AI infrastructure is concentrated in just a handful of countries, creating a widening digital divide that affects science, the economy, and geopolitical dependencies—and could further accelerate global inequality.
A new study from the University of Oxford shows that only 32 countries have specialized AI data centers. More than 150 nations have virtually no access to the "compute" power required for training modern AI models. Researchers Vili Lehdonvirta, Zoe Jay Hawkins, and Boxi Wu identified the United States, China, and the European Union as the main beneficiaries: US companies such as Amazon, Google, and Microsoft operate 87 AI hubs worldwide, Chinese providers run 39, while Europe lags behind with just six.
The Oxford team systematically mapped the global distribution of AI infrastructure by analyzing customer websites of nine leading cloud providers. The findings are stark: the United States dominates by a huge margin, while Africa and South America are almost entirely left out of AI development. Most of the chips used in these data centers are supplied by Nvidia—the US company whose GPUs have become foundational to the AI boom.
The New York Times highlights the gap with real-world examples: OpenAI CEO Sam Altman recently visited a $60 billion project in Texas, part of the Stargate initiative, while computer science professor Nicolás Wolovick at the University of Córdoba in Argentina conducts AI research in a converted classroom using outdated chips. In Kenya, startups like Qhala and Amini face a shortage of local computing resources and resort to working at night to rent overseas capacity at lower costs.
Most local needs remain unmet
The United States and China are leveraging their technological lead as a tool of geopolitical influence. Washington controls global access to high-performance chips through export restrictions, while Beijing offers state-backed loans to promote its own hardware. In the United Arab Emirates, Chinese technology has been excluded in exchange for access to Nvidia and Microsoft products. In Africa, companies like Huawei are trying to retrofit existing data centers with Chinese chips, according to the Times. While China is still behind Nvidia technologically, Lacina Koné of Smart Africa views this as a practical workaround, noting that Africa is open to deals with any supplier able to deliver GPUs.
Co-author Lehdonvirta warns that countries with computing power could wield influence comparable to oil producers. Nvidia's costly GPUs are hard to obtain, forcing many nations to rent compute resources from distant data centers. In response, several countries are now investing billions to build their own AI infrastructure.
Governments in India, Brazil, and the EU are funding local data centers and AI models. In Africa, Cassava Technologies plans to build five data centers with support from Nvidia and Google. Yet, by Cassava's own estimate, these efforts will only meet a fraction of the continent’s needs. The Oxford researchers caution that without broader access to compute resources, the gap is likely to widen further. As Koné puts it, compute is becoming the foundation of digital sovereignty.
