Machinepower

Machinepower Index / Dossiers / South Africa

Rank 38 of 50 · Tier 3

Rank 36–48 across the scoring rules tested. How this is tested

South Africa

Diffusion Advantage — For nations that cannot build the frontier but can use it faster than anyone: import the models and win on deployment, across the state and the economy, with public consent.

Score3.92
Potential4.87
Bottleneck cost−0.94
Weakest cell: Hardware & Silicon Access at 2.0 Binding layer: Watts Since the October fix: No change logged Edition: Q4 2026 · updated October 2026

How the Score, Potential and bottleneck cost are worked out: methodology.

The three layers

WattsMaterial power · holds the weakest cell
4.0 / 10

Energy, grids, data centres and silicon — the physical capacity to run models at scale.

Epoch AI records no large AI clusters or data centres here yet.

WeightsIntelligence power
4.7 / 10

Models, alignment capability, standards influence and the talent that produces them.

WillPolitical power
5.9 / 10

The capacity to wire AI into the state and the economy while holding public consent.

Where the score comes from

LayerSub-dimensionBasisScore
WattsPlanning & PermittingAssessed4.2
Firm Power CapacityDataset5.2
Sovereign ComputeDataset4.6
Hardware & Silicon AccessDataset2.0
WeightsFrontier Model OutputAssessed3.0
Alignment CapabilityAssessed3.5
Regulatory StandingDataset7.7
Talent DensityAssessed4.6
WillPolicy CapacityDataset4.3
Public Sector AdoptionDataset6.5
Workforce TransitionDataset5.4
Public TrustDataset7.2

Assessed marks a measurement we judged rather than took from a dataset: 4 of the 12 here.

Sources

Ipsos AI Monitor 2026April 2026
Public Trust in AI
66% say AI products have more benefits than drawbacks

Share of adults agreeing that products and services using AI have more benefits than drawbacks; the Index scores public trust as 1 + share ÷ 10. Survey of 23,532 adults across 32 countries, March to April 2026.

View source →
Lelapa AI2025
Frontier models (assessed)
3.0/10, assessed

South Africa's most visible domestic model is Lelapa AI's InkubaLM, a small open model for African languages; there is no domestic large-scale model development.

View source →
TOP500.orgJune 2026
High-Performance Compute Capacity
Cassava AI Factory (Cape Town) — 77.8 PF (#36)

The Cassava AI Factory at Africa Data Centres CPT1 put the African continent onto the TOP500 at scale on the June 2026 list.

View source →
Oxford Insights GAIRI 2025January 2026
Government AI Readiness
Governance pillar 76.58 · PSA 64.93 (rank 65)

The validated January 2026 report rates South African governance capacity at 76.58/100 — far above what its infrastructure permits it to deploy.

View source →
IMF2023 data, published June 2024
AI Preparedness Index: human capital and labour market policies
0.120 of a possible 0.25 · 5.4/10 in the Index

Used for the workforce measurement. The IMF combines schooling, pay, labour mobility and active labour-market policies; the Index rescales it so the weakest of 173 economies scores 0 and the strongest 10.

View source →
Ember2025
Firm generating capacity
54 GW (2025)

Fossil, nuclear, hydro and other dispatchable plant; averaged with generation for firm power.

View source →
University of Melbourne; KPMGJanuary 2025
Trust in AI at work
64% trust AI at work

Second source for public trust, averaged with Ipsos.

View source →
Oxford InsightsMay 2026
Policy capacity
43/100

Feeds policy capacity.

View source →
Public First, AI Global 20262026
Consent to build data centres
57% want more built, 15% fewer

Half of public trust: support for building more data centres, scored as 1 + share ÷ 10 and averaged with trust in AI.

View source →

Bottlenecks

These hold South Africa below its Potential. Its two lowest scores cost it the most: hardware & silicon access at 2.0 (part of Watts) and frontier model output at 3.0 (part of Weights).

Watts (Material)Hardware & Silicon Access

Assessment

Africa's AI gateway on Africa's most unreliable grid. The Cassava AI Factory in Cape Town put the continent onto the TOP500 at scale for the first time, and Oxford rates South African governance capacity far above its infrastructure. But load-shedding is a sovereignty tax on everything: planning at 4.2 and hardware at 2.0 hold a capable state to the floor of the Index.

The art of the possible

The Cassava build proves the demand; the grid is the bottleneck with a name. The play is to ring-fence AI-grade power — renewables and battery islands around data-centre zones — and make Cape Town the continent's inference hub before anyone else claims it.

Strategic play — Diffusion Advantage

Ring-fence AI-grade power before anything else, with renewables and battery islands around data-centre zones, so the Cassava build becomes the continent's inference hub in Cape Town before anyone else claims it, and treat the platinum-group metals beneath it as the position it trades from.

Leverage — the nearest position. No chokepoint in the chain. The nearest position sits beneath it: platinum-group metals and manganese, minerals that electrification keeps repricing. Further from the chip than most — but a position, not an ambition.

Open South Africa in the live Index → — every source behind every number, and the comparison tool.