Last month, OpenAI CEO Sam Altman visited the construction site of a massive new data center in Texas, a project larger than New York’s Central Park. Backed by $60 billion and equipped with its own natural gas plant, the center is set to become one of the most powerful AI computing hubs in the world. This is just one example of how nations and tech giants are racing to build infrastructure to dominate the next frontier of artificial intelligence.

AI’s Unequal Geography

The world's most advanced AI systems are emerging mainly from regions with high compute power, notably the United States, China, and parts of Europe. These regions control over half of the planet’s largest AI data centers. Meanwhile, more than 150 countries, including nearly all of Africa and much of South America, lack any such infrastructure.

Language models like ChatGPT work better in English and Chinese because these countries not only develop the technology but also control the servers and chips that power it. This imbalance is influencing innovation, talent retention, and even global politics.

The Geopolitics of GPUs

At the heart of this divide are microchips known as GPUs, primarily made by Nvidia. These chips are essential for AI, but they’re expensive and in short supply. The United States and China have used their access to GPUs to exert global influence, sometimes offering or restricting chip access based on diplomatic alignments.

Trade restrictions, like those imposed by the U.S., have prevented countries like Kenya from gaining access to essential chips, even after high-profile diplomatic visits. Meanwhile, China is promoting Huawei's chips across Africa in an attempt to fill the gap.

Struggles from the Ground

Startups in the Global South, such as Kenya’s Qhala and Amini, face major hurdles due to limited local compute power. Without nearby data centers, they must rent server time from other continents, leading to high costs, slower speeds, and complex legal risks.

Nicolás Wolovick, a professor in Argentina, runs one of the country’s most advanced AI hubs from a converted room with aging chips. His students often leave for countries with better resources. “Sometimes I want to cry,” he says. “I need more GPUs.”

Local Solutions and Sovereign AI

To reduce dependency, countries like India, Brazil, and those in Africa are investing in “sovereign AI.” India is building a national language model, and Brazil has pledged $4 billion to AI initiatives. The EU plans to invest €200 billion in AI infrastructure.

In Africa, Cassava Technologies is leading a $500 million project to build data centers, supported by Nvidia and Google. Yet even their most advanced facility is expected to meet only 10–20% of regional demand.

A Future Divided or United?

While companies like Microsoft and OpenAI are making efforts to localize products and expand infrastructure, the computing divide continues to widen. Without urgent global collaboration, AI may not just change the world, it could fracture it even more deeply.