Why South Africa, with 88% of the world's platinum, is losing leverage in AI negotiations
South Africa holds 88% of the world's platinum, a metal critical for AI semiconductors, and has the continent's largest data center market. Microsoft and Huawei are competing for control of its infrastructure. Yet the draft policy sets no conditions for investment and does not protect data sovereignty. The minister withdrew the document over AI hallucinations. A new commission is supposed to rework the policy so that South Africa actually uses its leverage.
AI-processed from IEEE Spectrum AI; edited by Hamidun News
South Africa holds 88% of the world's platinum reserves, a metal without which modern semiconductors and AI data centers are impossible. Yet its AI governance policy does not use this critical strategic advantage.
Platinum as leverage
South Africa is an exception among developing countries. Platinum from the Bushveld Complex goes into the semiconductors and data centers that power the world's AI infrastructure. That gives South Africa bargaining power no other African country has.
Add the continent's largest data center market (valued at $2.16 billion) and existing partnerships with Microsoft, Google, AWS and Oracle, and the result is a rare combination: critical resources plus a hyperscaler infrastructure presence. A well-drafted policy could turn this into a negotiating tool.
But the draft AI policy, withdrawn in April, ignored all of these cards. Instead of requirements on data sovereignty and technology transfer, it contained empty "OPTION" placeholders in the most critical places. Instead of minimum conditions for foreign investment, there was nothing. The policy treated South Africa as a consumer of AI, not as a country with influence over global infrastructure.
The price of choosing: Huawei or Microsoft
Against the backdrop of the withdrawn policy, a real fight for control of the continent's infrastructure is unfolding.
Huawei offers a package of the DeepSeek LLM plus the company's cloud infrastructure, with discounts of up to 90% against standard rates. It is attractive, but there is a catch: data sits on infrastructure potentially accessible to Chinese authorities under Chinese law. This has already been documented with Huawei's Safe Cities networks across Africa.
Microsoft has announced $300 million in investment in cloud and AI solutions by 2027. Google, AWS and Oracle already operate cloud regions in the country. That is computing capacity, but with full API dependence on American companies, closed models and terms South Africa did not negotiate.
A third path requires negotiating conditions:
- Minimum investment requirements above $30 million
- Mandatory reporting on the use of computing resources
- Requirements for technology transfer and local development
- National guarantees for data protection and sovereignty
This is the only route to genuine sovereignty, but it demands tough negotiation, not hope for the benevolence of hyperscalers.
A systemic governance failure
In April, the communications minister withdrew the draft after journalists found AI hallucinations in the text: fabricated sources inserted by a neural network. The incident exposed a deep systemic problem: the government adopted a critical document without a verification process and skipped basic source checks before publication.
This is not just a political failure. It signals that the state lacks an "AI assurance" layer, a system for checking which components go into AI solutions before they are deployed. If the government cannot verify its own policy, it is unlikely to be able to verify AI systems when procuring, deploying and regulating them.
A new independent commission is supposed to rework the policy. Its makeup is encouraging: machine learning and data regulation professionals from the University of the Witwatersrand and CSIR. But there is no deadline and no schedule for the revised draft. Meanwhile, South Africa remains without a formal AI policy at exactly the critical moment of a geopolitical choice.
What this means
South Africa is a global test case. If it negotiates terms (data control, technology transfer, capacity reporting) as mandatory requirements for investment, a model emerges for the African continent. If it lets investment proceed on standard commercial terms, it will normalize an extractive model across Africa.
Platinum either becomes an instrument of sovereignty or remains mere raw material. The window is closing.
- Nov 1, 202616 models tested on 7 search tools: Exa averaged 10.5 out of 14, SearXNG about 3
- Nov 1, 2026Google's StreamVibe video agent beats other models by 10.7% on response timing and quality
- Nov 1, 2026OpenAI Rewrites GPT Behavior Rules: Five Values and a Chain of Command
- Nov 1, 2026Voice bots built on LLMs turn to Jev, a fast classifier, for decisions in 200–500 ms
Key Questions Answered
What are the key details on Platinum as leverage?
South Africa is an exception among developing countries. Platinum from the Bushveld Complex goes into the semiconductors and data centers that power the world's AI infrastructure. That gives South Africa bargaining power no other African country has. Add the continent's largest data center market (valued at $2.16 billion) and existing partnerships with Microsoft, Google, AWS and Oracle, and the result is a rare combination: critical resources plus a hyperscaler infrastructure presence. A well-drafted policy could turn this into a negotiating tool. But the draft AI policy, w
What are the key details on The price of choosing: Huawei or Microsoft?
Against the backdrop of the withdrawn policy, a real fight for control of the continent's infrastructure is unfolding. Huawei offers a package of the DeepSeek LLM plus the company's cloud infrastructure, with discounts of up to 90% against standard rates. It is attractive, but there is a catch: data sits on infrastructure potentially accessible to Chinese authorities under Chinese law. This has already been documented with Huawei's Safe Cities networks across Africa. Microsoft has announced $300 million in investment in cloud and AI solutions by 2027. Google, AWS and Oracle
What are the key details on A systemic governance failure?
In April, the communications minister withdrew the draft after journalists found AI hallucinations in the text: fabricated sources inserted by a neural network. The incident exposed a deep systemic problem: the government adopted a critical document without a verification process and skipped basic source checks before publication. This is not just a political failure. It signals that the state lacks an "AI assurance" layer, a system for checking which components go into AI solutions before they are deployed. If the government cannot verify its own policy, it is unlikely to
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.