Only 32 countries have the computing power to build advanced AI as Africa hopes to catch up

New research reveals that artificial intelligence is creating a stark global digital divide, with just 32 countries possessing the computing power necessary to build cutting-edge AI systems. The uneven distribution of AI infrastructure is fracturing the world between nations with advanced data centers and those forced to rely on remote access, fundamentally reshaping global power dynamics in the digital age.

The big picture: The European Union leads with 28 AI-capable data centers, followed by the US with 26 and China with 22, collectively housing more than half of the world’s most powerful facilities mapped in the Oxford University study.

• The vast majority of these centers are concentrated in the northern hemisphere, leaving entire continents largely excluded from direct AI development capabilities.
• “We have a computing divide at the heart of the AI revolution,” Lacina Koné, director general of Smart Africa, a continental digital policy group, told the New York Times. “It’s not merely a hardware problem. It’s the sovereignty of our digital future.”

Why costs matter: The astronomical expense of AI infrastructure creates insurmountable barriers for most nations and organizations.

• Graphics processing units (the specialized computer chips needed for AI training) cost tens of thousands of dollars each, with single data centers requiring tens of thousands of these chips.
• Elon Musk’s “Colossus” facility in Tennessee houses over 200,000 AI chips, illustrating the massive scale required for competitive AI development.
• The Trump administration’s proposed $500 billion Stargate AI infrastructure project demonstrates both the prohibitive costs and intense hype surrounding the technology.

Real-world impact: Companies in underserved regions face significant operational challenges that highlight the practical consequences of this divide.

• Qhala, a Kenya-based startup building large language models (AI systems that understand and generate text) for African languages, must schedule work during early morning hours when American programmers are asleep to access faster computing speeds.
• “Proximity is essential,” Qhala founder Shikoh Gitau explained, emphasizing how geographic distance from data centers creates fundamental disadvantages.
• Kate Kallot, former Nvidia executive and founder of Kenyan AI startup Amini, noted: “If you don’t have the resources for compute to process the data and to build your AI models, then you can’t go anywhere.”

Global power dynamics: The concentration of AI computing power is reshaping international influence patterns similar to how oil resources have historically affected geopolitics.

• The world is splitting into two camps: nations that rely on China’s AI infrastructure and those dependent on the United States.
• Both superpowers are positioned to build far more data centers than other countries, further cementing their technological dominance.
• “Oil-producing countries have had an oversized influence on international affairs; in an AI-powered near future, compute producers could have something similar since they control access to a critical resource,” Oxford professor Vili Lehdonvirta explained.

The counterargument: Despite the rush to secure AI capabilities, significant questions remain about the technology’s long-term viability and value.

• Generative AI models continue to hallucinate frequently, pose public safety risks, and show no clear path to profitability.
• Companies may be approaching a “collapse” point as they exhaust available training data for model development.
• Some analysts suggest an AI bubble may be forming, potentially making current massive investments less valuable than anticipated.