating facilities and compute services	Provider evidence available	This is a starter sample, not a facility census. GPU counts, current rental quotes and latency measurements remain gaps.

Download observations CSV · Compute evidence CSV · Research coverage CSV · Definitions and provenance JSON

Collection method and remaining work · Collection revisions

Explore this data

Compare records, build a chart and download your selection in the Data Explorer.

Open Data Explorer →